1
00:00:00,000 --> 00:00:01,670


2
00:00:01,670 --> 00:00:02,890
URS GASSER: All right.

3
00:00:02,890 --> 00:00:04,690
Hello, everyone and welcome.

4
00:00:04,690 --> 00:00:06,450
How are you?

5
00:00:06,450 --> 00:00:07,140
Let's try again.

6
00:00:07,140 --> 00:00:08,490
How are you this morning?

7
00:00:08,490 --> 00:00:10,765
[APPLAUSE]

8
00:00:10,765 --> 00:00:15,060


9
00:00:15,060 --> 00:00:17,970
A very warm welcome to
the Berkman Klein Center,

10
00:00:17,970 --> 00:00:20,790
to Harvard Law
School on this sunny

11
00:00:20,790 --> 00:00:23,790
but windy and
slightly chilly day.

12
00:00:23,790 --> 00:00:26,100
We're really delighted
to have you here.

13
00:00:26,100 --> 00:00:27,330
My name is Urs Gasser.

14
00:00:27,330 --> 00:00:29,040
I serve as the
executive director

15
00:00:29,040 --> 00:00:30,810
of the Berkman Klein Center.

16
00:00:30,810 --> 00:00:32,340
I'm also on the
Harvard Law School

17
00:00:32,340 --> 00:00:36,180
faculty with my colleague, that
I will introduce just in a bit.

18
00:00:36,180 --> 00:00:39,960
And of course, I am
very, very pleased

19
00:00:39,960 --> 00:00:44,970
to moderate this special
conversation about the ethics

20
00:00:44,970 --> 00:00:47,290
of digital transformation.

21
00:00:47,290 --> 00:00:50,490
And I'm of course, even
more pleased and actually

22
00:00:50,490 --> 00:00:54,930
honored to welcome the
federal president of Germany,

23
00:00:54,930 --> 00:00:56,590
Frank-Walter Steinmeier.

24
00:00:56,590 --> 00:00:57,090
[APPLAUSE]

25
00:00:57,090 --> 00:00:58,382
FRANK-WALTER STEINMEIER: Hello.

26
00:00:58,382 --> 00:01:06,350


27
00:01:06,350 --> 00:01:08,020
URS GASSER: [SPEAKING GERMAN]

28
00:01:08,020 --> 00:01:08,945


29
00:01:08,945 --> 00:01:10,733
FRANK-WALTER
STEINMEIER: [INAUDIBLE]..

30
00:01:10,733 --> 00:01:12,650
URS GASSER: We are joined
by a wonderful group

31
00:01:12,650 --> 00:01:14,480
of colleagues and experts.

32
00:01:14,480 --> 00:01:18,210
And I will-- if you
are not mad with me,

33
00:01:18,210 --> 00:01:20,280
I will just briefly
introduce you,

34
00:01:20,280 --> 00:01:22,610
and we will get to know each
other a bit better as we

35
00:01:22,610 --> 00:01:26,870
go along and talk about
your work, and of course

36
00:01:26,870 --> 00:01:29,180
as we engage in the
opening conversations.

37
00:01:29,180 --> 00:01:32,510
So this should be interactive
as much as we can.

38
00:01:32,510 --> 00:01:38,000
Eva Weber-Gurska is a ethicist,
a philosopher currently

39
00:01:38,000 --> 00:01:40,970
at Ruhr University in Bochum.

40
00:01:40,970 --> 00:01:45,200
She's doing amazing work, and
I'm already looking forward

41
00:01:45,200 --> 00:01:47,510
to learning from you today.

42
00:01:47,510 --> 00:01:49,610
Quite often, these
debates about ethics

43
00:01:49,610 --> 00:01:53,050
happen without having
philosophers around us.

44
00:01:53,050 --> 00:01:55,250
I'm grateful that you're here.

45
00:01:55,250 --> 00:02:00,260
Matthew Liao is at NYU, and
is a professor of bioethics.

46
00:02:00,260 --> 00:02:03,220
He also runs a center
on the same topic,

47
00:02:03,220 --> 00:02:07,280
and I'm particularly
curious to hear also

48
00:02:07,280 --> 00:02:09,620
some of the lessons
learned from past cycles

49
00:02:09,620 --> 00:02:11,720
of technological
innovation, as we now

50
00:02:11,720 --> 00:02:16,360
talk about digital things
and AI, IT, and the like.

51
00:02:16,360 --> 00:02:18,650
Jeanette Hofmann, welcome back.

52
00:02:18,650 --> 00:02:20,210
Great to have you here.

53
00:02:20,210 --> 00:02:25,190
Jeanette is a professor of
internet policy at the Free

54
00:02:25,190 --> 00:02:27,680
University in Berlin.

55
00:02:27,680 --> 00:02:32,120
She's also the director of
the Alexander von Humboldt

56
00:02:32,120 --> 00:02:36,412
Institute for Internet
and Society in Berlin.

57
00:02:36,412 --> 00:02:37,995
I've introduced the
president already,

58
00:02:37,995 --> 00:02:41,490
and he doesn't need
an introduction.

59
00:02:41,490 --> 00:02:44,740
So next we have
Dean Melissa Nobles,

60
00:02:44,740 --> 00:02:47,780
who is the Dean of the
School of Humanities, Arts,

61
00:02:47,780 --> 00:02:50,930
and Social Sciences at MIT.

62
00:02:50,930 --> 00:02:55,990
Really great to have you here,
as we will hear more about MIT

63
00:02:55,990 --> 00:02:58,210
is building a new
school of computing.

64
00:02:58,210 --> 00:02:59,210
MELISSA NOBLES: Correct.

65
00:02:59,210 --> 00:03:00,380
URS GASSER: The
Schwarzman College.

66
00:03:00,380 --> 00:03:00,890
MELISSA NOBLES: That's right.

67
00:03:00,890 --> 00:03:02,473
URS GASSER: Lots of
interesting things

68
00:03:02,473 --> 00:03:03,950
happening there at
the intersection

69
00:03:03,950 --> 00:03:06,550
of engineering and ethics.

70
00:03:06,550 --> 00:03:10,340
So we're looking forward to your
thoughts on this conversation.

71
00:03:10,340 --> 00:03:12,170
Wolfgang Schulz,
professor for media

72
00:03:12,170 --> 00:03:14,900
and public law at the
University of Hamburg.

73
00:03:14,900 --> 00:03:17,270
He's also the director--
and now I have to read that,

74
00:03:17,270 --> 00:03:19,610
because I still
cannot remember it--

75
00:03:19,610 --> 00:03:23,660
the director of the Leibnitz
Institute for Media Research,

76
00:03:23,660 --> 00:03:26,540
which is known to me as
the Hans-Bredow-Institut,

77
00:03:26,540 --> 00:03:29,360
but I learned that it's
important to emphasize

78
00:03:29,360 --> 00:03:30,800
the Leibnitz part.

79
00:03:30,800 --> 00:03:34,340
Crystal Yang, really great
faculty colleague here,

80
00:03:34,340 --> 00:03:37,460
professor of law at
Harvard Law School.

81
00:03:37,460 --> 00:03:41,510
Does wonderful work, important
work on criminal justice

82
00:03:41,510 --> 00:03:44,160
and the use of algorithms
and data in that area.

83
00:03:44,160 --> 00:03:45,750
We'll talk more about that.

84
00:03:45,750 --> 00:03:48,950
So as you can see,
a fantastic lineup.

85
00:03:48,950 --> 00:03:53,300
And of course, I'm so grateful
to you, Mr. President,

86
00:03:53,300 --> 00:03:56,390
that you joined this
group as a participant,

87
00:03:56,390 --> 00:03:59,540
and I get a sense already
that you're ready to jump in

88
00:03:59,540 --> 00:04:01,670
and will take over
the moderation

89
00:04:01,670 --> 00:04:04,250
function in due course,
which is totally fine

90
00:04:04,250 --> 00:04:08,520
and will make my job easier.

91
00:04:08,520 --> 00:04:10,790
So one or two logistical notes.

92
00:04:10,790 --> 00:04:14,750
First, we will end
at roughly 11:30.

93
00:04:14,750 --> 00:04:15,560
That's the plan.

94
00:04:15,560 --> 00:04:19,190
Some seconds of the
conversation may be in German.

95
00:04:19,190 --> 00:04:22,190
You should have translation.

96
00:04:22,190 --> 00:04:24,620
If it's OK with
you, I will continue

97
00:04:24,620 --> 00:04:26,840
to moderate in
English, and I think

98
00:04:26,840 --> 00:04:29,570
the reason is straightforward,
because my Swiss accent is

99
00:04:29,570 --> 00:04:32,330
so strong when I speak
German, that it's

100
00:04:32,330 --> 00:04:36,280
easier for the Germans
when I talk English, so

101
00:04:36,280 --> 00:04:38,270
just to make that [? clear. ?]

102
00:04:38,270 --> 00:04:40,530
[APPLAUSE]

103
00:04:40,530 --> 00:04:43,480


104
00:04:43,480 --> 00:04:47,110
So with that, Herr
Bundespresident,

105
00:04:47,110 --> 00:04:50,010
here's the question for
you to start us off.

106
00:04:50,010 --> 00:04:55,890
We met the last time in 2012 in
Berlin, and had a conversation

107
00:04:55,890 --> 00:04:59,250
about what does it mean
to make good policies

108
00:04:59,250 --> 00:05:01,650
for the internet age?

109
00:05:01,650 --> 00:05:05,120
And I googled this
morning actually,

110
00:05:05,120 --> 00:05:08,610
and tried to remember what
happened in 2012, right?

111
00:05:08,610 --> 00:05:12,300
It seems like in internet
times it's more like 70 years

112
00:05:12,300 --> 00:05:13,342
ago than seven years ago.

113
00:05:13,342 --> 00:05:14,592
FRANK-WALTER STEINMEIER: Yeah.

114
00:05:14,592 --> 00:05:17,070
URS GASSER: And when I googled
things that happened there

115
00:05:17,070 --> 00:05:21,090
in the technology space was
the Google Glass project

116
00:05:21,090 --> 00:05:22,710
was kicked off.

117
00:05:22,710 --> 00:05:25,200
The iPad Mini was introduced.

118
00:05:25,200 --> 00:05:27,590
Facebook went public.

119
00:05:27,590 --> 00:05:30,360
And a bill was
signed in California

120
00:05:30,360 --> 00:05:35,400
that self-driving cars are
now allowed and regulated.

121
00:05:35,400 --> 00:05:38,640
So I'm wondering, that seems
to be a very different stage

122
00:05:38,640 --> 00:05:41,130
in our digital
transformation process,

123
00:05:41,130 --> 00:05:43,600
if you look back only a few
years and now fast forward

124
00:05:43,600 --> 00:05:45,450
to 2019.

125
00:05:45,450 --> 00:05:46,590
Where have we arrived?

126
00:05:46,590 --> 00:05:48,610
What are you thinking about?

127
00:05:48,610 --> 00:05:50,080
What are your concerns?

128
00:05:50,080 --> 00:05:53,010
What are your hopes and
how does that connect

129
00:05:53,010 --> 00:05:54,918
with the topic of today?

130
00:05:54,918 --> 00:05:56,120
Mr. President.

131
00:05:56,120 --> 00:05:57,870
FRANK-WALTER STEINMEIER:
[SPEAKING GERMAN]

132
00:05:57,870 --> 00:05:59,745
INTERPRETER: Well, thank
you very much indeed

133
00:05:59,745 --> 00:06:01,680
for these kind words of welcome.

134
00:06:01,680 --> 00:06:03,660
Ladies and gentlemen,
dear students,

135
00:06:03,660 --> 00:06:05,880
I think it is fantastic.

136
00:06:05,880 --> 00:06:08,130
You know, you've been
given the alternative

137
00:06:08,130 --> 00:06:13,330
to enjoy a sunshiny day, though
somewhat windy, late autumn,

138
00:06:13,330 --> 00:06:15,690
but you've nevertheless
decided in favor

139
00:06:15,690 --> 00:06:18,420
of the alternative-- coming
here into an enclosed room

140
00:06:18,420 --> 00:06:19,170
to listen to us.

141
00:06:19,170 --> 00:06:20,490
Thank you very much for that.

142
00:06:20,490 --> 00:06:24,030
And thank you also for
reminding me of the year 2012,

143
00:06:24,030 --> 00:06:27,060
which I remember well for
quite different reasons,

144
00:06:27,060 --> 00:06:29,050
and I'll get back
to that in a second.

145
00:06:29,050 --> 00:06:32,580
You know, my visit to
Boston in [? 1220. ?]

146
00:06:32,580 --> 00:06:36,300
But allow me to begin by
saying that I haven't been here

147
00:06:36,300 --> 00:06:38,220
for the very first
time, and I'm always

148
00:06:38,220 --> 00:06:43,920
happy to be back in this
academic, scientific center,

149
00:06:43,920 --> 00:06:47,130
a center not only with regard
to the United States of America

150
00:06:47,130 --> 00:06:50,970
but also in a much
broader sense, because it

151
00:06:50,970 --> 00:06:55,470
is a center that is exemplary
in bringing together researchers

152
00:06:55,470 --> 00:06:58,290
and academics from all
parts of the world,

153
00:06:58,290 --> 00:07:01,710
from all countries of the
world, to make them work

154
00:07:01,710 --> 00:07:03,930
on subjects of common concern.

155
00:07:03,930 --> 00:07:08,040
And when I remember that visit
in 2012, or earlier visits

156
00:07:08,040 --> 00:07:09,810
too and later visits,
you know, no matter

157
00:07:09,810 --> 00:07:13,310
whether we talked about foreign
policy issues or other issues,

158
00:07:13,310 --> 00:07:15,240
either here in the
hall or in other places

159
00:07:15,240 --> 00:07:18,420
in Harvard, whether we
talked about questions to do

160
00:07:18,420 --> 00:07:21,720
with climate policy,
about the state of affairs

161
00:07:21,720 --> 00:07:24,360
of transatlantic
relations, rest assured,

162
00:07:24,360 --> 00:07:28,740
every time that I came
here, I returned home

163
00:07:28,740 --> 00:07:31,230
having benefited
to a large extent

164
00:07:31,230 --> 00:07:34,480
from the discussions
I had at Harvard.

165
00:07:34,480 --> 00:07:38,550
There was one exception, and
that brings me back to 2012,

166
00:07:38,550 --> 00:07:39,060
really.

167
00:07:39,060 --> 00:07:46,980
Just once did I risk
myself, my whole existence

168
00:07:46,980 --> 00:07:48,810
here in Harvard,
because I allowed myself

169
00:07:48,810 --> 00:07:53,785
to be talked into throwing
the first pitch in a baseball

170
00:07:53,785 --> 00:07:54,285
match.

171
00:07:54,285 --> 00:07:57,930


172
00:07:57,930 --> 00:08:00,480
And I was extremely naive.

173
00:08:00,480 --> 00:08:04,410
Never ever had I before I
attended a baseball match,

174
00:08:04,410 --> 00:08:09,300
held a baseball in my hands,
nor been in a baseball stadium.

175
00:08:09,300 --> 00:08:13,720
And my then colleague, secretary
of state Condoleezza Rice,

176
00:08:13,720 --> 00:08:16,950
we all remember her,
was aghast when she

177
00:08:16,950 --> 00:08:19,720
heard what I was about to do.

178
00:08:19,720 --> 00:08:23,310
And the only comment she gave
to me was, "don't do it."

179
00:08:23,310 --> 00:08:26,610
But you know, that is
typical of us Germans.

180
00:08:26,610 --> 00:08:30,730
I had to accepted and I didn't
want to go back on my promise.

181
00:08:30,730 --> 00:08:38,460
So once we entered the stadium
on the afternoon of that day,

182
00:08:38,460 --> 00:08:43,320
I got an inkling of why
my colleague Condoleezza

183
00:08:43,320 --> 00:08:46,800
Rice was so aghast, because
the stadium was filled

184
00:08:46,800 --> 00:08:51,930
to the very last seat, 40,000 or
50,000 people in the audience.

185
00:08:51,930 --> 00:08:55,470
And I had a certain feeling
that they hadn't come just

186
00:08:55,470 --> 00:08:59,010
because of me, but
they'd come because they

187
00:08:59,010 --> 00:09:02,340
wanted to watch the
match, a match that

188
00:09:02,340 --> 00:09:04,440
was the match really.

189
00:09:04,440 --> 00:09:06,540
Here in the United
States, the Red Sox

190
00:09:06,540 --> 00:09:10,300
were playing the
New York Yankees.

191
00:09:10,300 --> 00:09:13,560
And I realized all of a
sudden that this is not just

192
00:09:13,560 --> 00:09:14,280
any match.

193
00:09:14,280 --> 00:09:18,690
It's about religious
issues, really.

194
00:09:18,690 --> 00:09:22,260
Still, it worked out somehow.

195
00:09:22,260 --> 00:09:24,000
I survived it.

196
00:09:24,000 --> 00:09:25,650
And having survived
that experience,

197
00:09:25,650 --> 00:09:27,600
I was happy to come
back every time.

198
00:09:27,600 --> 00:09:30,920
I wasn't shy of returning
to Boston, to Harvard.

199
00:09:30,920 --> 00:09:32,670
But today, it's a
different topic, really,

200
00:09:32,670 --> 00:09:36,090
that brings me here, different
from the topics we focused on

201
00:09:36,090 --> 00:09:38,810
in the past years.

202
00:09:38,810 --> 00:09:42,240
We're no longer on the
threshold of the digitization,

203
00:09:42,240 --> 00:09:44,360
of digitization,
of the digital age.

204
00:09:44,360 --> 00:09:47,340
But we have already
entered that age.

205
00:09:47,340 --> 00:09:53,880
I've come here because the
topic we will be talking about

206
00:09:53,880 --> 00:09:55,740
directly refers
back to topics I'm

207
00:09:55,740 --> 00:09:58,260
focusing on in my
presidency, the future

208
00:09:58,260 --> 00:10:00,060
of liberal democracy, that is.

209
00:10:00,060 --> 00:10:05,300
How does the internet, how do
Facebook, Twitter, algorithms,

210
00:10:05,300 --> 00:10:10,580
anonymity in the internet,
how do all these things change

211
00:10:10,580 --> 00:10:13,490
the democratic culture
of debate, which

212
00:10:13,490 --> 00:10:16,220
is of such great importance
to us in Germany,

213
00:10:16,220 --> 00:10:20,030
just as much as you do in
the United States of America?

214
00:10:20,030 --> 00:10:23,630
Despite the daily
waves of outrage

215
00:10:23,630 --> 00:10:26,540
that you have to
live with, how can we

216
00:10:26,540 --> 00:10:30,200
make sure that we keep
a general overview?

217
00:10:30,200 --> 00:10:33,320
How can we distinguish
what is important from what

218
00:10:33,320 --> 00:10:35,360
is unimportant?

219
00:10:35,360 --> 00:10:38,390
And does this
culture of thinking

220
00:10:38,390 --> 00:10:42,110
in simple opposites, yes
or no, black or white,

221
00:10:42,110 --> 00:10:44,810
harsh approaches, whether
that takes away from us

222
00:10:44,810 --> 00:10:51,140
our ability to see the nuances
between black and white?

223
00:10:51,140 --> 00:10:52,910
Are we capable of doing that?

224
00:10:52,910 --> 00:10:57,050
Do we continue to be capable of
entering into compromise, which

225
00:10:57,050 --> 00:11:00,530
I believe to be vital
for any democracy,

226
00:11:00,530 --> 00:11:04,640
if we no longer have the time to
differentiate or to see things

227
00:11:04,640 --> 00:11:07,700
in nuances in carefully
weighing the pros and cons,

228
00:11:07,700 --> 00:11:10,670
because it's no longer popular?

229
00:11:10,670 --> 00:11:12,620
We talked about this
yesterday in Boston

230
00:11:12,620 --> 00:11:17,180
with American and German
academics in great depth.

