1
00:00:09,800 --> 00:00:12,960
Speaker 1: Well, yesterday we spoke with Peter Scott, who worked at

2
00:00:13,039 --> 00:00:15,720
Speaker 1: NASA's Jet Propulsion Laboratory, and he's at with a book

3
00:00:15,800 --> 00:00:19,239
Speaker 1: entitled Crisis of Control, How Artificial Superintelligence may destroy or

4
00:00:19,280 --> 00:00:22,720
Speaker 1: save the human race. It's a fascinating conversation. The financial

5
00:00:22,720 --> 00:00:26,840
Speaker 1: services industry rapidly ramping up its use of AI. Let's

6
00:00:26,840 --> 00:00:29,240
Speaker 1: talk about some of the concerns, the risks, the necessary

7
00:00:29,320 --> 00:00:34,640
Speaker 1: navigation as the industry does so. Steve Chabinski's partner and

8
00:00:34,720 --> 00:00:37,239
Speaker 1: chair of the Global Data, Privacy and Cybersecurity Practice at

9
00:00:37,280 --> 00:00:40,479
Speaker 1: White and White and Case. He advises global businesses on

10
00:00:40,560 --> 00:00:44,080
Speaker 1: data and network security, compliance and risk management issues, and

11
00:00:44,120 --> 00:00:46,720
Speaker 1: he has also guided past White Houses on America's national

12
00:00:46,760 --> 00:00:51,040
Speaker 1: cyber strategy. He joins us from Washington, d C. Steve,

13
00:00:51,120 --> 00:00:53,479
Speaker 1: nice to have you here with us, So good to

14
00:00:53,479 --> 00:00:55,360
Speaker 1: be with you. You know, we do see the financial

15
00:00:55,360 --> 00:00:58,240
Speaker 1: industry ramping up it's use of AI, and we hear about,

16
00:00:58,400 --> 00:00:59,960
Speaker 1: you know, how great it is and how it's good

17
00:01:00,040 --> 00:01:04,840
Speaker 1: for investors. But with it comes some concerns. No, definitely,

18
00:01:04,880 --> 00:01:06,640
Speaker 1: and also there's a lot of hype. So I think,

19
00:01:06,680 --> 00:01:09,080
Speaker 1: you know, we have to really figure out what is

20
00:01:09,280 --> 00:01:13,360
Speaker 1: artificial intelligence? How is it being used and what are

21
00:01:13,400 --> 00:01:16,959
Speaker 1: the prospects because, um, if this gets out of control,

22
00:01:17,120 --> 00:01:20,760
Speaker 1: there really are some tremendous ramifications not only do the

23
00:01:20,760 --> 00:01:24,560
Speaker 1: financial services sector, but to society at large well. And

24
00:01:24,640 --> 00:01:27,119
Speaker 1: you know, if you look at a kind of straightforward definition,

25
00:01:27,240 --> 00:01:30,280
Speaker 1: it's you know, computer systems able to perform tasks that

26
00:01:30,480 --> 00:01:33,560
Speaker 1: normally humans would be doing or that require human intelligence.

27
00:01:33,800 --> 00:01:35,960
Speaker 1: How do you see it though? As it relies our

28
00:01:36,200 --> 00:01:40,400
Speaker 1: our applies excuse me to the financial industry um specifically,

29
00:01:40,440 --> 00:01:42,800
Speaker 1: and what you what you're kind of overseeing and advising

30
00:01:42,840 --> 00:01:46,679
Speaker 1: your clients about. Yeah, you know that this is similar

31
00:01:46,720 --> 00:01:48,880
Speaker 1: when we harken back to cloud computing when everyone was

32
00:01:48,920 --> 00:01:51,000
Speaker 1: talking about it as a new thing and it turned

33
00:01:51,000 --> 00:01:53,040
Speaker 1: out that we were using the cloud all the time

34
00:01:53,080 --> 00:01:55,400
Speaker 1: for email, we just didn't call it that. And so

35
00:01:55,440 --> 00:01:59,240
Speaker 1: there are very simple AI applications with machine learning and

36
00:01:59,280 --> 00:02:02,440
Speaker 1: big data and a lytics um that we've been using

37
00:02:02,480 --> 00:02:05,080
Speaker 1: a lot in the financial services sector for quite some time.

