WEBVTT

1
00:00:00.080 --> 00:00:02.663
(Upbeat music)

2
00:00:07.380 --> 00:00:09.510
<v ->As one of the biggest players in the field</v>

3
00:00:09.510 --> 00:00:11.340
how are you thinking about responsible AI?

4
00:00:11.340 --> 00:00:12.930
<v ->I think there's a spectrum of alternatives.</v>

5
00:00:12.930 --> 00:00:15.690
One is I'm gonna do an open experiment

6
00:00:15.690 --> 00:00:18.510
on the human population and see what happens.

7
00:00:18.510 --> 00:00:21.630
There's another that is full head in the sand:

8
00:00:21.630 --> 00:00:23.850
We're just gonna avoid this technology.

9
00:00:23.850 --> 00:00:25.830
We think that there is a thoughtful middle ground

10
00:00:25.830 --> 00:00:28.530
that allows you to use the potential of this technology

11
00:00:28.530 --> 00:00:30.840
but do it in a responsible way.

12
00:00:30.840 --> 00:00:32.100
And part of that is we do think

13
00:00:32.100 --> 00:00:35.130
that the government should play a strong role in regulation.

14
00:00:35.130 --> 00:00:37.050
We have a slightly different perspective.

15
00:00:37.050 --> 00:00:38.760
We don't think that there should be, you know

16
00:00:38.760 --> 00:00:42.390
as some have lobbed out, you know, the IAEA of AI.

17
00:00:42.390 --> 00:00:43.770
We think about it slightly differently.

18
00:00:43.770 --> 00:00:46.170
We think all the agencies that are regulating

19
00:00:46.170 --> 00:00:50.310
different parts of the economy need to build their AI skills

20
00:00:50.310 --> 00:00:54.270
in each of those agencies to effectively regulate it

21
00:00:54.270 --> 00:00:57.060
in a way that allows those industry sectors

22
00:00:57.060 --> 00:00:58.410
to function at full efficiency.

23
00:00:58.410 --> 00:01:01.650
So we're a very strong advocate

24
00:01:01.650 --> 00:01:04.020
of government regulation here, but we want to do it

25
00:01:04.020 --> 00:01:06.690
in a thoughtful way that gets most of the benefit

26
00:01:06.690 --> 00:01:09.480
of this technology in a way that's responsible for society.

27
00:01:09.480 --> 00:01:12.240
I think it's always easier to imagine

28
00:01:12.240 --> 00:01:14.160
the efficiency side of these technologies,

29
00:01:14.160 --> 00:01:16.590
all the jobs that might be changed or removed.

30
00:01:16.590 --> 00:01:19.080
It's very hard to foresee the effectiveness side of it.

31
00:01:19.080 --> 00:01:21.630
So just like, you know, if it was the internet,

32
00:01:21.630 --> 00:01:24.720
if it was mobile, like they've created so many more jobs

33
00:01:24.720 --> 00:01:26.220
and had productivity benefits

34
00:01:26.220 --> 00:01:29.220
on society that have been greater than the downside.

35
00:01:29.220 --> 00:01:32.520
So we'll see the innovator community

36
00:01:32.520 --> 00:01:35.340
find the opportunities we can create with this technology.

37
00:01:35.340 --> 00:01:37.020
So we are betting on the field.

38
00:01:37.020 --> 00:01:39.090
We announced our Watson X platform

39
00:01:39.090 --> 00:01:42.390
which is our foundation model platform.

40
00:01:42.390 --> 00:01:45.270
The other thing that we've bet on is that for enterprises

41
00:01:45.270 --> 00:01:48.900
there'll be a set of use cases that you can use an API

42
00:01:48.900 --> 00:01:51.870
and generate the first draft of something.

43
00:01:51.870 --> 00:01:54.060
But if it's running the core business,

44
00:01:54.060 --> 00:01:55.500
if it's a medical diagnosis,

45
00:01:55.500 --> 00:01:58.560
things where there's real business outcomes or lives

46
00:01:58.560 --> 00:02:00.840
on the line, you're gonna want to have a higher level

47
00:02:00.840 --> 00:02:02.820
of trust and governance certainty on that model.

48
00:02:02.820 --> 00:02:05.190
Yeah. So we've put a lot into everything

49
00:02:05.190 --> 00:02:08.640
from a set of models we build with kind of data provenance

50
00:02:08.640 --> 00:02:10.770
so they begin from clean data so you know

51
00:02:10.770 --> 00:02:13.743
that what you'll get out won't be a bad answer,

52
00:02:15.240 --> 00:02:17.760
but then governance to watch the bias and risk

53
00:02:17.760 --> 00:02:19.830
around it when it's in operations for the enterprise.

54
00:02:19.830 --> 00:02:22.983
So a lot of our bet has been around that at the core.

55
00:02:24.050 --> 00:02:26.633
(Upbeat music)