231
00:11:17,180 --> 00:11:20,750
Today, though, we
are again talking

232
00:11:20,750 --> 00:11:23,240
about digital
transformation, how

233
00:11:23,240 --> 00:11:26,960
that has changed our lives
and daily experiences.

234
00:11:26,960 --> 00:11:29,480
But as Mr. Gasser
kindly indicated,

235
00:11:29,480 --> 00:11:32,500
we will be focusing on
a different priority.

236
00:11:32,500 --> 00:11:34,250
We're not really
focusing on the question,

237
00:11:34,250 --> 00:11:36,650
whether we need
digital technologies.

238
00:11:36,650 --> 00:11:38,180
They're there, anyway.

239
00:11:38,180 --> 00:11:40,400
No one is denying the
fact that they open up

240
00:11:40,400 --> 00:11:43,100
enormous opportunities
for all of us, when

241
00:11:43,100 --> 00:11:45,100
it comes to fighting
poverty, for example,

242
00:11:45,100 --> 00:11:49,070
when it comes to tackling the
impact of climate change, when

243
00:11:49,070 --> 00:11:53,360
it comes to combating
diseases and their effects.

244
00:11:53,360 --> 00:11:55,340
Undoubtedly,
Germany is a country

245
00:11:55,340 --> 00:11:59,480
that has no resources
over its own,

246
00:11:59,480 --> 00:12:02,900
outside the humanities
and human resources.

247
00:12:02,900 --> 00:12:07,130
We want to be a country that
has technology to offer,

248
00:12:07,130 --> 00:12:11,180
and we want to participate in
the developments they entail.

249
00:12:11,180 --> 00:12:14,450
And that is a kind of
introductory remark on my part.

250
00:12:14,450 --> 00:12:15,860
As regards to the
topic we intend

251
00:12:15,860 --> 00:12:20,120
to talk about today the
ethic-- a code of ethics

252
00:12:20,120 --> 00:12:22,250
for the digital
transformation, I

253
00:12:22,250 --> 00:12:27,730
would like to just
briefly focus on

254
00:12:27,730 --> 00:12:30,380
why this topic is
so important to me.

255
00:12:30,380 --> 00:12:33,260
I actually have come
from two visits,

256
00:12:33,260 --> 00:12:35,270
and I refer back
to those visits.

257
00:12:35,270 --> 00:12:39,110
I visited Stanford last
year, focusing again

258
00:12:39,110 --> 00:12:40,700
on the future of digitization.

259
00:12:40,700 --> 00:12:43,730
When we traveled there, a few
days before we left we read

260
00:12:43,730 --> 00:12:50,660
in the papers that Elon Musk
had bought up a company that was

261
00:12:50,660 --> 00:12:53,600
engaging in the research
in brain implants,

262
00:12:53,600 --> 00:12:59,460
and that was doing very
well in that regard,

263
00:12:59,460 --> 00:13:03,380
that this might help
tackle diseases like

264
00:13:03,380 --> 00:13:08,180
Parkinson's and Alzheimer's.

265
00:13:08,180 --> 00:13:11,070
I learned a lot about the
imagination of researchers

266
00:13:11,070 --> 00:13:13,640
during my encounters
there, how one

267
00:13:13,640 --> 00:13:18,740
can influence brain
activities with the help

268
00:13:18,740 --> 00:13:22,190
of implants and algorithms.

269
00:13:22,190 --> 00:13:26,660
This has undeniable and
obvious consequences.

270
00:13:26,660 --> 00:13:29,240
But at the end of the
discussion, it wasn't I,

271
00:13:29,240 --> 00:13:31,620
but it was someone who is
very well known in the United

272
00:13:31,620 --> 00:13:33,110
States, George Shultz.

273
00:13:33,110 --> 00:13:38,480
That is the former Secretary of
State under Ronald Reagan, who

274
00:13:38,480 --> 00:13:42,410
also is or was at the time a
member of the board of Stanford

275
00:13:42,410 --> 00:13:43,190
University.

276
00:13:43,190 --> 00:13:45,950
He said at the end of the
discussion, guys, really,

277
00:13:45,950 --> 00:13:50,880
I am fascinated by the scenario
you have been painting,

278
00:13:50,880 --> 00:13:51,900
drawing of the future.

279
00:13:51,900 --> 00:13:56,870
But let's not forget, we
are living in a democracy.

280
00:13:56,870 --> 00:14:03,030
And democracy relies on
independent, self-determined,

281
00:14:03,030 --> 00:14:07,617
confident human beings
if it is to survive

282
00:14:07,617 --> 00:14:10,200
And he addressed himself to the
researchers and the academics.

283
00:14:10,200 --> 00:14:12,710
So when developing
these technologies,

284
00:14:12,710 --> 00:14:17,870
don't forget to think of the
consequences of your inventions

285
00:14:17,870 --> 00:14:20,975
and how that fits into
democracy and its principles.

286
00:14:20,975 --> 00:14:23,030
In my second trip,
and I'm going to be

287
00:14:23,030 --> 00:14:26,870
brief about this, with
my visit to China, again,

288
00:14:26,870 --> 00:14:29,460
we also [? focused ?]
on this topic.

289
00:14:29,460 --> 00:14:34,670
And we also talked about social
scoring, the opportunities,

290
00:14:34,670 --> 00:14:39,390
the perspectives that result
for the members of a society.

291
00:14:39,390 --> 00:14:44,190
The debates we had were not
easy, because at the beginning

292
00:14:44,190 --> 00:14:45,680
the Chinese didn't
understand why

293
00:14:45,680 --> 00:14:47,570
we were asking these
questions at all,

294
00:14:47,570 --> 00:14:50,690
and why we would find some
of these things complicated

295
00:14:50,690 --> 00:14:53,000
that come up in the
context of social scoring.

296
00:14:53,000 --> 00:14:56,192
Because they said, we have 80,
90% of support, popular support

297
00:14:56,192 --> 00:14:56,900
for these topics.

298
00:14:56,900 --> 00:14:59,950
Why are you against it?

299
00:14:59,950 --> 00:15:06,820
You know, we live under
different political

300
00:15:06,820 --> 00:15:08,770
circumstances, are
scared and shocked

301
00:15:08,770 --> 00:15:12,670
by the idea of having to
submit to a total surveillance

302
00:15:12,670 --> 00:15:19,930
of all aspects of our lives,
that no matter what we do,

303
00:15:19,930 --> 00:15:22,690
this might be linked
up to a system

304
00:15:22,690 --> 00:15:26,620
that assesses our performance
in a negative or a positive way,

305
00:15:26,620 --> 00:15:29,470
and that this, of course, has an
effect on the way we developed

306
00:15:29,470 --> 00:15:31,780
as human beings,
that hopes, wishes,

307
00:15:31,780 --> 00:15:34,300
and dreams are
becoming externalized,

308
00:15:34,300 --> 00:15:36,760
that they are stored
on a software I no

309
00:15:36,760 --> 00:15:39,910
longer have any influence over.

310
00:15:39,910 --> 00:15:45,430
For our concept of
individual responsibility

311
00:15:45,430 --> 00:15:49,660
and of personal freedom is
being called into question

312
00:15:49,660 --> 00:15:52,480
by such an approach.

313
00:15:52,480 --> 00:15:55,870
We, however, know that
this is not a problem that

314
00:15:55,870 --> 00:15:57,770
is exclusive to the Chinese.

315
00:15:57,770 --> 00:16:01,660
German companies,
American companies

316
00:16:01,660 --> 00:16:03,460
that invest in
the United States,

317
00:16:03,460 --> 00:16:05,930
that employ people
in the United States,

318
00:16:05,930 --> 00:16:09,040
they will be working under
the very same conditions.

319
00:16:09,040 --> 00:16:10,720
And thus we have
to have an interest

320
00:16:10,720 --> 00:16:12,480
in what is happening here.

321
00:16:12,480 --> 00:16:15,520
But let me close by saying
that the debate in China

322
00:16:15,520 --> 00:16:18,180
hasn't yet come to an end yet.

323
00:16:18,180 --> 00:16:19,310
It is still ongoing.

324
00:16:19,310 --> 00:16:24,700
We don't know what will be
the outcome, the result of all

325
00:16:24,700 --> 00:16:27,588
those tests and
experiments that are being

326
00:16:27,588 --> 00:16:28,880
carried out in China right now.

327
00:16:28,880 --> 00:16:31,510
But the obvious question
is on the table.

328
00:16:31,510 --> 00:16:34,090
Is there something
like a minimum

329
00:16:34,090 --> 00:16:38,440
of morals for the digital age?

330
00:16:38,440 --> 00:16:42,070
Shouldn't we work to have
something like that like, like

331
00:16:42,070 --> 00:16:46,550
a common expression
of the limits

332
00:16:46,550 --> 00:16:52,570
of the digital future in the
decades or centuries to come?

333
00:16:52,570 --> 00:16:57,190
Which brings me down to the
question of whether we do not

334
00:16:57,190 --> 00:17:01,090
really need a much more
intensive exchange of thoughts

335
00:17:01,090 --> 00:17:06,609
between the tech community,
the political scientists

336
00:17:06,609 --> 00:17:09,970
about the philosophy
of the individual than

337
00:17:09,970 --> 00:17:14,650
is happening at this point in
time, at least as I see it.

338
00:17:14,650 --> 00:17:18,190
Well, you know, if I were to
choose, I would very much like

339
00:17:18,190 --> 00:17:21,849
to be in a position where
I could leave Boston today,

340
00:17:21,849 --> 00:17:24,280
having received the
confirmation from all of you

341
00:17:24,280 --> 00:17:27,640
that I need not be afraid,
that I need not be concerned,

342
00:17:27,640 --> 00:17:31,720
that the debate is taking
place in the very intensity

343
00:17:31,720 --> 00:17:34,060
that I would wish to
see attributed to it.

344
00:17:34,060 --> 00:17:36,580
But whether that
is the case or not,

345
00:17:36,580 --> 00:17:40,520
we will have to hear
and see from you.

346
00:17:40,520 --> 00:17:42,610
I very much look
forward to this debate.

347
00:17:42,610 --> 00:17:44,632
[APPLAUSE]

348
00:17:44,632 --> 00:17:49,092


349
00:17:49,092 --> 00:17:50,800
URS GASSER: Thank you
so much for setting

350
00:17:50,800 --> 00:17:52,240
the stage so beautifully.

351
00:17:52,240 --> 00:17:55,660
And I realized while
you were speaking

352
00:17:55,660 --> 00:17:59,030
that my American colleagues
didn't have translation,

353
00:17:59,030 --> 00:18:00,800
but you followed it very nicely.

354
00:18:00,800 --> 00:18:03,050
CRYSTAL YANG: And I very
much appreciated the baseball

355
00:18:03,050 --> 00:18:04,050
reference.

356
00:18:04,050 --> 00:18:04,995
MELISSA NOBLES: Yeah,
I got that part.

357
00:18:04,995 --> 00:18:07,050
There were a couple of words
in there I kind of got.

358
00:18:07,050 --> 00:18:09,300
URS GASSER: Yes, but you
really set the stage so well.

359
00:18:09,300 --> 00:18:13,075


360
00:18:13,075 --> 00:18:14,600
Yes, that may be helpful.

361
00:18:14,600 --> 00:18:16,627
MELISSA NOBLES:
Great, thank you.

362
00:18:16,627 --> 00:18:18,460
FRANK-WALTER STEINMEIER:
Should I repeat it?

363
00:18:18,460 --> 00:18:22,308
MELISSA NOBLES: Yeah, thank you.

364
00:18:22,308 --> 00:18:23,270
Bit of a summary.

365
00:18:23,270 --> 00:18:25,160
I heard "Stanford."

366
00:18:25,160 --> 00:18:25,800
I heard--

367
00:18:25,800 --> 00:18:26,842
URS GASSER: Right, right.

368
00:18:26,842 --> 00:18:30,820
And of course you picked up
on that, which is exactly

369
00:18:30,820 --> 00:18:32,435
the segue to my question.

370
00:18:32,435 --> 00:18:33,310
MELISSA NOBLES: Sure.

371
00:18:33,310 --> 00:18:36,280
URS GASSER: So the
president was putting

372
00:18:36,280 --> 00:18:39,280
sort of the societal change
that we're going through,

373
00:18:39,280 --> 00:18:41,650
where technologies
of different sorts

374
00:18:41,650 --> 00:18:43,420
play such a vital
role in the larger

375
00:18:43,420 --> 00:18:45,380
context of the
future of democracy,

376
00:18:45,380 --> 00:18:49,390
and the question of how do
we want to live our lives

377
00:18:49,390 --> 00:18:54,310
and interact with each
other and shape our future.

378
00:18:54,310 --> 00:18:57,280
And within that, he
also referred to,

379
00:18:57,280 --> 00:19:01,520
as you picked up on,
a trip to Stanford,

380
00:19:01,520 --> 00:19:03,930
and pointed out
already, and that's

381
00:19:03,930 --> 00:19:06,013
a theme I want to follow
up on for a few minutes--

382
00:19:06,013 --> 00:19:06,888
MELISSA NOBLES: Sure.

383
00:19:06,888 --> 00:19:09,440
URS GASSER: --that there are
tremendous opportunities,

384
00:19:09,440 --> 00:19:11,620
although currently
the focus is really

385
00:19:11,620 --> 00:19:14,410
on the risks of new
technologies essentially

386
00:19:14,410 --> 00:19:18,180
in public discourse, for
sure, particularly in Europe.

387
00:19:18,180 --> 00:19:21,370
But before we go into
risk mode and talk

388
00:19:21,370 --> 00:19:24,580
about all the pitfalls of
these new technologies,

389
00:19:24,580 --> 00:19:29,040
I would like to
pause and really zoom

390
00:19:29,040 --> 00:19:32,980
in a little bit on this
question, what can technology

391
00:19:32,980 --> 00:19:35,230
do for climate change
and other areas

392
00:19:35,230 --> 00:19:36,610
that the president mentioned.

393
00:19:36,610 --> 00:19:38,310
FRANK-WALTER STEINMEIER:
Against climate change.

394
00:19:38,310 --> 00:19:39,770
URS GASSER: Against,
yes, against.

395
00:19:39,770 --> 00:19:40,600
MELISSA NOBLES:
Right, right, exactly.

396
00:19:40,600 --> 00:19:43,510
URS GASSER: To address some
of the big, big challenges

397
00:19:43,510 --> 00:19:44,200
of our time.

398
00:19:44,200 --> 00:19:45,075
MELISSA NOBLES: Sure.

399
00:19:45,075 --> 00:19:47,940
URS GASSER: And there is this
other place closer to home,

400
00:19:47,940 --> 00:19:48,895
MIT--

401
00:19:48,895 --> 00:19:49,240
MELISSA NOBLES: Right.

402
00:19:49,240 --> 00:19:51,250
URS GASSER: --where many
of these technologies are

403
00:19:51,250 --> 00:19:53,833
developed in the lab, and I was
wondering whether you would be

404
00:19:53,833 --> 00:19:57,400
willing to share maybe two,
three examples from also

405
00:19:57,400 --> 00:19:58,608
your humanities perspective--

406
00:19:58,608 --> 00:19:59,483
MELISSA NOBLES: Sure.

407
00:19:59,483 --> 00:20:02,080
URS GASSER: --that give you
hope and optimism, maybe.

408
00:20:02,080 --> 00:20:03,830
MELISSA NOBLES: Sure,
I'm glad to do that.

409
00:20:03,830 --> 00:20:05,080
Good morning, everyone.

410
00:20:05,080 --> 00:20:07,185
Well, you know, one of
the things about MIT, I

411
00:20:07,185 --> 00:20:08,560
kind of hesitate
in a certain way

412
00:20:08,560 --> 00:20:10,990
to be able to say two or
three, since the Institute

413
00:20:10,990 --> 00:20:14,400
is connected to
technological innovation.

414
00:20:14,400 --> 00:20:16,540
So think I'd rather
say a bit about what

415
00:20:16,540 --> 00:20:20,710
has made MIT such a leader
in thinking innovatively.

416
00:20:20,710 --> 00:20:23,950
And a big part of that has been
the commitment to collaboration

417
00:20:23,950 --> 00:20:25,330
across all five schools.

418
00:20:25,330 --> 00:20:28,288
So it's in recognition that many
of the problems that the world

419
00:20:28,288 --> 00:20:29,830
faces are obviously
global in nature,

420
00:20:29,830 --> 00:20:32,950
and they require knowledge
from all domains.

421
00:20:32,950 --> 00:20:35,050
It isn't just a
scientific problem.

422
00:20:35,050 --> 00:20:36,910
It isn't just an
engineering problem.

423
00:20:36,910 --> 00:20:40,510
It isn't just an economic
problem or a social problem.

424
00:20:40,510 --> 00:20:42,530
It is all of these
things together.

425
00:20:42,530 --> 00:20:46,840
And part of our strengths have
been putting together research

426
00:20:46,840 --> 00:20:48,170
programs to deal with these.

427
00:20:48,170 --> 00:20:52,000
So we have, for example,
the MIT Energy Initiative,

428
00:20:52,000 --> 00:20:58,240
which brings together professors
from engineering, science,

429
00:20:58,240 --> 00:21:00,520
humanities, also social
sciences, the Sloane School,

430
00:21:00,520 --> 00:21:02,230
to look at the economics
and the business

431
00:21:02,230 --> 00:21:06,550
models of what is sustainable
and what not, as well

432
00:21:06,550 --> 00:21:08,620
as architecture and
planning to look

433
00:21:08,620 --> 00:21:11,020
at the ways in which climate
change and the way we

434
00:21:11,020 --> 00:21:13,790
use energy is changing
how we structure cities.

435
00:21:13,790 --> 00:21:16,360
So it is the scope
of the problems

436
00:21:16,360 --> 00:21:19,600
and a commitment to putting
intellectual energies that

437
00:21:19,600 --> 00:21:23,410
are commensurate with
them that, I think,

438
00:21:23,410 --> 00:21:26,920
has set MIT in a better way
for thinking about the future.

439
00:21:26,920 --> 00:21:29,770
So I hesitate to
say any particular,

440
00:21:29,770 --> 00:21:34,030
except to say that the
problems are so massive,

441
00:21:34,030 --> 00:21:37,870
there is no way that technology
cannot be a part of it.

442
00:21:37,870 --> 00:21:40,740
And the issue is, how
do we think creatively

443
00:21:40,740 --> 00:21:42,890
about technology to make
sure that's happening.

444
00:21:42,890 --> 00:21:45,310
And that's a big part of
what education has to do,

445
00:21:45,310 --> 00:21:47,290
to connect students
to understand

446
00:21:47,290 --> 00:21:50,110
that the technology
is an expression,

447
00:21:50,110 --> 00:21:51,543
is a human endeavor, right?

448
00:21:51,543 --> 00:21:52,585
We create the technology.

449
00:21:52,585 --> 00:21:54,670
The technology
doesn't create us,

450
00:21:54,670 --> 00:21:56,980
and we have to start with
some basic commitments.

451
00:21:56,980 --> 00:21:59,720
So that's where we are now.

452
00:21:59,720 --> 00:22:02,140
And I look forward to
saying a bit more later

453
00:22:02,140 --> 00:22:03,725
on about the College
of Computing.

454
00:22:03,725 --> 00:22:04,600
URS GASSER: Fabulous.

455
00:22:04,600 --> 00:22:05,410
Thank you so much.

456
00:22:05,410 --> 00:22:07,940
That's very helpful.

457
00:22:07,940 --> 00:22:10,270
Some sort of an
iteration on this theme,

458
00:22:10,270 --> 00:22:14,240
and taking your point,
where you argue,

459
00:22:14,240 --> 00:22:18,910
well, there's no future without
getting technology right

460
00:22:18,910 --> 00:22:21,790
in a way that helps
us to address some

461
00:22:21,790 --> 00:22:25,240
of these big challenges
we face as humanity,

462
00:22:25,240 --> 00:22:28,780
but also to embrace
the opportunities.