38
00:02:05,120 --> 00:02:07,360
Speaker 1: And those are the areas that show a lot of

39
00:02:07,400 --> 00:02:11,000
Speaker 1: hope where we have enormous amounts of data that can

40
00:02:11,040 --> 00:02:14,160
Speaker 1: be ingested by computers, and computers do a great job

41
00:02:14,160 --> 00:02:18,440
Speaker 1: of not complaining and getting through data. Really quickly analyzing

42
00:02:18,480 --> 00:02:20,520
Speaker 1: it and in the best case, and what we're seeing

43
00:02:20,520 --> 00:02:24,280
Speaker 1: in the financial services sector doing predictive analysis, meaning they've

44
00:02:24,320 --> 00:02:26,720
Speaker 1: seen this before and they think that they know what's

45
00:02:26,720 --> 00:02:28,840
Speaker 1: going to happen next, and kind of at the end

46
00:02:28,960 --> 00:02:31,519
Speaker 1: pops out the answer that this stock is going to

47
00:02:31,600 --> 00:02:34,920
Speaker 1: go up right based on market factors UM. And the

48
00:02:34,960 --> 00:02:37,720
Speaker 1: way the financial services sector is taking advantage of this

49
00:02:38,080 --> 00:02:41,480
Speaker 1: is because at the same time that this analytic capability is,

50
00:02:41,680 --> 00:02:45,440
Speaker 1: UM is kind of achieving its capability. UM so is

51
00:02:45,440 --> 00:02:48,240
Speaker 1: the ability to gather data and share it quickly so

52
00:02:48,280 --> 00:02:50,720
Speaker 1: you have current data. A problem that was happening in

53
00:02:50,760 --> 00:02:52,720
Speaker 1: the past is that you had all this kind of

54
00:02:52,840 --> 00:02:55,360
Speaker 1: stores of bad data. Then you do analytics on it

55
00:02:55,360 --> 00:02:57,800
Speaker 1: with humans and it makes it even more stale, and

56
00:02:57,800 --> 00:02:59,440
Speaker 1: then you come up with the conclusion and it just

57
00:02:59,520 --> 00:03:02,320
Speaker 1: doesn't mark get ready because the market has moved on. Well,

58
00:03:02,320 --> 00:03:04,079
Speaker 1: so what do you warrant or what do you say

59
00:03:04,120 --> 00:03:07,640
Speaker 1: to your clients, your financial services clients who are working

60
00:03:07,680 --> 00:03:09,240
Speaker 1: in the world of air. Like you said, it's all

61
00:03:09,240 --> 00:03:10,880
Speaker 1: over the place. It's not a new thing. I mean,

62
00:03:10,919 --> 00:03:12,679
Speaker 1: we talk about it all the time. I think we

63
00:03:12,760 --> 00:03:14,680
Speaker 1: take it for granted at this point, but how do

64
00:03:14,720 --> 00:03:18,680
Speaker 1: you advise them? So there are a couple of initial questions.

65
00:03:18,720 --> 00:03:22,239
Speaker 1: One is is the company able to actually use big

66
00:03:22,320 --> 00:03:25,359
Speaker 1: data without artificial intelligence? Meaning what is what is their

67
00:03:25,400 --> 00:03:29,080
Speaker 1: sophistication so that they actually understand what good data is.

68
00:03:29,280 --> 00:03:31,680
Speaker 1: This point that we were bringing up before that if

69
00:03:31,680 --> 00:03:33,760
Speaker 1: they don't understand how to use big data at all,

70
00:03:33,800 --> 00:03:36,520
Speaker 1: then it's too soon to even think about artificial intelligence.

71
00:03:36,760 --> 00:03:38,240
Speaker 1: And then there are a lot of data questions that

72
00:03:38,280 --> 00:03:41,160
Speaker 1: are legal questions. Who owns the data? Are they even

73
00:03:41,200 --> 00:03:44,000
Speaker 1: allowed to take the data that they have and use

74
00:03:44,040 --> 00:03:45,960
Speaker 1: it for the purposes that they want to use it.