463
00:22:28,780 --> 00:22:32,110
And I was wondering
whether you would

464
00:22:32,110 --> 00:22:35,290
be willing to share your
thinking around this topic.

465
00:22:35,290 --> 00:22:37,660
Much of the ethical
debates of these days

466
00:22:37,660 --> 00:22:43,360
are focused on ethics
in the sense of telling

467
00:22:43,360 --> 00:22:44,740
us what not to do, right?

468
00:22:44,740 --> 00:22:46,120
What lines not to cross.

469
00:22:46,120 --> 00:22:48,393
And we will definitely
return to that,

470
00:22:48,393 --> 00:22:50,185
and this will be the
key part of the panel.

471
00:22:50,185 --> 00:22:54,440
But before we go
there, I was wondering,

472
00:22:54,440 --> 00:22:57,430
is there some sort of
an ethical obligation

473
00:22:57,430 --> 00:23:00,190
for the good use of technology?

474
00:23:00,190 --> 00:23:05,590
And basically a moral
imperative that would almost

475
00:23:05,590 --> 00:23:08,440
be in contrast to the
precautionary principle that's

476
00:23:08,440 --> 00:23:10,850
so popular in Europe
these days, and say,

477
00:23:10,850 --> 00:23:13,900
no, we have to double down
on developing technologies

478
00:23:13,900 --> 00:23:16,510
for the social good and
in the public interest.

479
00:23:16,510 --> 00:23:20,310
How does a philosopher or
ethicist think about that?

480
00:23:20,310 --> 00:23:23,237
EVA WEBER-GURSKA: Yeah,
thank you for that question.

481
00:23:23,237 --> 00:23:24,070
I'm happy to answer.

482
00:23:24,070 --> 00:23:26,850
So I think there are
at least two ways

483
00:23:26,850 --> 00:23:28,440
to understand your question.

484
00:23:28,440 --> 00:23:33,780
First, we may ask if there
is a moral obligation

485
00:23:33,780 --> 00:23:36,960
to generally use
digital technology now

486
00:23:36,960 --> 00:23:39,660
that it has been
invented and developed up

487
00:23:39,660 --> 00:23:44,230
to a point where so much
concrete applications are

488
00:23:44,230 --> 00:23:45,850
possible.

489
00:23:45,850 --> 00:23:49,070
But my answer to
this would be no.

490
00:23:49,070 --> 00:23:51,210
There is no general
moral obligation

491
00:23:51,210 --> 00:23:54,240
to do what can be done.

492
00:23:54,240 --> 00:23:57,320
Because digitization
is just a means,

493
00:23:57,320 --> 00:24:01,290
and moral obligation refers to
ends, to purposes, not the way

494
00:24:01,290 --> 00:24:02,650
we get there.

495
00:24:02,650 --> 00:24:06,030
And so it is an open
question if digitization

496
00:24:06,030 --> 00:24:08,460
is the best way to get
there, where we want to go,

497
00:24:08,460 --> 00:24:10,680
to our moral purpose,
which is, as you already

498
00:24:10,680 --> 00:24:14,320
pointed out, good democracy,
human flourishing, and so on.

499
00:24:14,320 --> 00:24:18,000
And we have just to see exactly
where digitization is helpful

500
00:24:18,000 --> 00:24:19,800
and where not.

501
00:24:19,800 --> 00:24:22,020
But on the other
hand, if you ask

502
00:24:22,020 --> 00:24:28,080
if we theorists, theorists
like us here on the panel,

503
00:24:28,080 --> 00:24:31,980
should point out possible
positive uses of technology

504
00:24:31,980 --> 00:24:35,820
more often, I would say yes,
and that's important, too.

505
00:24:35,820 --> 00:24:37,620
Because otherwise,
the development

506
00:24:37,620 --> 00:24:40,860
of digital technology
mostly is driven

507
00:24:40,860 --> 00:24:43,710
by interests for
financial profit,

508
00:24:43,710 --> 00:24:47,610
and that is not the best
premise for the best

509
00:24:47,610 --> 00:24:51,070
outcome in a moral perspective,
from a moral perspective.

510
00:24:51,070 --> 00:24:54,430
So it would be surely good to
have more people pointing out

511
00:24:54,430 --> 00:24:58,980
the positive uses, but I
think there are already

512
00:24:58,980 --> 00:25:01,620
quite examples for that, too.

513
00:25:01,620 --> 00:25:04,840
And also where reflection and
realization goes together.

514
00:25:04,840 --> 00:25:06,990
For example, at the
Weizenbaum Institute

515
00:25:06,990 --> 00:25:10,820
there was a fellow
in summer in Berlin.

516
00:25:10,820 --> 00:25:13,740
A young colleague
went and developed

517
00:25:13,740 --> 00:25:18,420
an app, which enables
people from different parts

518
00:25:18,420 --> 00:25:22,200
of the political spectrum
to chat and discuss it

519
00:25:22,200 --> 00:25:25,590
with each other
online, for example.

520
00:25:25,590 --> 00:25:28,665
And yeah, I mean, there
are a lot of opportunities,

521
00:25:28,665 --> 00:25:31,890
and we should point
them out, but also I

522
00:25:31,890 --> 00:25:36,930
want to add that although
these projects all

523
00:25:36,930 --> 00:25:41,115
have to be chosen carefully,
because you mentioned already

524
00:25:41,115 --> 00:25:45,360
with climate change, we can
do good things against climate

525
00:25:45,360 --> 00:25:47,170
change with digitalization.

526
00:25:47,170 --> 00:25:49,770
But on the other hand, we also
have to be aware of the fact

527
00:25:49,770 --> 00:25:52,800
that digitization, all
digital technologies

528
00:25:52,800 --> 00:25:56,430
themselves are consuming
masses of energy.

529
00:25:56,430 --> 00:26:02,280
So it would be best to choose
only those projects which

530
00:26:02,280 --> 00:26:04,580
have a really urgent reason.

531
00:26:04,580 --> 00:26:07,410
There has to be something
important at stake

532
00:26:07,410 --> 00:26:11,170
that we invent and
apply new technologies.

533
00:26:11,170 --> 00:26:14,310
And I remember the British
philosopher Dave Parfit

534
00:26:14,310 --> 00:26:18,220
saying that all people,
all humans with healthy,

535
00:26:18,220 --> 00:26:21,090
with two healthy legs should
use the stairs instead

536
00:26:21,090 --> 00:26:24,230
of the elevator in
order to save energy.

537
00:26:24,230 --> 00:26:26,930
Because he said elevators are
just made for people who cannot

538
00:26:26,930 --> 00:26:27,860
walk.

539
00:26:27,860 --> 00:26:30,850
And a bit in a similar
way, we should always

540
00:26:30,850 --> 00:26:35,190
watch out where are the
urgent reasons that we invent

541
00:26:35,190 --> 00:26:37,080
digital technologies for.

542
00:26:37,080 --> 00:26:39,720
And what is urgent,
what is important always

543
00:26:39,720 --> 00:26:41,180
depends on the domain.

544
00:26:41,180 --> 00:26:42,960
It's different in every domain.

545
00:26:42,960 --> 00:26:46,830
In medicine, for example, it's
the diminishing of suffering.

546
00:26:46,830 --> 00:26:48,570
In law, it's justice.

547
00:26:48,570 --> 00:26:53,130
In a democracy, it's
participation and about

548
00:26:53,130 --> 00:26:55,880
[? founded ?] formation
of political opinion.

549
00:26:55,880 --> 00:26:58,920
And only then, when we have
identified precise moral

550
00:26:58,920 --> 00:27:02,280
purposes and we see that
we cannot attain them

551
00:27:02,280 --> 00:27:05,670
but by digital technologies,
then I think then we might

552
00:27:05,670 --> 00:27:08,960
be seen as obliged to use them.

553
00:27:08,960 --> 00:27:10,080
URS GASSER: Wonderful.

554
00:27:10,080 --> 00:27:11,130
Great segue.

555
00:27:11,130 --> 00:27:14,140
You pointed out sort
of the big questions,

556
00:27:14,140 --> 00:27:16,680
but also that these
questions can only

557
00:27:16,680 --> 00:27:20,070
be answered or worked through
in a particular application

558
00:27:20,070 --> 00:27:20,880
context.

559
00:27:20,880 --> 00:27:22,590
You mentioned already a few.

560
00:27:22,590 --> 00:27:26,050
And if I may, to get a
little bit more specific

561
00:27:26,050 --> 00:27:29,850
and put the
conversation from 34--

562
00:27:29,850 --> 00:27:33,480
you know, go from 30,000 feet
a bit lower to 10,000 feet,

563
00:27:33,480 --> 00:27:35,820
maybe, and take
two examples that

564
00:27:35,820 --> 00:27:37,770
illustrate some sort
of the struggle, how

565
00:27:37,770 --> 00:27:43,150
we embrace opportunities, but
also protect against risks.

566
00:27:43,150 --> 00:27:46,560
And Matthew and
Crystal, as I already

567
00:27:46,560 --> 00:27:49,860
mentioned in the
introduction, you

568
00:27:49,860 --> 00:27:53,250
have interesting work
that's sort of serves

569
00:27:53,250 --> 00:27:54,780
as a case study in our context.

570
00:27:54,780 --> 00:27:58,860
Matthew, focusing on health
and public health and the role

571
00:27:58,860 --> 00:28:02,250
of technology, whether
it's AI or IoT,

572
00:28:02,250 --> 00:28:05,730
how are some of these
questions that you've

573
00:28:05,730 --> 00:28:09,370
identified crystallizing, and
where do you see things going?

574
00:28:09,370 --> 00:28:11,160
What are some of the concerns?

575
00:28:11,160 --> 00:28:12,410
What's the state of play?

576
00:28:12,410 --> 00:28:14,368
MATTHEW LIAO: Yeah, so
good morning, everybody.

577
00:28:14,368 --> 00:28:18,280
So as a Professor Gasser
has said, I'm a philosopher.

578
00:28:18,280 --> 00:28:20,960
And I have a book
coming out called

579
00:28:20,960 --> 00:28:23,030
The Ethics of Artificial
Intelligence coming out

580
00:28:23,030 --> 00:28:23,840
next March.

581
00:28:23,840 --> 00:28:26,780
And we cover a number of
these different issues

582
00:28:26,780 --> 00:28:28,970
in the ethics of AI.

583
00:28:28,970 --> 00:28:31,595
And one of the applications
of the ethics of AI

584
00:28:31,595 --> 00:28:33,830
is in the realm of health care.

585
00:28:33,830 --> 00:28:36,800
They are actually a lot of
really exciting opportunities

586
00:28:36,800 --> 00:28:39,260
and a lot of development,
a lot of things being done

587
00:28:39,260 --> 00:28:41,070
in the area of health care.

588
00:28:41,070 --> 00:28:44,270
So for example, machine
learning is being deployed

589
00:28:44,270 --> 00:28:46,940
to screen cancer cells.

590
00:28:46,940 --> 00:28:51,800
It's found that it's almost
as effective as radiologists.

591
00:28:51,800 --> 00:28:54,680
It's also been used
in ophthalmology.

592
00:28:54,680 --> 00:29:00,260
It's been used to screen, to
figure out whether a embryo is

593
00:29:00,260 --> 00:29:03,350
going to be viable or not.

594
00:29:03,350 --> 00:29:05,480
Natural language
processing is being

595
00:29:05,480 --> 00:29:08,180
used to figure
out whether people

596
00:29:08,180 --> 00:29:10,290
are having suicidal thoughts.

597
00:29:10,290 --> 00:29:12,740
So there are a lot of
really exciting developments

598
00:29:12,740 --> 00:29:15,650
that are currently underway.

599
00:29:15,650 --> 00:29:18,560
And what that means
for us is that it

600
00:29:18,560 --> 00:29:24,230
can really, for example, reduce
health care costs in the US.

601
00:29:24,230 --> 00:29:28,250
I think we spend about $3 to
$4 trillion in health care

602
00:29:28,250 --> 00:29:29,340
each year.

603
00:29:29,340 --> 00:29:33,170
And so one of the things
that machine learning can do

604
00:29:33,170 --> 00:29:36,920
is reduce administrative costs
in health care, for example.

605
00:29:36,920 --> 00:29:41,960
It can also assist
facilitating drug discovery.

606
00:29:41,960 --> 00:29:45,740
And finally, another
example is, it

607
00:29:45,740 --> 00:29:48,720
can really realize the
vision of precision medicine.

608
00:29:48,720 --> 00:29:52,340
So for example,
Fitbits, wearables,

609
00:29:52,340 --> 00:29:55,040
to figure out
healthy lifestyles,

610
00:29:55,040 --> 00:29:57,650
what you should be
eating, your calorie

611
00:29:57,650 --> 00:29:59,850
intakes and so on and so forth.

612
00:29:59,850 --> 00:30:03,180
So all those are really,
really exciting developments.

613
00:30:03,180 --> 00:30:05,570
I'm an ethicist, so I
also think about some

614
00:30:05,570 --> 00:30:06,810
of the ethical problems.

615
00:30:06,810 --> 00:30:08,600
And I just want to
very quickly share

616
00:30:08,600 --> 00:30:12,360
some of the ethical
concerns with you as well.

617
00:30:12,360 --> 00:30:15,380
So one of the biggest
challenges with machine learning

618
00:30:15,380 --> 00:30:18,390
is that it requires
a lot of data.

619
00:30:18,390 --> 00:30:20,600
And so what that
means is, someone's

620
00:30:20,600 --> 00:30:22,760
got to go out there and
collect all these data.

621
00:30:22,760 --> 00:30:25,650
And then you get into
issues about privacy,

622
00:30:25,650 --> 00:30:26,860
especially in health care.

623
00:30:26,860 --> 00:30:29,250
It's personal data that
we're talking about.

624
00:30:29,250 --> 00:30:34,100
So one obvious example
is Facebook and Cambridge

625
00:30:34,100 --> 00:30:37,220
Analytica collecting
a lot of information

626
00:30:37,220 --> 00:30:39,510
from Facebook users.

627
00:30:39,510 --> 00:30:42,000
Another example is
GlaxoSmithKline.

628
00:30:42,000 --> 00:30:45,140
They just recently bought
this company 23andMe,

629
00:30:45,140 --> 00:30:49,400
which is ancestry type, you
upload your information,

630
00:30:49,400 --> 00:30:51,000
it gets your
genetic information.

631
00:30:51,000 --> 00:30:53,140
So now they have
all the database.

632
00:30:53,140 --> 00:30:56,000
And so one of the things we
need to really be worried about

633
00:30:56,000 --> 00:31:00,740
is whether they're collecting
the data appropriately,

634
00:31:00,740 --> 00:31:03,620
are they violating rights,
what are the implications

635
00:31:03,620 --> 00:31:05,520
of the individuals?

636
00:31:05,520 --> 00:31:08,400
Another issue is going
to be the garbage in,

637
00:31:08,400 --> 00:31:09,800
garbage out problem.

638
00:31:09,800 --> 00:31:13,970
So the algorithms
that we're using today

639
00:31:13,970 --> 00:31:17,580
are going to only be as
good as the data themselves.

640
00:31:17,580 --> 00:31:19,880
And so what we're
finding is that

641
00:31:19,880 --> 00:31:22,800
sometimes the data sets that
we're collecting are not--

642
00:31:22,800 --> 00:31:25,400
they don't have
accurate representations

643
00:31:25,400 --> 00:31:26,660
of the subjects.

644
00:31:26,660 --> 00:31:29,360
So take for example
self-driving cars.

645
00:31:29,360 --> 00:31:32,320
It turns out that
self-driving cars,

646
00:31:32,320 --> 00:31:35,900
they're not so good at
detecting people of color,

647
00:31:35,900 --> 00:31:41,480
because the training sets, the
training data that they use,

648
00:31:41,480 --> 00:31:47,150
they don't have enough of a set
of people of color in the data

649
00:31:47,150 --> 00:31:47,760
set.

650
00:31:47,760 --> 00:31:52,100
And so that's a problem when we
deploy those sort of data, sort

651
00:31:52,100 --> 00:31:54,620
of the algorithm in the wild.

652
00:31:54,620 --> 00:31:56,750
And I'll just say
one more thing.

653
00:31:56,750 --> 00:32:00,230
The biggest concern I have
with machine learning right now

654
00:32:00,230 --> 00:32:02,100
is there's something
called deep learning.

655
00:32:02,100 --> 00:32:04,260
And deep learning is
actually a technical term.

656
00:32:04,260 --> 00:32:12,650
It just means that it's using
a big network to figure out

657
00:32:12,650 --> 00:32:15,770
how a machine should act.

658
00:32:15,770 --> 00:32:20,960
And it's powered a lot of the
recent developments since 2012.

659
00:32:20,960 --> 00:32:24,940
It's powered a lot of
the new breakthroughs.

660
00:32:24,940 --> 00:32:26,690
But one of the problems
with deep learning

661
00:32:26,690 --> 00:32:29,340
is that it just doesn't
capture the causality,

662
00:32:29,340 --> 00:32:30,920
the causal relations.

663
00:32:30,920 --> 00:32:33,930
It doesn't really
understand what it's doing.

664
00:32:33,930 --> 00:32:38,130
And so it's linear regression.

665
00:32:38,130 --> 00:32:39,620
It's a lot of math.

666
00:32:39,620 --> 00:32:42,140
But here's one problem.

667
00:32:42,140 --> 00:32:46,710
So there's something called
generative adversarial network.

668
00:32:46,710 --> 00:32:48,420
It's a type of single--

669
00:32:48,420 --> 00:32:50,540
so one type of
attack is something

670
00:32:50,540 --> 00:32:53,900
called a single pixel attack.

671
00:32:53,900 --> 00:32:57,300
So machine learning is very
good at image classification.

672
00:32:57,300 --> 00:33:01,110
They can take images, and they
classify them very accurately.

673
00:33:01,110 --> 00:33:03,410
But science--
researchers have found

674
00:33:03,410 --> 00:33:08,000
that if you just take an
image, say an image of a car,

675
00:33:08,000 --> 00:33:09,950
and you just take one
pixel and you change it

676
00:33:09,950 --> 00:33:12,470
from black to white,
the machine learning

677
00:33:12,470 --> 00:33:14,040
will completely screw it up.

678
00:33:14,040 --> 00:33:16,990
So for example, with
the image of the car,

679
00:33:16,990 --> 00:33:23,910
it will now classify that image
as a dog with 99% confidence.

680
00:33:23,910 --> 00:33:26,830
And just imagine deploying
that type of machine learning

681
00:33:26,830 --> 00:33:29,900
in the context of health
care, when people's lives are

682
00:33:29,900 --> 00:33:34,040
at stake, or in the context
of self-driving cars, right?

683
00:33:34,040 --> 00:33:36,490
And so I think
we're going to get

684
00:33:36,490 --> 00:33:38,500
into more of these
discussions later,

685
00:33:38,500 --> 00:33:41,950
but I think that's where we have
to be careful about rolling out

686
00:33:41,950 --> 00:33:45,043
these technologies.

687
00:33:45,043 --> 00:33:46,960
URS GASSER: Crystal,
does that sound familiar,

688
00:33:46,960 --> 00:33:49,270
listening to these
stories from health

689
00:33:49,270 --> 00:33:52,690
when you look at your work on
the use of algorithms and data

690
00:33:52,690 --> 00:33:55,023
in the criminal justice system
or where are differences?

691
00:33:55,023 --> 00:33:56,398
CRYSTAL YANG:
Yeah, I think there

692
00:33:56,398 --> 00:33:57,860
are a lot of similarities.

693
00:33:57,860 --> 00:34:00,220
And I think as some
of the other panelists

694
00:34:00,220 --> 00:34:03,460
have pointed out, while
algorithms are now basically

695
00:34:03,460 --> 00:34:06,140
used in so many
parts of society,

696
00:34:06,140 --> 00:34:07,930
one of the areas
where they have had

697
00:34:07,930 --> 00:34:10,420
a very dramatic
increase in usage

698
00:34:10,420 --> 00:34:13,210
is the United States
criminal justice system.

699
00:34:13,210 --> 00:34:16,630
And the algorithms here we
often call risk assessment

700
00:34:16,630 --> 00:34:20,139
instruments, because what the
algorithms are trying to do

701
00:34:20,139 --> 00:34:23,230
is predict somebody's
future criminality.

702
00:34:23,230 --> 00:34:25,420
And these instruments
now are used

703
00:34:25,420 --> 00:34:29,170
at various stages of the
criminal justice system, things

704
00:34:29,170 --> 00:34:33,070
from policing, to pretrial
and bail decisions,

705
00:34:33,070 --> 00:34:37,000
to sentencing, to probation
and parole as well.

706
00:34:37,000 --> 00:34:41,000
And just some examples,
take predictive policing.

707
00:34:41,000 --> 00:34:42,790
One of the common
technologies is

708
00:34:42,790 --> 00:34:46,270
called Pred Poll, which is
used by the Los Angeles Police

709
00:34:46,270 --> 00:34:49,090
Department and over 60
other police departments

710
00:34:49,090 --> 00:34:50,699
across the United States.

711
00:34:50,699 --> 00:34:53,710
It uses historical
data on crime types,

712
00:34:53,710 --> 00:34:55,870
where crimes have
happened, to predict

713
00:34:55,870 --> 00:34:59,830
the future incidents of
different criminal incidents.

714
00:34:59,830 --> 00:35:02,320
In sentencing, now
many states allow

715
00:35:02,320 --> 00:35:04,810
judges to consider
risk scores that

716
00:35:04,810 --> 00:35:07,360
are meant to predict the
future risk of committing

717
00:35:07,360 --> 00:35:09,190
new criminal behavior.

718
00:35:09,190 --> 00:35:12,410
One common algorithm here, which
has been in the news a lot,

719
00:35:12,410 --> 00:35:14,980
you may have heard of,
is the Compass algorithm.

720
00:35:14,980 --> 00:35:17,470
It's a proprietary
algorithm, so we actually

721
00:35:17,470 --> 00:35:20,200
don't know exactly the
underlying algorithmic

722
00:35:20,200 --> 00:35:23,230
structure, that
classifies individuals

723
00:35:23,230 --> 00:35:26,860
on a scale of 1 to 10 in terms
of their predictive likelihood

724
00:35:26,860 --> 00:35:31,020
of recidivism, using how
that person answers questions

725
00:35:31,020 --> 00:35:34,310
on a 137 question survey.

726
00:35:34,310 --> 00:35:36,310
So these are just
some of the examples,

727
00:35:36,310 --> 00:35:40,870
and I think they raise a huge
host of issues and challenges.

728
00:35:40,870 --> 00:35:44,590
One that I think requires
a lot of understanding

729
00:35:44,590 --> 00:35:48,790
from philosophers and ethicists
is, do these risk assessment

730
00:35:48,790 --> 00:35:53,150
tools have a role to play, even,
in the criminal justice system?

731
00:35:53,150 --> 00:35:57,310
I think some view the endeavor
of predicting future risk

732
00:35:57,310 --> 00:36:01,210
as wrongheaded, and believe
that because of this input data

733
00:36:01,210 --> 00:36:06,130
garbage in, garbage out type of
problem, that using algorithms

734
00:36:06,130 --> 00:36:09,700
to predict future risk will
only entrench or potentially

735
00:36:09,700 --> 00:36:13,120
exacerbate inequalities
and inequities that we

736
00:36:13,120 --> 00:36:15,980
see in society at large.

737
00:36:15,980 --> 00:36:18,780
On the other hand, and I place
myself more in this camp,

738
00:36:18,780 --> 00:36:21,880
while there is acknowledgment
that the algorithms are often

739
00:36:21,880 --> 00:36:25,390
imperfect, I think it's
also important to consider

740
00:36:25,390 --> 00:36:27,330
the relevant counterfactual.

741
00:36:27,330 --> 00:36:30,340
The counterfactual is not
a world free of inequality,

742
00:36:30,340 --> 00:36:31,360
of inequity.

743
00:36:31,360 --> 00:36:34,600
It's a counterfactual in which
we have human decision makers.

744
00:36:34,600 --> 00:36:35,380
And guess what?

745
00:36:35,380 --> 00:36:38,440
There's lots of evidence
that human decision makers

746
00:36:38,440 --> 00:36:41,830
have a big role to play in
perpetuating inequalities

747
00:36:41,830 --> 00:36:43,970
through bias and inconsistency.

748
00:36:43,970 --> 00:36:46,240
So there's a role to, I
think, to consider what are

749
00:36:46,240 --> 00:36:48,280
we comparing the algorithms to?

750
00:36:48,280 --> 00:36:49,990
It's not a perfect world.

751
00:36:49,990 --> 00:36:51,100
It's humans.

752
00:36:51,100 --> 00:36:53,170
I think another set
of design questions

753
00:36:53,170 --> 00:36:56,540
that Matthew has gotten at is,
in the criminal justice system,

754
00:36:56,540 --> 00:36:59,410
there are lots of open,
unresolved questions

755
00:36:59,410 --> 00:37:01,870
about how do we
design an algorithm?

756
00:37:01,870 --> 00:37:04,150
If we're going to
predict risk, can we

757
00:37:04,150 --> 00:37:06,190
consider individual
characteristics

758
00:37:06,190 --> 00:37:08,710
like somebody's race or
ethnicity, what we often

759
00:37:08,710 --> 00:37:10,750
call protected characteristics?

760
00:37:10,750 --> 00:37:14,440
If you can't, can you consider
non-protected characteristics,

761
00:37:14,440 --> 00:37:17,530
things like education,
where somebody lives,

762
00:37:17,530 --> 00:37:22,660
which can effectively proxy for
a person's race or ethnicity?

763
00:37:22,660 --> 00:37:24,700
There is also
complicated questions

764
00:37:24,700 --> 00:37:26,980
about how to evaluate
if an algorithm is

765
00:37:26,980 --> 00:37:29,050
doing what we want it to do.

766
00:37:29,050 --> 00:37:32,890
What does it mean, for instance,
for an algorithm to be fair?

767
00:37:32,890 --> 00:37:35,710
It turns out here the
law has not so much

768
00:37:35,710 --> 00:37:39,560
to say so far about how to
define or measure fairness.

769
00:37:39,560 --> 00:37:43,150
And even outside the law,
there's a very lively computer

770
00:37:43,150 --> 00:37:46,450
science debate about algorithmic
fairness, where there's

771
00:37:46,450 --> 00:37:48,530
very different
definitions of fairness,

772
00:37:48,530 --> 00:37:50,770
that in many
circumstances we could all

773
00:37:50,770 --> 00:37:53,810
say that one sounds great
or that one sounds great,

774
00:37:53,810 --> 00:37:55,750
but it's actually been
shown mathematically

775
00:37:55,750 --> 00:37:59,440
that in many instances it is
impossible to simultaneously

776
00:37:59,440 --> 00:38:02,870
satisfy all those notions
of algorithmic fairness.

777
00:38:02,870 --> 00:38:05,650
And so then that requires
a normative choice by us

778
00:38:05,650 --> 00:38:08,740
as a society or a
legal system to choose

779
00:38:08,740 --> 00:38:11,740
which of those definitions
of algorithmic fairness

780
00:38:11,740 --> 00:38:12,490
should dominate.

781
00:38:12,490 --> 00:38:15,520
So I think those are just
a couple of the key issues

782
00:38:15,520 --> 00:38:17,967
and challenges that I see in
the criminal justice system.

783
00:38:17,967 --> 00:38:19,550
URS GASSER: No
shortage of challenges.

784
00:38:19,550 --> 00:38:21,133
CRYSTAL YANG: No
shortage, absolutely.

785
00:38:21,133 --> 00:38:24,005
URS GASSER: It was
clear from both stories.

786
00:38:24,005 --> 00:38:30,670
I'd love to build up on
this and ask Wolfgang,

787
00:38:30,670 --> 00:38:34,450
Crystal made the
point that part of it

788
00:38:34,450 --> 00:38:37,220
is a story about
technology, but part of it

789
00:38:37,220 --> 00:38:39,980
also seems to be a story
about society at large,

790
00:38:39,980 --> 00:38:44,300
about the institutions
we already have in place.

791
00:38:44,300 --> 00:38:46,940
Part of it seems to
be about human nature,

792
00:38:46,940 --> 00:38:48,530
with our own biases.

793
00:38:48,530 --> 00:38:52,910
So how do you think about that
as we have this intense debate,

794
00:38:52,910 --> 00:38:58,430
debates about AI decision making
versus human decision making?

795
00:38:58,430 --> 00:39:02,990
And should we replace
judges by AIs or not?

796
00:39:02,990 --> 00:39:06,977
How much is it about
technology, really?

797
00:39:06,977 --> 00:39:08,810
WOLFGANG SCHULZ: I think
to respond to that,

798
00:39:08,810 --> 00:39:11,870
I have to go to flight
level 10,000 again.

799
00:39:11,870 --> 00:39:14,690
I would say that I
am descending later.

800
00:39:14,690 --> 00:39:16,608


801
00:39:16,608 --> 00:39:17,900
URS GASSER: Only in a good way.

802
00:39:17,900 --> 00:39:18,530
Don't worry.

803
00:39:18,530 --> 00:39:20,030
WOLFGANG SCHULZ:
In a good way, yes.

804
00:39:20,030 --> 00:39:20,800
I hope.

805
00:39:20,800 --> 00:39:25,530
So when we talk about
technology in expert circles

806
00:39:25,530 --> 00:39:28,970
but in society at large,
then we very often

807
00:39:28,970 --> 00:39:30,930
have a distinction between
here is the society,

808
00:39:30,930 --> 00:39:32,390
here is the technology.

809
00:39:32,390 --> 00:39:34,910
And that is a dangerous
thing, because then we

810
00:39:34,910 --> 00:39:38,420
frame technology as a
kind of natural disaster

811
00:39:38,420 --> 00:39:41,770
that is coming, and we have to
build walls to cope with that.

812
00:39:41,770 --> 00:39:45,470
And we are not in
the mode of creating

813
00:39:45,470 --> 00:39:47,990
the technology as a
society and together

814
00:39:47,990 --> 00:39:49,710
with the different disciplines.

815
00:39:49,710 --> 00:39:51,453
So I think we have
to be very careful

816
00:39:51,453 --> 00:39:53,120
how we talk about
these things and where

817
00:39:53,120 --> 00:39:55,500
we talk about tensions.

818
00:39:55,500 --> 00:40:00,320
And I can, I think, build on
what Crystal said, because we

819
00:40:00,320 --> 00:40:03,530
are doing some research on
the criminal justice system

820
00:40:03,530 --> 00:40:05,930
as well, and have done recently.

821
00:40:05,930 --> 00:40:08,930
And what I find
interesting is that when

822
00:40:08,930 --> 00:40:13,340
we talk about technology
coming into processes, then

823
00:40:13,340 --> 00:40:15,710
we start thinking
about what our quality

824
00:40:15,710 --> 00:40:18,230
measures are as a society here.

825
00:40:18,230 --> 00:40:20,810
And I had a discussion
with German judges

826
00:40:20,810 --> 00:40:24,140
a couple of months ago, and we
are talking about sentencing.

827
00:40:24,140 --> 00:40:28,010
And we are talking about
AI supporting that,

828
00:40:28,010 --> 00:40:31,010
and then I raised the
question of explainability,

829
00:40:31,010 --> 00:40:33,660
which is one of
the issues in AI,

830
00:40:33,660 --> 00:40:37,790
that say we cannot really see
and explain what happens there.

831
00:40:37,790 --> 00:40:40,460
And then one of the judges
said, wait a moment.

832
00:40:40,460 --> 00:40:42,950
Ask me, can I explain
what I am doing

833
00:40:42,950 --> 00:40:45,120
when I come to this decision?

834
00:40:45,120 --> 00:40:47,030
And I'm not sure that
I really can do that.

835
00:40:47,030 --> 00:40:51,660
I can give a reason that is
valid in the legal system,

836
00:40:51,660 --> 00:40:55,370
but I cannot really explain
what my motives were here.

837
00:40:55,370 --> 00:40:58,770
And then we had a debate on
what are the factors there.

838
00:40:58,770 --> 00:41:02,360
And in the German legal
system and the criminal law,

839
00:41:02,360 --> 00:41:05,330
it's not very well elaborated
what the criteria are.

840
00:41:05,330 --> 00:41:07,250
So it's very, very vague.

841
00:41:07,250 --> 00:41:08,930
And so we had a
very fruitful debate

842
00:41:08,930 --> 00:41:11,120
on what the values actually are.

843
00:41:11,120 --> 00:41:14,240
And you can have the same
in other fields of society.

844
00:41:14,240 --> 00:41:17,750
We have next week
or the week after,

845
00:41:17,750 --> 00:41:20,540
a workshop with
computer scientists

846
00:41:20,540 --> 00:41:24,800
and people from communication
science and law,

847
00:41:24,800 --> 00:41:31,310
talking about how to understand
diversity in recommender

848
00:41:31,310 --> 00:41:32,990
systems for the media.

849
00:41:32,990 --> 00:41:36,720
And we want to come up with
ideas what that actually is.

850
00:41:36,720 --> 00:41:38,390
And then you have
to go back, and what

851
00:41:38,390 --> 00:41:40,920
do you want as a
society actually,

852
00:41:40,920 --> 00:41:42,990
when you talk about diversity?

853
00:41:42,990 --> 00:41:46,280
So I think that's a good
thing that technology forces

854
00:41:46,280 --> 00:41:51,710
us to ask those hard questions
about societal values

855
00:41:51,710 --> 00:41:54,920
and to better understand what
makes human decision making

856
00:41:54,920 --> 00:41:55,880
so special.

857
00:41:55,880 --> 00:41:58,610
We are talking a lot about
things like tested knowledge

858
00:41:58,610 --> 00:42:02,420
and tested norms, things that
we all understand because we

859
00:42:02,420 --> 00:42:04,020
are part of the society.

860
00:42:04,020 --> 00:42:06,380
And we cannot really
explain why we do that this

861
00:42:06,380 --> 00:42:10,310
way or that way, because it's
tacit knowledge or tacit norms.

862
00:42:10,310 --> 00:42:13,460
And that is something that
you cannot really, now,

863
00:42:13,460 --> 00:42:15,500
I would say, build
into technology.

864
00:42:15,500 --> 00:42:18,500
That would require technology
to be part of society

865
00:42:18,500 --> 00:42:21,900
and learn in interaction, and
I think we are far from that

866
00:42:21,900 --> 00:42:22,400
so far.

867
00:42:22,400 --> 00:42:27,470
So I believe that this
is a twist of the debate

868
00:42:27,470 --> 00:42:30,860
that very often we
do not really include

869
00:42:30,860 --> 00:42:34,640
in our conversation, when
we talk about this society

870
00:42:34,640 --> 00:42:38,740
here, technology there aspect.

871
00:42:38,740 --> 00:42:40,610
URS GASSER: So if
Wolfgang is right,

872
00:42:40,610 --> 00:42:43,670
and he is most often
right, as we know--

873
00:42:43,670 --> 00:42:45,170
JEANETTE HOFMANN:
I don't deny that.

874
00:42:45,170 --> 00:42:48,470
URS GASSER: And technology is
deeply embedded in society,

875
00:42:48,470 --> 00:42:52,220
and as we heard the president
opening, in his opening

876
00:42:52,220 --> 00:42:54,740
remarks, we as societies
are in a learning

877
00:42:54,740 --> 00:42:58,550
process ourselves, how to
cope with massive challenges

878
00:42:58,550 --> 00:43:02,180
and transformations
of all sorts.

879
00:43:02,180 --> 00:43:04,520
Based on the work you've
been doing, following

880
00:43:04,520 --> 00:43:07,160
early debates around
internet regulation

881
00:43:07,160 --> 00:43:11,510
and approaches to
governance, what's currently

882
00:43:11,510 --> 00:43:14,480
happening in this
societal learning process

883
00:43:14,480 --> 00:43:18,980
as we try to identify
and agree and regulate

884
00:43:18,980 --> 00:43:21,890
good uses versus bad uses
across different contexts,

885
00:43:21,890 --> 00:43:23,980
and we only highlighted
two examples

886
00:43:23,980 --> 00:43:26,050
and could add many more.

887
00:43:26,050 --> 00:43:29,420
What sorts of norms
are emerging and what

888
00:43:29,420 --> 00:43:34,125
some of the dynamics around
these norms as you observe?

889
00:43:34,125 --> 00:43:35,500
JEANETTE HOFMANN:
Thank you, Urs.

890
00:43:35,500 --> 00:43:37,990
I could talk for hours
on this question.

891
00:43:37,990 --> 00:43:40,900
I really like it.

892
00:43:40,900 --> 00:43:42,610
Let me go one step back.

893
00:43:42,610 --> 00:43:46,120
At the time when the internet
and digital technologies--

894
00:43:46,120 --> 00:43:47,470
URS GASSER: Switch off the mic.

895
00:43:47,470 --> 00:43:49,560
On the mic.

896
00:43:49,560 --> 00:43:56,000
FRANK-WALTER STEINMEIER:
[SPEAKING GERMAN]

897
00:43:56,000 --> 00:43:57,700
JEANETTE HOFMANN: OK.

898
00:43:57,700 --> 00:43:58,408
Around the time--

899
00:43:58,408 --> 00:44:00,283
FRANK-WALTER STEINMEIER:
Well, that's better.

900
00:44:00,283 --> 00:44:02,770
JEANETTE HOFMANN: --Digital
technologies really

901
00:44:02,770 --> 00:44:06,800
became more present
in our societies,

902
00:44:06,800 --> 00:44:10,720
Western societies went
through a long period

903
00:44:10,720 --> 00:44:16,420
of privatization and liberation
from old state monopolies.

904
00:44:16,420 --> 00:44:22,390
And we thought of that
the force of the internet

905
00:44:22,390 --> 00:44:27,550
as a form of liberalization,
and that kind of idea

906
00:44:27,550 --> 00:44:32,380
of self-regulation and "let the
markets determine the future,"

907
00:44:32,380 --> 00:44:35,950
we thought that this was
a very good alternative.

908
00:44:35,950 --> 00:44:40,960
And this we have driven
to a point where we now

909
00:44:40,960 --> 00:44:45,070
regard digital
technologies nearly

910
00:44:45,070 --> 00:44:48,100
as a self-driving
autonomous force.

911
00:44:48,100 --> 00:44:54,070
We ascribe a lot of power and
agency to digital technologies

912
00:44:54,070 --> 00:44:58,120
themselves and the
companies who develop them.

913
00:44:58,120 --> 00:45:00,790
I would say that the
debate we see now

914
00:45:00,790 --> 00:45:05,170
about AI and ethical
frameworks is

915
00:45:05,170 --> 00:45:10,360
an echo of that, the idea
that ethical principles might

916
00:45:10,360 --> 00:45:14,980
be good enough to give us an
orientation for the future

917
00:45:14,980 --> 00:45:17,080
of artificial intelligence.

918
00:45:17,080 --> 00:45:20,980
But we need to ask
ourselves whether we

919
00:45:20,980 --> 00:45:25,690
get enough accountability
out of ethical guidelines

920
00:45:25,690 --> 00:45:27,310
and frameworks.

921
00:45:27,310 --> 00:45:29,710
I just came back
from the West Coast,

922
00:45:29,710 --> 00:45:33,730
where you see really
a change of wind.

923
00:45:33,730 --> 00:45:36,760
Companies now begin
to wonder whether they

924
00:45:36,760 --> 00:45:42,910
do not need a legal framework
for the future development.

925
00:45:42,910 --> 00:45:46,300
Such a legal framework
could be, for example,

926
00:45:46,300 --> 00:45:50,110
anchored in human rights,
and legislation could

927
00:45:50,110 --> 00:45:52,970
build on fundamental rights.

928
00:45:52,970 --> 00:45:57,350
They could set limits
to future developments,

929
00:45:57,350 --> 00:46:00,560
also to make us see
that finally it is

930
00:46:00,560 --> 00:46:02,980
society that shapes technology.

931
00:46:02,980 --> 00:46:06,100
It's not that technology
sets its own rules.

932
00:46:06,100 --> 00:46:09,760
But we are not really aware
of it, I think, at the moment.

933
00:46:09,760 --> 00:46:12,670
We nearly have
lost the capability

934
00:46:12,670 --> 00:46:16,990
to see and to recognize
how we change technologies

935
00:46:16,990 --> 00:46:18,410
as societies.

936
00:46:18,410 --> 00:46:22,990
So we need to perhaps
turn around a bit,

937
00:46:22,990 --> 00:46:26,590
give up this idea of
complete self-regulation,

938
00:46:26,590 --> 00:46:30,520
and come to new models
that sit somewhere

939
00:46:30,520 --> 00:46:35,350
in between a market approach
and a pure government approach.

940
00:46:35,350 --> 00:46:38,140
We need new
regulatory frameworks

941
00:46:38,140 --> 00:46:41,650
that need to work across
national boundaries,

942
00:46:41,650 --> 00:46:45,380
even though we can,
I think, not hope

943
00:46:45,380 --> 00:46:47,350
for multilateral approaches.

944
00:46:47,350 --> 00:46:51,850
We need something below, and
the GDPR, the General Data

945
00:46:51,850 --> 00:46:56,260
Protection Regulation that the
European Commission introduced,

946
00:46:56,260 --> 00:46:59,050
is often mentioned
as a gold standard

947
00:46:59,050 --> 00:47:00,820
for that kind of approach.

948
00:47:00,820 --> 00:47:03,700
Perhaps some countries
can get together,

949
00:47:03,700 --> 00:47:08,810
build a legal framework and
export it via trade agreements.

950
00:47:08,810 --> 00:47:10,980
[APPLAUSE]

951
00:47:10,980 --> 00:47:12,138


952
00:47:12,138 --> 00:47:13,680
URS GASSER: That is
some good advice.

953
00:47:13,680 --> 00:47:15,385
Thank you, Jeanette.

954
00:47:15,385 --> 00:47:20,050
So a couple of things that I
would like to follow up on.

955
00:47:20,050 --> 00:47:22,450
One is this role of
the ethics principles.

956
00:47:22,450 --> 00:47:24,190
You mentioned there
is a flourishing

957
00:47:24,190 --> 00:47:26,590
of ethical principles
around AI in particular.

958
00:47:26,590 --> 00:47:30,610
I think 130 or
something are out there.

959
00:47:30,610 --> 00:47:35,920
We tried to map some of them,
but it's getting quite a task.

960
00:47:35,920 --> 00:47:39,310
But on the other
hand side, given also

961
00:47:39,310 --> 00:47:43,330
Wolfgang's remarks and the
opening statement by Herr

962
00:47:43,330 --> 00:47:47,230
Bundespresident, there is
value to these ethical debates,

963
00:47:47,230 --> 00:47:48,570
nonetheless, right?

964
00:47:48,570 --> 00:47:50,260
And you also make
this point, of course,

965
00:47:50,260 --> 00:47:53,980
that we need all different
approaches and tools, probably

966
00:47:53,980 --> 00:47:55,520
including law but also ethics.

967
00:47:55,520 --> 00:47:59,680
And if I may ask
you, how do you think

968
00:47:59,680 --> 00:48:01,390
about these ethical principles?

969
00:48:01,390 --> 00:48:05,080
What's the value in
these ethical norms

970
00:48:05,080 --> 00:48:07,720
crystallizing guidelines
and things like that,

971
00:48:07,720 --> 00:48:09,550
whether it's for
companies or an act

972
00:48:09,550 --> 00:48:12,730
by international organizations
like OECD or even

973
00:48:12,730 --> 00:48:15,860
by nation states?

974
00:48:15,860 --> 00:48:17,410
What's the promise,
but also what

975
00:48:17,410 --> 00:48:19,810
are the limitations
of ethical approaches

976
00:48:19,810 --> 00:48:24,610
of this sort when we deal with
these complex, messy problems?

977
00:48:24,610 --> 00:48:27,380
EVA WEBER-GURSKA: Yeah, so
ethics and law of course

978
00:48:27,380 --> 00:48:30,230
have to be distinguished,
although they are connected.

979
00:48:30,230 --> 00:48:34,490
Ethics is, I would say,
the explicit formulation

980
00:48:34,490 --> 00:48:37,610
of implicit norms
that guide or should

981
00:48:37,610 --> 00:48:43,700
guide our everyday actions in
our life, our living together.

982
00:48:43,700 --> 00:48:48,500
And law, it's the core
of the organization

983
00:48:48,500 --> 00:48:53,390
of a state or a nation,
transforms some of these norms

984
00:48:53,390 --> 00:48:56,810
into concrete rules, the
infringement of which

985
00:48:56,810 --> 00:48:59,900
then is bound up with
sanctions by the state.

986
00:48:59,900 --> 00:49:01,550
So this is something different.

987
00:49:01,550 --> 00:49:04,910
And not all moral norms are
legal norms and vice versa,

988
00:49:04,910 --> 00:49:05,960
of course.

989
00:49:05,960 --> 00:49:10,400
But ethical guidelines now for
new topics like digitization

990
00:49:10,400 --> 00:49:13,190
can be, I think,
helpful first steps

991
00:49:13,190 --> 00:49:18,800
to show something that then can
be transformed into law, too.

992
00:49:18,800 --> 00:49:20,500
URS GASSER: Speaking
of law, what's

993
00:49:20,500 --> 00:49:23,420
your hope that, looking
at your area of research,

994
00:49:23,420 --> 00:49:29,180
the law will evolve in this
dynamic situation, where

995
00:49:29,180 --> 00:49:31,970
maybe ethical principles
may lead the way?

996
00:49:31,970 --> 00:49:33,620
Where do you see
the promise of law

997
00:49:33,620 --> 00:49:36,140
in these debates,
where we're facing

998
00:49:36,140 --> 00:49:39,110
this shift from the human
towards the machine?

999
00:49:39,110 --> 00:49:42,170
CRYSTAL YANG: Yeah, I think
law has a very important role

1000
00:49:42,170 --> 00:49:42,950
to play here.

1001
00:49:42,950 --> 00:49:45,500
I think I share
Jeanette's general sense

1002
00:49:45,500 --> 00:49:48,020
that self-regulation
is probably not

1003
00:49:48,020 --> 00:49:50,480
going to be a sufficient
solution, that there

1004
00:49:50,480 --> 00:49:52,800
have to be legal interventions.

1005
00:49:52,800 --> 00:49:55,010
And the law is both
instrumental in

1006
00:49:55,010 --> 00:49:58,580
that it will undoubtedly,
by deciding what to permit

1007
00:49:58,580 --> 00:50:01,820
and what to prohibit, shape
the behavior of governments,

1008
00:50:01,820 --> 00:50:03,740
private companies,
in terms of how

1009
00:50:03,740 --> 00:50:07,280
they design algorithms, how they
implement them on the ground.

1010
00:50:07,280 --> 00:50:10,340
The law, I think, also has
important expressive principles

1011
00:50:10,340 --> 00:50:13,580
maybe related to ethics,
where if the law allows

1012
00:50:13,580 --> 00:50:16,430
for something, then
citizens, members of society,

1013
00:50:16,430 --> 00:50:19,410
will view something as maybe
more socially acceptable.

1014
00:50:19,410 --> 00:50:22,520
So I think the law here
has a big role to play.

1015
00:50:22,520 --> 00:50:24,650
Coming back to the criminal
justice system, though,

1016
00:50:24,650 --> 00:50:27,800
I think there are many ways
in which the current law,

1017
00:50:27,800 --> 00:50:29,930
certainly in the
United States, falls

1018
00:50:29,930 --> 00:50:33,290
short for a lot of the
new challenges that

1019
00:50:33,290 --> 00:50:35,180
might come with algorithms.

1020
00:50:35,180 --> 00:50:37,820
So to give you some
examples, many people

1021
00:50:37,820 --> 00:50:40,940
are troubled by the
use of disparities

1022
00:50:40,940 --> 00:50:44,480
that can emerge when you use
an algorithm to make decisions.

1023
00:50:44,480 --> 00:50:47,210
And that could be because
of the data or the structure

1024
00:50:47,210 --> 00:50:48,690
of the algorithm.

1025
00:50:48,690 --> 00:50:50,300
Now, it turns out
there's probably

1026
00:50:50,300 --> 00:50:53,300
pretty limited legal
remedies for addressing

1027
00:50:53,300 --> 00:50:54,950
those disparities.

1028
00:50:54,950 --> 00:50:58,610
Under current US law, a
finding of discrimination

1029
00:50:58,610 --> 00:51:02,030
under the Equal Protection
Clause of the US Constitution

1030
00:51:02,030 --> 00:51:05,390
would require a showing
of discriminatory intent

1031
00:51:05,390 --> 00:51:06,620
or purpose.

1032
00:51:06,620 --> 00:51:10,100
And that's hard, because when
an algorithmic designer chooses

1033
00:51:10,100 --> 00:51:12,870
to use a variable or
certain types of data,

1034
00:51:12,870 --> 00:51:15,590
there's probably often
no discriminatory intent

1035
00:51:15,590 --> 00:51:16,550
or purpose.

1036
00:51:16,550 --> 00:51:18,255
And yet because
so many variables

1037
00:51:18,255 --> 00:51:21,440
can be proxies for
things we're troubled by,

1038
00:51:21,440 --> 00:51:24,390
there's often maybe no
direct legal remedy.

1039
00:51:24,390 --> 00:51:26,120
And so this
traditional requirement

1040
00:51:26,120 --> 00:51:28,670
we've had in the US
Constitution and case

1041
00:51:28,670 --> 00:51:31,540
law of requiring
intent and motive

1042
00:51:31,540 --> 00:51:33,890
is often ill suited
to addressing

1043
00:51:33,890 --> 00:51:37,490
the new types of problems that
the algorithms can introduce.

1044
00:51:37,490 --> 00:51:39,890
Moreover, it's actually
been the case in the US

1045
00:51:39,890 --> 00:51:43,880
that many have interpreted
the case law on discrimination

1046
00:51:43,880 --> 00:51:47,450
as requiring or prohibiting
the use of characteristics

1047
00:51:47,450 --> 00:51:48,780
like race or ethnicity.

1048
00:51:48,780 --> 00:51:52,080
You can not use them in
any way, shape, or form.

1049
00:51:52,080 --> 00:51:55,610
But the reality is that because
of the complex statistical

1050
00:51:55,610 --> 00:52:00,230
relationships underlying many
variables, I, other computer

1051
00:52:00,230 --> 00:52:02,870
scientists, economists
have written and shown

1052
00:52:02,870 --> 00:52:06,080
that those proxy effects
that we may be worried about

1053
00:52:06,080 --> 00:52:09,890
are often created because of
the prohibition on the use

1054
00:52:09,890 --> 00:52:11,400
of those characteristics.

1055
00:52:11,400 --> 00:52:14,180
And that once you take
statistics into account,

1056
00:52:14,180 --> 00:52:17,330
you may actually want to use
protected characteristics

1057
00:52:17,330 --> 00:52:21,330
in certain forms to actually
remedy those disparities.

1058
00:52:21,330 --> 00:52:23,610
And so it's actually
this problem right now,

1059
00:52:23,610 --> 00:52:26,930
where I think the law is
pushing companies, governments

1060
00:52:26,930 --> 00:52:29,960
to develop versions of
algorithms that may actually

1061
00:52:29,960 --> 00:52:33,200
be counterproductive to
our larger societal goal

1062
00:52:33,200 --> 00:52:35,690
of equality and opportunity.

1063
00:52:35,690 --> 00:52:38,780
And I think, to the earlier
point about human decision

1064
00:52:38,780 --> 00:52:42,320
making, the law often does
not consider counterfactuals

1065
00:52:42,320 --> 00:52:43,850
in a very easy way.

1066
00:52:43,850 --> 00:52:47,510
It often seems to require
perfection for algorithms,

1067
00:52:47,510 --> 00:52:48,560
explainability.

1068
00:52:48,560 --> 00:52:51,380
But as you point out,
what is more a black box

1069
00:52:51,380 --> 00:52:53,390
than what is in a judge's mind?

1070
00:52:53,390 --> 00:52:56,300
Perhaps the judge's mind
is more of a black box

1071
00:52:56,300 --> 00:52:59,830
than a neural network or other
forms of machine learning.

1072
00:52:59,830 --> 00:53:03,710
And so I worry that the law, by
sometimes requiring perfection

1073
00:53:03,710 --> 00:53:06,170
and not considering
the counterfactual,

1074
00:53:06,170 --> 00:53:11,060
will often chill and deter what
may be innovative and good uses

1075
00:53:11,060 --> 00:53:14,310
of algorithmic decision making.

1076
00:53:14,310 --> 00:53:16,250
URS GASSER: So also
the relationship

1077
00:53:16,250 --> 00:53:19,910
between technology and
law and law and ethics

1078
00:53:19,910 --> 00:53:22,540
is very complicated
and bi-directional,

1079
00:53:22,540 --> 00:53:25,340
with unintended
consequences included.

1080
00:53:25,340 --> 00:53:27,160
FRANK-WALTER STEINMEIER:
[SPEAKING GERMAN]

1081
00:53:27,160 --> 00:53:32,100
INTERPRETER: Still,
if I may, some of you

1082
00:53:32,100 --> 00:53:38,940
have put this on the same, on
a par in a way that does not

1083
00:53:38,940 --> 00:53:43,160
convince me of any decision
taken by an algorithm

1084
00:53:43,160 --> 00:53:44,450
being plausible.

1085
00:53:44,450 --> 00:53:46,830
You know, the fact
that we don't always

1086
00:53:46,830 --> 00:53:52,080
understand algorithms
and software that

1087
00:53:52,080 --> 00:53:58,040
is guided by algorithms, that
leaves me very concerned.

1088
00:53:58,040 --> 00:54:00,030
We say, and the
situation in America

1089
00:54:00,030 --> 00:54:03,130
is slightly different from
the situation in Germany.

1090
00:54:03,130 --> 00:54:06,960
A judge passes a
sentence or a judgment,

1091
00:54:06,960 --> 00:54:10,140
he or she has to
justify that decision.

1092
00:54:10,140 --> 00:54:13,600
Not every ruling or
decision has to be accepted.

1093
00:54:13,600 --> 00:54:15,240
People may have a
different opinion.

1094
00:54:15,240 --> 00:54:17,950
But as a rule, as far as
the tradition in Germany

1095
00:54:17,950 --> 00:54:21,330
is concerned, you do have
a very extensive duty

1096
00:54:21,330 --> 00:54:24,690
to justify your
sentence or your ruling.

1097
00:54:24,690 --> 00:54:29,280
And that is what lacks when
you talk about algorithms.

1098
00:54:29,280 --> 00:54:31,650
That is one of the questions
that we need to discuss,

1099
00:54:31,650 --> 00:54:33,570
I believe.

1100
00:54:33,570 --> 00:54:39,290
Is it conceivable at
all that algorithms,

1101
00:54:39,290 --> 00:54:41,730
that the control
of our algorithms

1102
00:54:41,730 --> 00:54:44,280
can become or be made
more transparent?

1103
00:54:44,280 --> 00:54:48,310
Of course, not towards each
and every individual person,

1104
00:54:48,310 --> 00:54:51,720
but perhaps with
regard to those who

1105
00:54:51,720 --> 00:54:53,790
consider themselves
the representative

1106
00:54:53,790 --> 00:54:54,600
of the government.

1107
00:54:54,600 --> 00:54:58,440
You know, the body in question
responsible for protecting

1108
00:54:58,440 --> 00:55:03,300
the rights and the
freedoms of the individual.

1109
00:55:03,300 --> 00:55:04,290
Can you hear me?

1110
00:55:04,290 --> 00:55:07,110


1111
00:55:07,110 --> 00:55:09,390
And a second remark
that I'd like

1112
00:55:09,390 --> 00:55:11,190
to make against the
backdrop of what

1113
00:55:11,190 --> 00:55:15,990
has just been said, it is
good that we have a debate

1114
00:55:15,990 --> 00:55:19,290
up here on the
rostrum, so to speak,

1115
00:55:19,290 --> 00:55:24,580
about the ethical principles
for digital transformation.

1116
00:55:24,580 --> 00:55:26,400
But what struck
me, and I've tried

1117
00:55:26,400 --> 00:55:29,850
to refer to that in my
introductory remarks,

1118
00:55:29,850 --> 00:55:35,940
when I bring together a group
of experts in my office,

1119
00:55:35,940 --> 00:55:39,990
I have experts briefing me about
the technological potential

1120
00:55:39,990 --> 00:55:42,415
of AI.

1121
00:55:42,415 --> 00:55:45,030
But, you know, I have an idea
after these talks of what

1122
00:55:45,030 --> 00:55:46,760
is doable, what is conceivable.

1123
00:55:46,760 --> 00:55:50,580
But when I have talks about the
ethical limits of digitization,

1124
00:55:50,580 --> 00:55:52,980
it brings together a wholly
different group of people,

1125
00:55:52,980 --> 00:55:57,610
because as a rule, I do not
meet IT experts or engineers,

1126
00:55:57,610 --> 00:55:59,910
but I meet social
scientists, philosophers,

1127
00:55:59,910 --> 00:56:05,850
political scientists, which is
an indication, to some extent,

1128
00:56:05,850 --> 00:56:08,220
of something that keeps
me deeply troubled.

1129
00:56:08,220 --> 00:56:10,373
And that is that we
have a debate, not

1130
00:56:10,373 --> 00:56:14,200
that that debate takes
place within closed circles,

1131
00:56:14,200 --> 00:56:15,090
closed communities.

1132
00:56:15,090 --> 00:56:19,440
That is to say, we have a
debate about the ethical limits

1133
00:56:19,440 --> 00:56:24,330
modality of a digital
age and that we

1134
00:56:24,330 --> 00:56:26,820
have to bear in mind with
limits that we should

1135
00:56:26,820 --> 00:56:29,760
not to surpass or overstep.

1136
00:56:29,760 --> 00:56:33,540
We have a similar debate about
the function of democracy,

1137
00:56:33,540 --> 00:56:38,700
but it is not carried beyond
the respective communities.

1138
00:56:38,700 --> 00:56:39,870
Please tell me if I'm wrong.

1139
00:56:39,870 --> 00:56:41,940
I'm happy to hear you
point that out to me.

1140
00:56:41,940 --> 00:56:48,030
But as I see it, and
as I wish to see it,

1141
00:56:48,030 --> 00:56:50,080
the two communities that
I've been mentioning,

1142
00:56:50,080 --> 00:56:52,020
the tech community
on the one hand

1143
00:56:52,020 --> 00:56:56,770
and the more
philosophical community

1144
00:56:56,770 --> 00:56:59,100
bringing together social
scientists and philosophers,

1145
00:56:59,100 --> 00:57:02,120
we don't have a discussion that
brings both groups together.

1146
00:57:02,120 --> 00:57:04,120
We haven't been able to
link up that discussion.

1147
00:57:04,120 --> 00:57:06,007
Is that impression
that I have correct?

1148
00:57:06,007 --> 00:57:07,090
Would you agree with that?

1149
00:57:07,090 --> 00:57:09,630
Is it limited to
Germany, or would you

1150
00:57:09,630 --> 00:57:12,150
say that this is also
transferable to the debate

1151
00:57:12,150 --> 00:57:13,560
in the United States?

1152
00:57:13,560 --> 00:57:15,550
URS GASSER: Change that
right, over at MIT?

1153
00:57:15,550 --> 00:57:17,130
Can you can share your thoughts?

1154
00:57:17,130 --> 00:57:20,680
And then I would like to open
up for a number of questions.

1155
00:57:20,680 --> 00:57:22,220
So be ready with your questions.

1156
00:57:22,220 --> 00:57:22,590
MELISSA NOBLES: Sure.

1157
00:57:22,590 --> 00:57:24,450
So the Schwarzman, the
College of Computing,

1158
00:57:24,450 --> 00:57:26,140
which was announced
last year, is

1159
00:57:26,140 --> 00:57:28,170
intended to get at
just this issue,

1160
00:57:28,170 --> 00:57:31,430
that much of the technology
that is obviously being created

1161
00:57:31,430 --> 00:57:34,800
at MIT, we recognize
that there has

1162
00:57:34,800 --> 00:57:38,255
to be a bridge between
technology and the humanities

1163
00:57:38,255 --> 00:57:42,510
and the social sciences in an
intentional, deliberate way.

1164
00:57:42,510 --> 00:57:44,640
And part of why the
school was established

1165
00:57:44,640 --> 00:57:46,822
is, tons of students are
interested in computing.

1166
00:57:46,822 --> 00:57:47,530
They're doing it.

1167
00:57:47,530 --> 00:57:49,920
They're coming in, wanting
to major in computer science,

1168
00:57:49,920 --> 00:57:52,260
but many of them don't want to
only be computer scientists.

1169
00:57:52,260 --> 00:57:54,427
They want to apply that
knowledge to something else,

1170
00:57:54,427 --> 00:57:56,820
but they want it to be guided
by some domain knowledge

1171
00:57:56,820 --> 00:57:59,410
outside of computer science.

1172
00:57:59,410 --> 00:58:03,120
And so the goal of the college
is to eventually-- we're

1173
00:58:03,120 --> 00:58:06,450
beginning to see joint, blended
degrees between computer

1174
00:58:06,450 --> 00:58:09,600
science and economics, computer
science and urban studies,

1175
00:58:09,600 --> 00:58:11,940
computer science and music.

1176
00:58:11,940 --> 00:58:15,010
Not all students are doing
this, but the interest is great,

1177
00:58:15,010 --> 00:58:18,420
and it's intended to
allow for this connection

1178
00:58:18,420 --> 00:58:21,040
in a more organic way
from the beginning,

1179
00:58:21,040 --> 00:58:23,617
such that students will
have the type of skills,

1180
00:58:23,617 --> 00:58:25,950
so that we won't be talking
about disparate communities,

1181
00:58:25,950 --> 00:58:27,570
but that students
will have enough

1182
00:58:27,570 --> 00:58:30,780
of an openness, at least
an exposure understanding

1183
00:58:30,780 --> 00:58:33,855
that this is what it means
to be a computer scientist.

1184
00:58:33,855 --> 00:58:36,280
And conversely, for
my own discipline,

1185
00:58:36,280 --> 00:58:38,220
I'm a political
scientist, this is

1186
00:58:38,220 --> 00:58:40,160
what it means to be a
political scientist,

1187
00:58:40,160 --> 00:58:41,650
is to know something about this.

1188
00:58:41,650 --> 00:58:44,550
So all of us are going to
have to learn more and be

1189
00:58:44,550 --> 00:58:48,780
open to learning more, if we're
going to successfully deal

1190
00:58:48,780 --> 00:58:50,053
with this issue.

1191
00:58:50,053 --> 00:58:51,720
One other thing I'd
like to say about it

1192
00:58:51,720 --> 00:58:55,650
is, we're starting to have
these conversations on campus.

1193
00:58:55,650 --> 00:58:58,830
They are not easy
conversations to have.

1194
00:58:58,830 --> 00:59:01,560
As much as we try
to be collaborative,

1195
00:59:01,560 --> 00:59:04,470
we've really had to work.

1196
00:59:04,470 --> 00:59:06,960
We may be using the
same terms, but we

1197
00:59:06,960 --> 00:59:08,820
speak a different language.

1198
00:59:08,820 --> 00:59:11,850
And it requires
patience to do this.

1199
00:59:11,850 --> 00:59:14,280
So part of what
we're doing is also

1200
00:59:14,280 --> 00:59:17,460
learning some other principles
of generosity and patience

1201
00:59:17,460 --> 00:59:18,910
as we deal with one another.

1202
00:59:18,910 --> 00:59:21,320
Because if we want to
solve this problem,

1203
00:59:21,320 --> 00:59:24,235
deal with technology in a
way that we all want to see,

1204
00:59:24,235 --> 00:59:25,860
then that's what it's
going to require,

1205
00:59:25,860 --> 00:59:27,860
some other human qualities
that we have to bring

1206
00:59:27,860 --> 00:59:29,710
to bear for this to happen.

1207
00:59:29,710 --> 00:59:32,490
And I think especially
just inclusion

1208
00:59:32,490 --> 00:59:35,280
is really important for
us for our undergraduates,

1209
00:59:35,280 --> 00:59:39,310
since many of them will be
going into leadership positions.

1210
00:59:39,310 --> 00:59:40,750
They know the technologies.

1211
00:59:40,750 --> 00:59:42,750
We need them working on
the congressional staff.

1212
00:59:42,750 --> 00:59:46,770
If you all saw the hearings
with Mark Zuckerberg

1213
00:59:46,770 --> 00:59:50,933
in the Congress, people didn't
know what Google was, right?

1214
00:59:50,933 --> 00:59:52,350
Or they didn't
know the difference

1215
00:59:52,350 --> 00:59:53,465
between Samsung and Apple.

1216
00:59:53,465 --> 00:59:54,840
I mean, they didn't
know anything

1217
00:59:54,840 --> 00:59:56,280
about the technologies.

1218
00:59:56,280 --> 00:59:59,562
If they don't understand
how to open their phones,

1219
00:59:59,562 --> 01:00:01,020
how can you imagine
that they could

1220
01:00:01,020 --> 01:00:03,780
be responsible and entrusted
to do the kinds of things

1221
01:00:03,780 --> 01:00:05,820
that you all are describing?

1222
01:00:05,820 --> 01:00:09,300
Some of what we also
need, our students

1223
01:00:09,300 --> 01:00:13,990
to be able to play those kind
of roles, precisely because--

1224
01:00:13,990 --> 01:00:16,880
but we don't want them to do
it only knowing the technology.

1225
01:00:16,880 --> 01:00:19,000
They will also have to
understand economics, also

1226
01:00:19,000 --> 01:00:21,160
have to understand
political science, and such.

1227
01:00:21,160 --> 01:00:23,650
So that is the task
of the college.

1228
01:00:23,650 --> 01:00:25,753
And we're just getting
started, so stay tuned.

1229
01:00:25,753 --> 01:00:26,920
URS GASSER: That's exciting.

1230
01:00:26,920 --> 01:00:27,900
That's exciting.

1231
01:00:27,900 --> 01:00:28,950
Thank you.

1232
01:00:28,950 --> 01:00:31,900
OK, let's open up
for a few questions.

1233
01:00:31,900 --> 01:00:35,400
[? Becca, ?] our mic-runner
is ready and fast on her legs.

1234
01:00:35,400 --> 01:00:36,720
Who has a question?

1235
01:00:36,720 --> 01:00:39,240
And please end with
a question mark.

1236
01:00:39,240 --> 01:00:40,492
That would be good.

1237
01:00:40,492 --> 01:00:42,947
[LAUGHTER]

1238
01:00:42,947 --> 01:00:48,840


1239
01:00:48,840 --> 01:00:50,090
AUDIENCE: Me?

1240
01:00:50,090 --> 01:00:52,860
First thank you so much for
all the interesting insights

1241
01:00:52,860 --> 01:00:54,690
you shared today.

1242
01:00:54,690 --> 01:00:57,420
My question is, regarding
the fact that a lot of you

1243
01:00:57,420 --> 01:00:59,790
mentioned today that
there's an urgency

1244
01:00:59,790 --> 01:01:03,840
to craft tech specific
ethics regulations as soon

1245
01:01:03,840 --> 01:01:05,080
as possible.

1246
01:01:05,080 --> 01:01:07,470
So in a way, this is
really a moral discussion

1247
01:01:07,470 --> 01:01:08,760
with a deadline.

1248
01:01:08,760 --> 01:01:11,130
When would you say
is this deadline?

1249
01:01:11,130 --> 01:01:17,340
When do we have to formalize our
thoughts and put it into law?

1250
01:01:17,340 --> 01:01:20,700
EVA WEBER-GURSKA: Oh, was
that a question for me?

1251
01:01:20,700 --> 01:01:22,064
URS GASSER: Now it is, yes.

1252
01:01:22,064 --> 01:01:23,920
[LAUGHTER]

1253
01:01:23,920 --> 01:01:28,740
EVA WEBER-GURSKA: No, of course
I think the deadline is--

1254
01:01:28,740 --> 01:01:32,010
it's not far ahead, but
it's just right now.

1255
01:01:32,010 --> 01:01:35,580
But there are already a
lot of ethic guidelines

1256
01:01:35,580 --> 01:01:36,960
being written right now.

1257
01:01:36,960 --> 01:01:40,455
So we really have on national
level a different one

1258
01:01:40,455 --> 01:01:40,955
in Germany.

1259
01:01:40,955 --> 01:01:42,720
Then there are
international levels,

1260
01:01:42,720 --> 01:01:45,150
like, for example,
the high level export

1261
01:01:45,150 --> 01:01:51,600
group that got tasked by
the European Commission.

1262
01:01:51,600 --> 01:01:53,160
And they wrote
something, and then

1263
01:01:53,160 --> 01:01:56,880
we have already part of this
ethic guidelines, for example,

1264
01:01:56,880 --> 01:02:01,390
extracted, and the
G20 group signed it.

1265
01:02:01,390 --> 01:02:04,610
And so there are
already guidelines,

1266
01:02:04,610 --> 01:02:08,140
but I think the guidelines
are only the first step.

1267
01:02:08,140 --> 01:02:11,400
And then, of course,
they're our first step

1268
01:02:11,400 --> 01:02:14,320
to transform them into
law, as we heard already

1269
01:02:14,320 --> 01:02:15,840
with the [GERMAN].

1270
01:02:15,840 --> 01:02:21,460
And so yeah, it's right now,
and we should go further ahead.

1271
01:02:21,460 --> 01:02:24,425
But it's still already
something happening, I think.

1272
01:02:24,425 --> 01:02:25,300
URS GASSER: Wolfgang?

1273
01:02:25,300 --> 01:02:27,717
WOLFGANG SCHULZ: Yeah, maybe
I can add a legal perspective

1274
01:02:27,717 --> 01:02:30,480
to that, because one of
the problems of the law

1275
01:02:30,480 --> 01:02:34,900
is that the function of
law is to have stability.

1276
01:02:34,900 --> 01:02:38,350
And what we need here is
some flexibility as well.

1277
01:02:38,350 --> 01:02:40,830
And so what we are struggling
with, as legal scholars,

1278
01:02:40,830 --> 01:02:44,910
is to find ways to
make law more flexible,

1279
01:02:44,910 --> 01:02:47,780
to have constant evaluation,
to have sunset clauses

1280
01:02:47,780 --> 01:02:51,343
and things like that, so
that we do not have to wait.

1281
01:02:51,343 --> 01:02:53,760
I think it was Susan Crawford
said the marvelous sentence,

1282
01:02:53,760 --> 01:02:57,130
we have to regulate things
that we don't understand.

1283
01:02:57,130 --> 01:03:01,410
And I think we are-- can't
wait until all of the lawmakers

1284
01:03:01,410 --> 01:03:04,560
have understood
what it actually is.

1285
01:03:04,560 --> 01:03:05,910
You have to act before.

1286
01:03:05,910 --> 01:03:09,000
But then we need different
instruments, especially when

1287
01:03:09,000 --> 01:03:11,460
you take into account
what you said,

1288
01:03:11,460 --> 01:03:14,130
that when you are talking with
lawmakers, even if they really

1289
01:03:14,130 --> 01:03:17,250
try hard, the problems
are really high tech.

1290
01:03:17,250 --> 01:03:19,840
Only a couple of people really
understands what's happening

1291
01:03:19,840 --> 01:03:23,820
there, and even the best
member of parliament cannot

1292
01:03:23,820 --> 01:03:25,480
be an expert in this field.

1293
01:03:25,480 --> 01:03:29,850
So we need constant
evaluation and some mechanism

1294
01:03:29,850 --> 01:03:32,360
seems to deal with.

1295
01:03:32,360 --> 01:03:35,360
Every day, even
we as researchers,

1296
01:03:35,360 --> 01:03:39,290
every week I have a new
understanding of how algorithms

1297
01:03:39,290 --> 01:03:40,650
interact with society.

1298
01:03:40,650 --> 01:03:43,220
Every meeting, I have an
interaction with some software

1299
01:03:43,220 --> 01:03:45,590
engineer and oh, OK, it's
a little bit different

1300
01:03:45,590 --> 01:03:46,670
than I thought before.

1301
01:03:46,670 --> 01:03:50,420
And on this basis to create
law, that's, I would say,

1302
01:03:50,420 --> 01:03:52,955
a fundamentally new challenge.

1303
01:03:52,955 --> 01:03:54,705
FRANK-WALTER STEINMEIER:
[SPEAKING GERMAN]

1304
01:03:54,705 --> 01:03:57,810
INTERPRETER: But if I may
add to what you've just said,

1305
01:03:57,810 --> 01:03:59,510
the situation is changing.

1306
01:03:59,510 --> 01:04:01,950
When you take a look at
the German legal culture,

1307
01:04:01,950 --> 01:04:04,430
you know, it's based
on the assumption

1308
01:04:04,430 --> 01:04:06,980
that a law that is passed
is, until kingdom come,

1309
01:04:06,980 --> 01:04:09,140
it's for eternity, you know?

1310
01:04:09,140 --> 01:04:12,960
When we know look at the
area of internet law,

1311
01:04:12,960 --> 01:04:16,910
something interesting
is happening that is

1312
01:04:16,910 --> 01:04:18,440
quite difficult, complicated.

1313
01:04:18,440 --> 01:04:21,380
When you look to the
relationship between the one

1314
01:04:21,380 --> 01:04:24,230
passing the law, making
the law, and the public,

1315
01:04:24,230 --> 01:04:28,070
you know, public comments
about the Network Enforcement

1316
01:04:28,070 --> 01:04:31,220
Act in Germany, for example,
is an indication of that.

1317
01:04:31,220 --> 01:04:34,190
In some areas of legislation,
we have already reached a point

1318
01:04:34,190 --> 01:04:40,160
where we can no longer
provide an eternal guarantee

1319
01:04:40,160 --> 01:04:42,130
for the legislation
that is being passed.

1320
01:04:42,130 --> 01:04:46,130
We are taking one step,
hesitatingly after the other,

1321
01:04:46,130 --> 01:04:50,540
carefully trying out to see
how that intervention is

1322
01:04:50,540 --> 01:04:52,950
going to affect the
reality in the future.

1323
01:04:52,950 --> 01:04:56,000
This careful and cautious,
tentative approach

1324
01:04:56,000 --> 01:05:00,555
in legislation that is
taking place right now,

1325
01:05:00,555 --> 01:05:02,930
it's not something that is
being greeted with enthusiasm.

1326
01:05:02,930 --> 01:05:05,660
And I do understand, but
there might be no alternative

1327
01:05:05,660 --> 01:05:06,950
to that approach, you know.

1328
01:05:06,950 --> 01:05:10,490
Trying to time and
again refer back

1329
01:05:10,490 --> 01:05:13,120
from the instruments to the
technology and vice versa

1330
01:05:13,120 --> 01:05:15,260
and to amend things
when necessary.

1331
01:05:15,260 --> 01:05:15,475
URS GASSER: One
of the dark sides

1332
01:05:15,475 --> 01:05:17,183
of having such an
amazing group of people

1333
01:05:17,183 --> 01:05:21,200
is, we could discuss 20
minutes just one question.

1334
01:05:21,200 --> 01:05:24,570
So you want to jump in quickly
with comments, Jeanette

1335
01:05:24,570 --> 01:05:26,245
and Matthew?

1336
01:05:26,245 --> 01:05:28,370
JEANETTE HOFMANN: I thought
perhaps one way forward

1337
01:05:28,370 --> 01:05:33,860
could be to pursue a
more procedural approach

1338
01:05:33,860 --> 01:05:35,300
to these problems.

1339
01:05:35,300 --> 01:05:39,020
For example, think of
ways to holding companies

1340
01:05:39,020 --> 01:05:42,140
accountable to the kinds
of technology development

1341
01:05:42,140 --> 01:05:44,250
they tried to bring
to the market,

1342
01:05:44,250 --> 01:05:47,780
introducing auditing
requirements

1343
01:05:47,780 --> 01:05:50,810
for certain type of algorithms.

1344
01:05:50,810 --> 01:05:56,270
Make it mandatory to only
use in certain areas machine

1345
01:05:56,270 --> 01:05:58,790
learning systems that
are self-explainable,

1346
01:05:58,790 --> 01:06:03,680
that explain, in at
least basic ways,

1347
01:06:03,680 --> 01:06:06,590
how they come to
certain recommendations

1348
01:06:06,590 --> 01:06:07,790
and predictions.

1349
01:06:07,790 --> 01:06:09,950
That seems to be a
way forward, rather

1350
01:06:09,950 --> 01:06:14,397
than relying just on rules.

1351
01:06:14,397 --> 01:06:16,230
MATTHEW LIAO: Maybe I
can also jump in here.

1352
01:06:16,230 --> 01:06:19,730
So I think Professor
Gasser's mentioned

1353
01:06:19,730 --> 01:06:25,160
that there are about 130
ethical principles that

1354
01:06:25,160 --> 01:06:28,970
are being presented
by different companies

1355
01:06:28,970 --> 01:06:30,150
and so on and so forth.

1356
01:06:30,150 --> 01:06:34,500
I think what we really
also need is a rationale,

1357
01:06:34,500 --> 01:06:35,870
or philosophical--

1358
01:06:35,870 --> 01:06:38,180
this is a plug
for philosophers--

1359
01:06:38,180 --> 01:06:41,890
a philosophical justification
for some of these principles.

1360
01:06:41,890 --> 01:06:44,270
So for example, they
talk about-- a lot

1361
01:06:44,270 --> 01:06:47,390
of these principles say things
like "we need explainability."

1362
01:06:47,390 --> 01:06:48,860
Why do we need explainability?

1363
01:06:48,860 --> 01:06:55,280
I mean, we've heard some of the
panelists asking this question

1364
01:06:55,280 --> 01:06:56,760
and various other things.

1365
01:06:56,760 --> 01:06:59,690
And so here in that line,
I'm very sympathetic to what

1366
01:06:59,690 --> 01:07:01,070
Professor Hofmann's
saying, which

1367
01:07:01,070 --> 01:07:07,250
is this idea of the human rights
framework, which says that, we

1368
01:07:07,250 --> 01:07:08,970
need to look towards the goal.

1369
01:07:08,970 --> 01:07:10,760
What are these algorithms for?

1370
01:07:10,760 --> 01:07:14,280
Fundamentally, they are about
promoting human well-being,

1371
01:07:14,280 --> 01:07:14,780
right?

1372
01:07:14,780 --> 01:07:17,540
We want to make sure that we
have a harmonious society, one

1373
01:07:17,540 --> 01:07:19,950
that works towards all of us.

1374
01:07:19,950 --> 01:07:22,130
And so a human rights
framework, I think,

1375
01:07:22,130 --> 01:07:26,322
can really move
towards that goal.

1376
01:07:26,322 --> 01:07:27,530
And there's a rich tradition.

1377
01:07:27,530 --> 01:07:29,270
There's a rich
literature on these

1378
01:07:29,270 --> 01:07:32,720
philosophical justifications
of the different rights.

1379
01:07:32,720 --> 01:07:35,972
And they go beyond just
discrimination, right?

1380
01:07:35,972 --> 01:07:37,430
They say-- they're
positive rights.

1381
01:07:37,430 --> 01:07:41,540
They're rights where
we just-- it's not just

1382
01:07:41,540 --> 01:07:44,180
about making sure that you don't
discriminate, by making sure

1383
01:07:44,180 --> 01:07:47,420
that your technologies
also work to help people

1384
01:07:47,420 --> 01:07:48,930
and so on and so forth.

1385
01:07:48,930 --> 01:07:50,600
And the other thing
about human rights

1386
01:07:50,600 --> 01:07:54,720
is that it's an
obligation on everybody.

1387
01:07:54,720 --> 01:07:57,380
So it's not just an
obligation on the engineers.

1388
01:07:57,380 --> 01:07:59,880
It's not just an
obligation on the company.

1389
01:07:59,880 --> 01:08:02,430
It's not just an obligation
on the government.

1390
01:08:02,430 --> 01:08:04,170
It's an obligation on all of us.

1391
01:08:04,170 --> 01:08:05,720
We need to
collectively make sure

1392
01:08:05,720 --> 01:08:09,680
that we're working
towards making sure

1393
01:08:09,680 --> 01:08:13,947
that these technologies
work well for everybody.

1394
01:08:13,947 --> 01:08:14,780
URS GASSER: Crystal.

1395
01:08:14,780 --> 01:08:16,040
CRYSTAL YANG: I promise
to be very brief.

1396
01:08:16,040 --> 01:08:17,770
I know there are
a lot of hands up.

1397
01:08:17,770 --> 01:08:19,359
This is such a
fascinating question.

1398
01:08:19,359 --> 01:08:22,077
I think, I wholeheartedly
endorse the perspectives others

1399
01:08:22,077 --> 01:08:23,660
have raised, especially
Mr. President,

1400
01:08:23,660 --> 01:08:25,520
that the laws must be adaptive.

1401
01:08:25,520 --> 01:08:28,640
They must be flexible, because
we are still learning how they

1402
01:08:28,640 --> 01:08:30,939
work, how the algorithms work.

1403
01:08:30,939 --> 01:08:34,090
And so we have to also study
when a new regulation goes

1404
01:08:34,090 --> 01:08:35,062
into effect.

1405
01:08:35,062 --> 01:08:36,520
What does that mean
about the types

1406
01:08:36,520 --> 01:08:39,240
of algorithms we're seeing
that flourish after that?

1407
01:08:39,240 --> 01:08:42,670
What types of algorithms are
now disappearing as a result?

1408
01:08:42,670 --> 01:08:44,270
To that point about
explainability,

1409
01:08:44,270 --> 01:08:48,069
I think we all have this
desire to understand

1410
01:08:48,069 --> 01:08:49,939
what the algorithm is doing.

1411
01:08:49,939 --> 01:08:51,700
And so we often,
then, might shift

1412
01:08:51,700 --> 01:08:55,069
towards regulation or
principles that the algorithm

1413
01:08:55,069 --> 01:08:56,380
be explainable.

1414
01:08:56,380 --> 01:08:58,090
The complication
there is, now there's

1415
01:08:58,090 --> 01:09:01,330
emerging work coming from
computer science and economics,

1416
01:09:01,330 --> 01:09:04,540
showing that when you force an
algorithm to be explainable,

1417
01:09:04,540 --> 01:09:06,850
you generally will choose
an algorithm that's

1418
01:09:06,850 --> 01:09:10,220
simpler, because it has to
be easier to understand.

1419
01:09:10,220 --> 01:09:13,450
But a simpler algorithm, as
it turns out in some contexts,

1420
01:09:13,450 --> 01:09:16,569
actually can lead to both
less efficient results

1421
01:09:16,569 --> 01:09:20,140
and less equitable results,
which again raises a conundrum

1422
01:09:20,140 --> 01:09:22,689
that I think no field
alone can address,

1423
01:09:22,689 --> 01:09:25,899
but just reveals that there
are inherent trade-offs

1424
01:09:25,899 --> 01:09:28,899
every time we make a
choice, like explainability.

1425
01:09:28,899 --> 01:09:30,729
And we have to confront
those trade-offs

1426
01:09:30,729 --> 01:09:33,880
and decide how do we weigh
competing values, which

1427
01:09:33,880 --> 01:09:36,690
are inevitably going
to be at stake.

1428
01:09:36,690 --> 01:09:37,640
URS GASSER: Great.

1429
01:09:37,640 --> 01:09:40,151
So let's collect
three questions,

1430
01:09:40,151 --> 01:09:41,109
and then we'll respond.

1431
01:09:41,109 --> 01:09:42,819
One question here,
and then maybe one

1432
01:09:42,819 --> 01:09:44,430
from over here, this area.

1433
01:09:44,430 --> 01:09:47,529
AUDIENCE: Yeah, I'd like
to add an observation.

1434
01:09:47,529 --> 01:09:51,010
I think it's also
observable on the podium

1435
01:09:51,010 --> 01:09:54,069
that we are missing
economists, and we are

1436
01:09:54,069 --> 01:09:55,780
missing behavioral scientists.

1437
01:09:55,780 --> 01:09:58,120
And it seems to me that
these two components are

1438
01:09:58,120 --> 01:10:02,710
crucial in understanding
the impact that AI has had

1439
01:10:02,710 --> 01:10:05,680
and will have on our
society and each of us.

1440
01:10:05,680 --> 01:10:06,740
Why do I say this?

1441
01:10:06,740 --> 01:10:10,540
Because AI has enormous
economic potency.

1442
01:10:10,540 --> 01:10:13,900
In this country, the
majority of the productivity

1443
01:10:13,900 --> 01:10:16,060
of this country comes from AI.

1444
01:10:16,060 --> 01:10:19,390
And why is it that Facebook
and Google and other companies

1445
01:10:19,390 --> 01:10:23,080
have been doing undaunted
what they have been doing

1446
01:10:23,080 --> 01:10:25,370
is because exactly of that.

1447
01:10:25,370 --> 01:10:28,160
And so that is one reality
that we have to face.

1448
01:10:28,160 --> 01:10:32,560
And this reality is deeply
immersed in research as well.

1449
01:10:32,560 --> 01:10:34,840
Where is most of
our funding going?

1450
01:10:34,840 --> 01:10:38,770
It is going to computer science,
to computer engineering.

1451
01:10:38,770 --> 01:10:41,050
And then we have
some alibi, excuse

1452
01:10:41,050 --> 01:10:43,810
the term, addition
of social sciences,

1453
01:10:43,810 --> 01:10:46,360
and if we are lucky,
behavioral sciences.

1454
01:10:46,360 --> 01:10:47,000
There's no--

1455
01:10:47,000 --> 01:10:47,150
URS GASSER: Thank you.

1456
01:10:47,150 --> 01:10:48,370
Sorry, we have to stop there.

1457
01:10:48,370 --> 01:10:49,995
AUDIENCE: It's just--
it's very-- well,

1458
01:10:49,995 --> 01:10:51,910
we've heard a lot
from the panel.

1459
01:10:51,910 --> 01:10:52,680
[LAUGHTER]

1460
01:10:52,680 --> 01:10:53,680
URS GASSER: Fair enough.

1461
01:10:53,680 --> 01:10:57,010
AUDIENCE: I'm sorry, but
there is no level playing

1462
01:10:57,010 --> 01:10:59,770
field between the
behavioral sciences

1463
01:10:59,770 --> 01:11:02,380
and all the psychological
dynamics that

1464
01:11:02,380 --> 01:11:05,950
are opened up by AI
and computer scientists

1465
01:11:05,950 --> 01:11:07,420
and computer engineering.

1466
01:11:07,420 --> 01:11:11,170
Unless we change these
funding structures-- and we've

1467
01:11:11,170 --> 01:11:14,290
heard a lot about the
necessity for other regulations

1468
01:11:14,290 --> 01:11:16,570
for companies, but
these funding structures

1469
01:11:16,570 --> 01:11:18,880
have enormous consequences.

1470
01:11:18,880 --> 01:11:20,860
President Steinmeier
was asking, is there

1471
01:11:20,860 --> 01:11:24,460
any example about,
at the beginning

1472
01:11:24,460 --> 01:11:27,580
of such an enterprise, to
bring disciplines together?

1473
01:11:27,580 --> 01:11:28,720
I would say, yes.

1474
01:11:28,720 --> 01:11:31,240
And actually at the
Ruhr Universitat Bochum,

1475
01:11:31,240 --> 01:11:33,550
there is one
competence cluster that

1476
01:11:33,550 --> 01:11:36,520
focuses on cybersecurity,
which tries

1477
01:11:36,520 --> 01:11:40,000
to give level-headed,
equal importance

1478
01:11:40,000 --> 01:11:43,240
to social behavioral science
on the one, economics

1479
01:11:43,240 --> 01:11:46,390
and computer science and
computer engineering.

1480
01:11:46,390 --> 01:11:48,250
So I just hope
that in the future,

1481
01:11:48,250 --> 01:11:52,780
such discussions move beyond
very important contributions

1482
01:11:52,780 --> 01:11:56,410
from philosophers,
ethicists, and lawyers

1483
01:11:56,410 --> 01:12:00,755
to also have a brighter view.

1484
01:12:00,755 --> 01:12:03,210
[APPLAUSE]

1485
01:12:03,210 --> 01:12:05,233


1486
01:12:05,233 --> 01:12:06,650
URS GASSER: Not
really a question.

1487
01:12:06,650 --> 01:12:08,060
AUDIENCE: I promise to be short.

1488
01:12:08,060 --> 01:12:10,940
URS GASSER: Back up.

1489
01:12:10,940 --> 01:12:12,600
We collect there, please.

1490
01:12:12,600 --> 01:12:13,310
Go ahead, please.

1491
01:12:13,310 --> 01:12:14,685
AUDIENCE: I'll
keep myself short.

1492
01:12:14,685 --> 01:12:15,810
URS GASSER: Yup.

1493
01:12:15,810 --> 01:12:17,270
AUDIENCE: As we are at a
German-American conference

1494
01:12:17,270 --> 01:12:19,820
right now, I just want to ask,
how does the transatlantic

1495
01:12:19,820 --> 01:12:23,580
relationship help us in
solving all these challenges?

1496
01:12:23,580 --> 01:12:26,280
What is needed for an effective
transatlantic relationship,

1497
01:12:26,280 --> 01:12:28,340
especially the
German-American one,

1498
01:12:28,340 --> 01:12:30,500
to solve all these
challenges together,

1499
01:12:30,500 --> 01:12:34,850
as like a world and not
as like separate states?

1500
01:12:34,850 --> 01:12:35,780
URS GASSER: OK.

1501
01:12:35,780 --> 01:12:36,920
Maybe one or two more?

1502
01:12:36,920 --> 01:12:37,580
Yes, please.

1503
01:12:37,580 --> 01:12:41,150


1504
01:12:41,150 --> 01:12:42,352
AUDIENCE: I have a question.

1505
01:12:42,352 --> 01:12:44,040
[LAUGHTER]

1506
01:12:44,040 --> 01:12:45,470
URS GASSER: I appreciate that.

1507
01:12:45,470 --> 01:12:49,660
AUDIENCE: Do you think,
talking about social media,

1508
01:12:49,660 --> 01:12:51,140
that the time will
come that there

1509
01:12:51,140 --> 01:12:54,980
will be a reliable algorithm
to identify hate speech?

1510
01:12:54,980 --> 01:12:58,580


1511
01:12:58,580 --> 01:13:00,317
URS GASSER: OK, over there.

1512
01:13:00,317 --> 01:13:00,900
Last question.

1513
01:13:00,900 --> 01:13:03,120
AUDIENCE: I promised
a short question.

1514
01:13:03,120 --> 01:13:05,540
AUDIENCE: I'm one of those
computer scientists writing

1515
01:13:05,540 --> 01:13:08,360
those messy
algorithms, and I know

1516
01:13:08,360 --> 01:13:09,900
that there are
people in my field

1517
01:13:09,900 --> 01:13:11,870
who think very
critical about this.

1518
01:13:11,870 --> 01:13:13,560
And I know there's
a lot of discussion.

1519
01:13:13,560 --> 01:13:16,040
So I can convince
you there are people

1520
01:13:16,040 --> 01:13:18,690
behind the current
talking about the things.

1521
01:13:18,690 --> 01:13:21,650
How can we reach
out to other people

1522
01:13:21,650 --> 01:13:23,090
who are thinking about this?

1523
01:13:23,090 --> 01:13:25,650


1524
01:13:25,650 --> 01:13:27,810
AUDIENCE: So my
question is short.

1525
01:13:27,810 --> 01:13:29,080
Thank you all.

1526
01:13:29,080 --> 01:13:31,900
This was delightful.

1527
01:13:31,900 --> 01:13:35,880
I would like to understand, are
we being human race empowered

1528
01:13:35,880 --> 01:13:36,840
by technology?

1529
01:13:36,840 --> 01:13:39,660
Or are we powering
technology by humans?

1530
01:13:39,660 --> 01:13:43,290


1531
01:13:43,290 --> 01:13:44,040
URS GASSER: Great.

1532
01:13:44,040 --> 01:13:45,090
All right.

1533
01:13:45,090 --> 01:13:48,210
We have 12 minutes left.

1534
01:13:48,210 --> 01:13:49,620
And I'm Swiss, as I said.

1535
01:13:49,620 --> 01:13:51,960
So I want to end on time.

1536
01:13:51,960 --> 01:13:56,290
So what I would suggest is that
we actually do a closing round

1537
01:13:56,290 --> 01:14:00,510
and pick the question that
you would like to address,

1538
01:14:00,510 --> 01:14:03,360
but put it into the
context also of your work

1539
01:14:03,360 --> 01:14:05,470
and what we've discussed here.

1540
01:14:05,470 --> 01:14:08,490
So we have the question of
transatlantic relationships.

1541
01:14:08,490 --> 01:14:13,016
We have the question
around social media

1542
01:14:13,016 --> 01:14:15,710
and the role of
technology in creating

1543
01:14:15,710 --> 01:14:19,890
a safer environment using
the example of hate speech.

1544
01:14:19,890 --> 01:14:22,440
And we have this
ultimate question,

1545
01:14:22,440 --> 01:14:25,440
is technology empowering
people, or are people here

1546
01:14:25,440 --> 01:14:27,670
somehow to empower technology.

1547
01:14:27,670 --> 01:14:31,080
So these are a
few of the themes.

1548
01:14:31,080 --> 01:14:33,430
Perhaps we start with Eva.

1549
01:14:33,430 --> 01:14:35,430
EVA WEBER-GURSKA: Mm-hm.

1550
01:14:35,430 --> 01:14:39,060
Yeah, maybe to the
two last questions.

1551
01:14:39,060 --> 01:14:41,790
Of course, I think that the
technology should empower

1552
01:14:41,790 --> 01:14:45,383
people, but for that
human machine interaction,

1553
01:14:45,383 --> 01:14:47,300
it is important that we
understand each other,

1554
01:14:47,300 --> 01:14:48,425
as we already talked about.

1555
01:14:48,425 --> 01:14:54,000
And maybe just into two aspects
that philosophy can contribute

1556
01:14:54,000 --> 01:14:59,040
here, and the one is, it's
not only about explainability

1557
01:14:59,040 --> 01:15:01,020
that you said, and it's
not only what is also

1558
01:15:01,020 --> 01:15:04,780
important about the ethic or
moral justification of why it's

1559
01:15:04,780 --> 01:15:08,760
so important to explain, but
also to see that in morality,

1560
01:15:08,760 --> 01:15:10,750
for example, it's
about reason giving.

1561
01:15:10,750 --> 01:15:16,530
So the whole validity
of moral norms

1562
01:15:16,530 --> 01:15:18,990
depends, I think, on
the fact that they

1563
01:15:18,990 --> 01:15:21,450
exist between
beings that can give

1564
01:15:21,450 --> 01:15:23,430
reasons and understand reasons.

1565
01:15:23,430 --> 01:15:28,095
And so one question would be, do
we want algorithmic structures

1566
01:15:28,095 --> 01:15:32,250
who cannot give reasons in an
empathic sense, for example.

1567
01:15:32,250 --> 01:15:35,700
And another topic would
be the topic of trust,

1568
01:15:35,700 --> 01:15:39,150
because the ethic
guidelines often highlight

1569
01:15:39,150 --> 01:15:41,870
trustworthy AI as a claim.

1570
01:15:41,870 --> 01:15:46,080
I would also be skeptical
of this as a best aim,

1571
01:15:46,080 --> 01:15:48,960
because trustworthiness
presupposes

1572
01:15:48,960 --> 01:15:53,730
also being a moral
subject, because trust

1573
01:15:53,730 --> 01:15:57,900
means to believe that someone
will hold to his or her

1574
01:15:57,900 --> 01:15:59,500
commitment to do something.

1575
01:15:59,500 --> 01:16:01,360
And this is also
something that is only

1576
01:16:01,360 --> 01:16:03,040
possible for a moral subject.

1577
01:16:03,040 --> 01:16:07,340
So AI systems cannot be
trustworthy agents or subjects.

1578
01:16:07,340 --> 01:16:08,257
URS GASSER: Thank you.

1579
01:16:08,257 --> 01:16:09,450
Matthew?

1580
01:16:09,450 --> 01:16:12,270
MATTHEW LIAO: Yes, I'll take
the question on hate speech.

1581
01:16:12,270 --> 01:16:17,430
So I mean, there are some
attempts using machine learning

1582
01:16:17,430 --> 01:16:22,170
algorithms to detect things like
fake news and things like that,

1583
01:16:22,170 --> 01:16:25,410
but I actually want to give
you a very grim picture.

1584
01:16:25,410 --> 01:16:28,710
This is like election
2.0, since we're coming up

1585
01:16:28,710 --> 01:16:30,580
to another election cycle.

1586
01:16:30,580 --> 01:16:31,740
So there's some evidence.

1587
01:16:31,740 --> 01:16:33,720
There's something
called deep fake, which

1588
01:16:33,720 --> 01:16:39,510
is being able to produce all
these videos that just look

1589
01:16:39,510 --> 01:16:42,660
like they can
superimpose your photo

1590
01:16:42,660 --> 01:16:44,860
or you onto another video.

1591
01:16:44,860 --> 01:16:48,930
And then they can get you to
talk and do various things.

1592
01:16:48,930 --> 01:16:51,450
And what people are
finding is that--

1593
01:16:51,450 --> 01:16:54,240
so there's the theory
that you tend to vote

1594
01:16:54,240 --> 01:16:56,895
for people who look like you.

1595
01:16:56,895 --> 01:17:01,860
OK, and so now they're creating
deep fake videos, where

1596
01:17:01,860 --> 01:17:07,170
they super impose your picture
onto a candidate's picture,

1597
01:17:07,170 --> 01:17:08,160
so it looks like you.

1598
01:17:08,160 --> 01:17:11,250
And now supposedly that's
going to influence your voting

1599
01:17:11,250 --> 01:17:13,320
behavior, because you're
more likely to vote

1600
01:17:13,320 --> 01:17:14,850
for people who look like you.

1601
01:17:14,850 --> 01:17:19,860
And so that's going to be
very worrying in the future.

1602
01:17:19,860 --> 01:17:23,040
And then the question is,
we're going to get to a point

1603
01:17:23,040 --> 01:17:25,490
where it's going to be
very hard for human eyes

1604
01:17:25,490 --> 01:17:27,420
to be able to detect
those differences,

1605
01:17:27,420 --> 01:17:29,690
and that's going to
be very worrying.

1606
01:17:29,690 --> 01:17:30,950
URS GASSER: Jeanette.

1607
01:17:30,950 --> 01:17:32,700
JEANETTE HOFMANN: I'd
also like to pick up

1608
01:17:32,700 --> 01:17:35,860
the question on hate speech.

1609
01:17:35,860 --> 01:17:38,460
What I find really good
about this question

1610
01:17:38,460 --> 01:17:40,710
is because we have
so many examples

1611
01:17:40,710 --> 01:17:49,110
that show how deeply ambiguous
we are about such wording.

1612
01:17:49,110 --> 01:17:52,840
Facebook once told me the
example of the term "bitch."

1613
01:17:52,840 --> 01:17:54,690
"Bitch" can be
really dismissive,

1614
01:17:54,690 --> 01:17:56,610
when you call a woman a bitch.

1615
01:17:56,610 --> 01:17:58,890
But nowadays, in
some circles, "bitch"

1616
01:17:58,890 --> 01:18:00,600
can also be appreciative.

1617
01:18:00,600 --> 01:18:05,070
Women might refer to
each other as bitches.

1618
01:18:05,070 --> 01:18:07,413
How is Facebook
supposed to regulate--

1619
01:18:07,413 --> 01:18:09,330
[? AUDIENCE: ?] Especially
on [? Halloween. ?]

1620
01:18:09,330 --> 01:18:11,122
JEANETTE HOFMANN: I
mean, how is Facebook--

1621
01:18:11,122 --> 01:18:12,960
URS GASSER: I did
not see that coming.

1622
01:18:12,960 --> 01:18:13,890
JEANETTE HOFMANN: --supposed
to regulate wordings that

1623
01:18:13,890 --> 01:18:15,870
have so different meanings?

1624
01:18:15,870 --> 01:18:18,930
And that, I think,
shows also the limit,

1625
01:18:18,930 --> 01:18:23,860
the limits of technical
filtering of language.

1626
01:18:23,860 --> 01:18:26,400
Language is changing
all the time,

1627
01:18:26,400 --> 01:18:29,730
and it differs across
cultures also very much.

1628
01:18:29,730 --> 01:18:31,050
So there are really limits.

1629
01:18:31,050 --> 01:18:35,940
Another point, if I may,
the question of empowering

1630
01:18:35,940 --> 01:18:38,130
versus disempowering.

1631
01:18:38,130 --> 01:18:41,500
I really like this question,
because it implicitly refers

1632
01:18:41,500 --> 01:18:44,240
to autonomy of human beings.

1633
01:18:44,240 --> 01:18:47,710
I think it's a mistake to
think autonomy needs to be

1634
01:18:47,710 --> 01:18:49,630
defended against technology.

1635
01:18:49,630 --> 01:18:53,800
Technology, in many ways,
enhances our autonomy.

1636
01:18:53,800 --> 01:18:55,700
Think of flying around.

1637
01:18:55,700 --> 01:18:56,860
Think of your watch.

1638
01:18:56,860 --> 01:18:59,170
I mean, we coordinate
as societies

1639
01:18:59,170 --> 01:19:00,940
through these technologies.

1640
01:19:00,940 --> 01:19:03,770
And at the same time, they
are disciplining ourselves.

1641
01:19:03,770 --> 01:19:05,290
So it's not an either/or.

1642
01:19:05,290 --> 01:19:09,040
And technologies and human
beings are not opposites.

1643
01:19:09,040 --> 01:19:12,190
It's the matter of how
we structure and shape

1644
01:19:12,190 --> 01:19:15,470
the relationship
between the two.

1645
01:19:15,470 --> 01:19:17,910
URS GASSER: Dean Nobles, is
it OK if we go last with you?

1646
01:19:17,910 --> 01:19:19,658
Mr. President, Dean Nobles.

1647
01:19:19,658 --> 01:19:21,700
MELISSA NOBLES: Sure, I'll
just take the question

1648
01:19:21,700 --> 01:19:23,710
of the computer
scientist who said

1649
01:19:23,710 --> 01:19:26,620
these kind of conversations
are also happening

1650
01:19:26,620 --> 01:19:29,170
among computer scientists.

1651
01:19:29,170 --> 01:19:33,670
But we need to get better
connection with others, who are

1652
01:19:33,670 --> 01:19:35,380
thinking critically about it.

1653
01:19:35,380 --> 01:19:38,600
I think that obviously education
plays a hugely important role

1654
01:19:38,600 --> 01:19:39,100
in this.

1655
01:19:39,100 --> 01:19:41,440
That is, early on
getting students

1656
01:19:41,440 --> 01:19:43,360
of different disciplines
to work together

1657
01:19:43,360 --> 01:19:45,880
and to learn together
in a way that addresses

1658
01:19:45,880 --> 01:19:48,250
these questions exactly.

1659
01:19:48,250 --> 01:19:50,530
The challenge of all
knowledge is making sure

1660
01:19:50,530 --> 01:19:52,900
that it doesn't stay
siloed, that we work

1661
01:19:52,900 --> 01:19:55,180
in a truly collaborative way.

1662
01:19:55,180 --> 01:19:57,920
And it seems to me that's the
challenge for the 21st century.

1663
01:19:57,920 --> 01:20:01,220


1664
01:20:01,220 --> 01:20:02,780
WOLFGANG SCHULZ:
Yeah, I think I'll

1665
01:20:02,780 --> 01:20:07,140
pick the comment, if
I may, on the funding

1666
01:20:07,140 --> 01:20:09,020
and interdisciplinary research.

1667
01:20:09,020 --> 01:20:11,660
I think we talk a lot about
interdisciplinary research,

1668
01:20:11,660 --> 01:20:13,760
and we need it to
solve problems,

1669
01:20:13,760 --> 01:20:15,920
but the academic
system is not really

1670
01:20:15,920 --> 01:20:18,740
designed to cater that need.

1671
01:20:18,740 --> 01:20:20,180
We still have
problems with that,

1672
01:20:20,180 --> 01:20:23,110
and I constantly get phone
calls from colleagues

1673
01:20:23,110 --> 01:20:27,070
that want to apply for
project funding next week.

1674
01:20:27,070 --> 01:20:29,790
And they said, oh, we have just
seen we need some ethics in it,

1675
01:20:29,790 --> 01:20:32,870
and we need a lawyer
or something like that.

1676
01:20:32,870 --> 01:20:34,590
Would you be available?

1677
01:20:34,590 --> 01:20:37,190
And normally I'd
say no, because it

1678
01:20:37,190 --> 01:20:39,640
has to be part of
the project question

1679
01:20:39,640 --> 01:20:41,900
and not just an icing
on the cake that

1680
01:20:41,900 --> 01:20:44,990
has already been baked.

1681
01:20:44,990 --> 01:20:46,590
That makes no sense.

1682
01:20:46,590 --> 01:20:50,030
And so I think we have issues
here in the academic system.

1683
01:20:50,030 --> 01:20:53,120
And maybe 30 seconds on
the transatlantic issue.

1684
01:20:53,120 --> 01:20:59,060
I think it's really helpful
and good for these questions

1685
01:20:59,060 --> 01:21:01,040
that there are really
stable research

1686
01:21:01,040 --> 01:21:03,860
relationships between
our American colleagues

1687
01:21:03,860 --> 01:21:06,755
and the research in Germany.

1688
01:21:06,755 --> 01:21:09,050
It's really great,
and that survives

1689
01:21:09,050 --> 01:21:13,390
even if there is a
political winter or autumn,

1690
01:21:13,390 --> 01:21:14,858
that we have this relationship.

1691
01:21:14,858 --> 01:21:16,400
And to solve these
kinds of problems,

1692
01:21:16,400 --> 01:21:18,257
I think that's
extremely helpful.

1693
01:21:18,257 --> 01:21:20,090
CRYSTAL YANG: Yeah,
I'll just follow up also

1694
01:21:20,090 --> 01:21:22,632
on the research question to the
excellent question over here.

1695
01:21:22,632 --> 01:21:24,710
I failed to mention, I
am actually an economist

1696
01:21:24,710 --> 01:21:25,610
as well as a lawyer.

1697
01:21:25,610 --> 01:21:29,300
And I would welcome many more
economists studying this area,

1698
01:21:29,300 --> 01:21:31,640
and I hope that the
funding structures as well

1699
01:21:31,640 --> 01:21:34,850
as the incentives do promote
that greater collaboration.

1700
01:21:34,850 --> 01:21:37,070
I think the computer
science community

1701
01:21:37,070 --> 01:21:38,390
is doing amazing work.

1702
01:21:38,390 --> 01:21:41,210
It's often siloed from what
the economics community

1703
01:21:41,210 --> 01:21:43,180
is thinking about, what
the legal community

1704
01:21:43,180 --> 01:21:43,930
is thinking about.

1705
01:21:43,930 --> 01:21:47,030
And so I think initiatives
like what Dean Nobles is doing

1706
01:21:47,030 --> 01:21:51,110
is probably a really great way
of bringing people together.

1707
01:21:51,110 --> 01:21:53,870
And to education,
there is such a need

1708
01:21:53,870 --> 01:21:57,890
for infusing this type of
learning in legal systems.

1709
01:21:57,890 --> 01:22:00,710
And I don't think that
US law schools, at least,

1710
01:22:00,710 --> 01:22:03,080
have really been at
the forefront of this.

1711
01:22:03,080 --> 01:22:06,080
In fact, many of the
decisions you read from state

1712
01:22:06,080 --> 01:22:09,380
supreme court judges who are
ruling on the use of algorithms

1713
01:22:09,380 --> 01:22:13,580
and making important case law
have explicit acknowledgments.

1714
01:22:13,580 --> 01:22:17,500
I'm paraphrasing here, but I'm
not paraphrasing so far of,

1715
01:22:17,500 --> 01:22:19,797
the judges in this case were
limited in their decision

1716
01:22:19,797 --> 01:22:21,380
making, because they
didn't understand

1717
01:22:21,380 --> 01:22:22,910
how the algorithm worked.

1718
01:22:22,910 --> 01:22:24,960
Well, that's a
really big problem.

1719
01:22:24,960 --> 01:22:27,185
And so we need to
train the lawyers who

1720
01:22:27,185 --> 01:22:30,140
will be deciding these cases,
working on behalf of clients

1721
01:22:30,140 --> 01:22:32,240
who are both creators
of algorithms

1722
01:22:32,240 --> 01:22:34,310
and individuals
adversely affected

1723
01:22:34,310 --> 01:22:39,030
by algorithms to understand
how algorithms work.

1724
01:22:39,030 --> 01:22:40,310
URS GASSER: [SPEAKING GERMAN]

1725
01:22:40,310 --> 01:22:41,480


1726
01:22:41,480 --> 01:22:44,370
INTERPRETER: Mr. President,
you have the final word.

1727
01:22:44,370 --> 01:22:44,870
Thank you.

1728
01:22:44,870 --> 01:22:45,860
Thank you, indeed.

1729
01:22:45,860 --> 01:22:48,500
I'm not attempting,
even attempting

1730
01:22:48,500 --> 01:22:51,190
to respond to all the
questions that were put to us,

1731
01:22:51,190 --> 01:22:53,780
but allow me begin by
the following remark.

1732
01:22:53,780 --> 01:22:57,410
The debate that we have
just been witnessing

1733
01:22:57,410 --> 01:23:00,460
with the participation
of the audience

1734
01:23:00,460 --> 01:23:03,350
would undoubtedly be
easier in the future

1735
01:23:03,350 --> 01:23:05,900
if we were to keep it clear
from any misunderstandings.

1736
01:23:05,900 --> 01:23:07,610
If I may come back
to the question

1737
01:23:07,610 --> 01:23:10,357
that you put at
the beginning, why

1738
01:23:10,357 --> 01:23:15,590
is there no economist
amongst people here?

1739
01:23:15,590 --> 01:23:18,230
You know the economic
potential of IT

1740
01:23:18,230 --> 01:23:21,980
and artificial intelligence
is being seen sufficiently,

1741
01:23:21,980 --> 01:23:22,540
I believe.

1742
01:23:22,540 --> 01:23:26,140
So if you take a
look at the expert,

1743
01:23:26,140 --> 01:23:28,220
the group of experts
brought together here,

1744
01:23:28,220 --> 01:23:30,620
you will undoubtedly
find, confirm

1745
01:23:30,620 --> 01:23:33,020
that everyone here is aware
of the economic potential.

1746
01:23:33,020 --> 01:23:35,300
Everyone is aware of the
technological potential.

1747
01:23:35,300 --> 01:23:37,490
Everyone is aware of
the potential that

1748
01:23:37,490 --> 01:23:41,120
exists when it comes to
fighting poverty, fighting

1749
01:23:41,120 --> 01:23:47,158
disease, fighting the
impact of climate change.

1750
01:23:47,158 --> 01:23:48,950
If we want to be
successful in those areas,

1751
01:23:48,950 --> 01:23:53,300
we need experts
at the top level.

1752
01:23:53,300 --> 01:23:56,205
And we in Germany intend to
participate in that development

1753
01:23:56,205 --> 01:23:58,220
just as much as you do.

1754
01:23:58,220 --> 01:24:03,750
But that is a kind
of advance remark.

1755
01:24:03,750 --> 01:24:05,700
I want to be very
clear, having said

1756
01:24:05,700 --> 01:24:09,110
what I've said doesn't
mean that we end up

1757
01:24:09,110 --> 01:24:15,080
in an age of unbridled
regulation, a crazy approach

1758
01:24:15,080 --> 01:24:16,230
towards--

1759
01:24:16,230 --> 01:24:18,080
a craze about regulation.

1760
01:24:18,080 --> 01:24:20,210
When you look at
the field of tension

1761
01:24:20,210 --> 01:24:25,450
between new technologies
and AI on the one hand

1762
01:24:25,450 --> 01:24:27,770
and what is the constituent
element of our societies,

1763
01:24:27,770 --> 01:24:30,080
and that is democratic
decision making

1764
01:24:30,080 --> 01:24:32,350
processes in Western societies.

1765
01:24:32,350 --> 01:24:34,190
Their is a field of tension.

1766
01:24:34,190 --> 01:24:36,830
Shouldn't we make that also
a topic of the discussion

1767
01:24:36,830 --> 01:24:37,750
every once in a while?

1768
01:24:37,750 --> 01:24:40,100
And that is why I suggested
to make that the topic

1769
01:24:40,100 --> 01:24:41,710
of our discussion today.

1770
01:24:41,710 --> 01:24:47,180
So no one should assume or be
afraid that this inherently

1771
01:24:47,180 --> 01:24:54,490
entails a secret wish to in some
way influence the development

1772
01:24:54,490 --> 01:24:58,020
or to slow down the
developments in the field of AI

1773
01:24:58,020 --> 01:25:00,410
and technologies
of digitization.

1774
01:25:00,410 --> 01:25:02,065
But that's not my
intention, really,

1775
01:25:02,065 --> 01:25:03,740
but there is this field
of tension I mentioned,

1776
01:25:03,740 --> 01:25:04,940
and we have to focus on it.

1777
01:25:04,940 --> 01:25:06,470
We have to deal
with it, and this

1778
01:25:06,470 --> 01:25:08,540
is equally true
for all those who

1779
01:25:08,540 --> 01:25:12,710
participate in the process
of technological development

1780
01:25:12,710 --> 01:25:15,950
of these means of communication.

1781
01:25:15,950 --> 01:25:19,040
This should not be
left to philosophers

1782
01:25:19,040 --> 01:25:20,540
or individual groups.

1783
01:25:20,540 --> 01:25:23,220
It has to be viewed as
a topic for all of us.

1784
01:25:23,220 --> 01:25:28,460
And if we pursue such an
approach, we will, I believe,

1785
01:25:28,460 --> 01:25:31,142
reach a point, and that
has become obvious here

1786
01:25:31,142 --> 01:25:32,600
as a consequence
of the discussion,

1787
01:25:32,600 --> 01:25:35,630
where we don't leave
it to appealing

1788
01:25:35,630 --> 01:25:38,330
to the morals and each
of every individual

1789
01:25:38,330 --> 01:25:41,390
and his or her responsibility.

1790
01:25:41,390 --> 01:25:46,700
But we need to have a debate
across borders, whether there

1791
01:25:46,700 --> 01:25:49,340
should be limits to
technological process

1792
01:25:49,340 --> 01:25:52,250
that we should not
surpass, because this,

1793
01:25:52,250 --> 01:25:55,430
at the end of the day,
is what it is all about.

1794
01:25:55,430 --> 01:25:59,150
It's difficult enough when you
look at Germany and the United

1795
01:25:59,150 --> 01:26:02,120
States of America, but it will
become even more difficult

1796
01:26:02,120 --> 01:26:07,130
when you think about
those countries that

1797
01:26:07,130 --> 01:26:14,210
have a completely different
social system or approach.

1798
01:26:14,210 --> 01:26:15,950
But we need to have that debate.

1799
01:26:15,950 --> 01:26:18,730
We need to have it with
a country like China.

1800
01:26:18,730 --> 01:26:21,950
And in saying that, I'm
not cherishing any illusion

1801
01:26:21,950 --> 01:26:27,280
about us having in 10 or five
years a kind of UN charter

1802
01:26:27,280 --> 01:26:28,720
on artificial intelligence.

1803
01:26:28,720 --> 01:26:30,240
We won't get that.

1804
01:26:30,240 --> 01:26:32,190
But nevertheless,
we should engage

1805
01:26:32,190 --> 01:26:35,990
in that kind of a
debate, just as much

1806
01:26:35,990 --> 01:26:39,890
as we have a debate
with China although we

1807
01:26:39,890 --> 01:26:42,680
have different views on
the issues of bioethics

1808
01:26:42,680 --> 01:26:45,440
and genetic engineering.

1809
01:26:45,440 --> 01:26:47,430
We are not in agreement
on these issues,

1810
01:26:47,430 --> 01:26:49,100
but nevertheless
we have succeeded

1811
01:26:49,100 --> 01:26:53,660
in defining some limits or
ceilings or restrictions.

1812
01:26:53,660 --> 01:26:56,450


1813
01:26:56,450 --> 01:26:59,370
Thus I am not discouraged,
you know, in any way,

1814
01:26:59,370 --> 01:27:01,370
when I look to the
possibility of such a debate,

1815
01:27:01,370 --> 01:27:05,000
although it's going to
be a complicated one.

1816
01:27:05,000 --> 01:27:08,530
But you know, this
was mentioned.

1817
01:27:08,530 --> 01:27:12,720
What we need is a transatlantic
debate on the subject matter,

1818
01:27:12,720 --> 01:27:13,220
too.

1819
01:27:13,220 --> 01:27:16,240


1820
01:27:16,240 --> 01:27:19,940
Apart from all the topics of the
day, the conflicts of the day,

1821
01:27:19,940 --> 01:27:22,930
and I don't want to
downplay their importance,

1822
01:27:22,930 --> 01:27:25,435
but we have to tackle the
question of the importance

1823
01:27:25,435 --> 01:27:26,810
of the freedom of
the individual,

1824
01:27:26,810 --> 01:27:30,290
of the democratic culture in
the states of the Western world.

1825
01:27:30,290 --> 01:27:33,290
And we count ourselves
amongst those, just as well as

1826
01:27:33,290 --> 01:27:34,790
the United States
of America, and we

1827
01:27:34,790 --> 01:27:38,100
need to have that debate amongst
the West first and foremost.

1828
01:27:38,100 --> 01:27:41,550
This is what I would wish, us
to have the opportunity, time

1829
01:27:41,550 --> 01:27:45,050
and again, as I have been trying
to seek it during my visit

1830
01:27:45,050 --> 01:27:49,730
here to engage in discussions
and debates that do not solely

1831
01:27:49,730 --> 01:27:53,560
involve focus on the
present conflicts, trade

1832
01:27:53,560 --> 01:27:55,220
conflicts being just
one case in point,

1833
01:27:55,220 --> 01:27:58,610
but to have a transatlantic
dialogue about the issues

1834
01:27:58,610 --> 01:28:02,230
that are really at the essence
of what links us and affects us

1835
01:28:02,230 --> 01:28:06,710
in the years to come and will be
affecting us in the future too.

1836
01:28:06,710 --> 01:28:08,900
I very much look
forward to my next visit

1837
01:28:08,900 --> 01:28:12,640
to Boston and to Harvard, and
thank you for having come here.

1838
01:28:12,640 --> 01:28:15,390
[APPLAUSE]

1839
01:28:15,390 --> 01:28:25,781