75
00:03:45,960 --> 00:03:49,000
Speaker 1: It sounds like an easy question, It's not so easy

76
00:03:49,080 --> 00:03:51,280
Speaker 1: because what we're seeing is in these big data pools,

77
00:03:51,760 --> 00:03:54,320
Speaker 1: no one knows where the data came from. It came

78
00:03:54,400 --> 00:03:58,000
Speaker 1: from come from a thousand different sources, and the providence

79
00:03:58,040 --> 00:04:00,360
Speaker 1: basically the origins of this data mad or is as

80
00:04:00,360 --> 00:04:02,520
Speaker 1: a matter of law, because some of the data was

81
00:04:02,560 --> 00:04:06,120
Speaker 1: collected under regulation and it couldn't be used for certain purposes.

82
00:04:06,320 --> 00:04:08,520
Speaker 1: Some of it can't be shared. There's other data where

83
00:04:08,520 --> 00:04:10,760
Speaker 1: the data subjects have the ability to tell you to

84
00:04:10,880 --> 00:04:13,360
Speaker 1: delete it or correct it. And if you have these

85
00:04:13,520 --> 00:04:15,320
Speaker 1: what I would say this kind of like dirty pool

86
00:04:15,320 --> 00:04:17,200
Speaker 1: of data where you don't know where it came from,

87
00:04:17,480 --> 00:04:20,040
Speaker 1: then you really shouldn't be building your business case. Well

88
00:04:20,080 --> 00:04:21,880
Speaker 1: here's the thing, Steve, I'm kind of waiting for the

89
00:04:21,880 --> 00:04:25,560
Speaker 1: big lawsuit, the first big case dealing data from one

90
00:04:25,600 --> 00:04:28,640
Speaker 1: of these you know, algorithms or AI or I don't

91
00:04:28,640 --> 00:04:30,960
Speaker 1: know where. Artificial intelligence plays a big factor, and it's like,

92
00:04:31,000 --> 00:04:35,760
Speaker 1: who's ultimately responsible? Yeah, so a responsibility. Well, let's start

93
00:04:35,760 --> 00:04:37,160
Speaker 1: with the fact that we're not at the point yet

94
00:04:37,160 --> 00:04:39,480
Speaker 1: where the computer is responsible and we're not sending robots

95
00:04:39,560 --> 00:04:41,880
Speaker 1: to jail yet, So there's going to be someone right

96
00:04:42,240 --> 00:04:45,240
Speaker 1: that is is viewed as the one that had to

97
00:04:45,240 --> 00:04:48,440
Speaker 1: figure out what the algorithm was. For example, in the

98
00:04:48,440 --> 00:04:52,599
Speaker 1: financial service sector, there are clear legal rules about avoiding

99
00:04:52,640 --> 00:04:56,839
Speaker 1: bias and discrimination with respect to profiling. And so that's

100
00:04:56,839 --> 00:04:59,360
Speaker 1: an area where the you know, the data holder, that

101
00:04:59,480 --> 00:05:01,960
Speaker 1: the per and the group that first got the data

102
00:05:02,080 --> 00:05:05,360
Speaker 1: is going to be liable to the data subject. And

103
00:05:05,400 --> 00:05:08,119
Speaker 1: then it could it could trickle down in liability because

104
00:05:08,120 --> 00:05:09,920
Speaker 1: if then that group shared it with someone else who

105
00:05:09,960 --> 00:05:13,040
Speaker 1: misused the data they got it under license, then they're

106
00:05:13,080 --> 00:05:15,400
Speaker 1: going to be liable. Um. And so you know, you

107
00:05:15,440 --> 00:05:18,479
Speaker 1: have this kind of wave all around where everyone's taking

108
00:05:18,560 --> 00:05:21,320
Speaker 1: care of taking advantage of data sets, but there might

109
00:05:21,360 --> 00:05:25,200
Speaker 1: be enormous consequences based on how it's used. And you

110
00:05:25,240 --> 00:05:27,320
Speaker 1: see this in trading as well. Right if you have

111
00:05:27,360 --> 00:05:29,360
Speaker 1: to figure out what the risk level is, and all

112
00:05:29,360 --> 00:05:33,120
Speaker 1: of a sudden you have funds that are literally allowing

113
00:05:33,200 --> 00:05:37,200
Speaker 1: computers to make decisions about the next investment um and

114
00:05:37,240 --> 00:05:39,600
Speaker 1: actually execute on them. Well that might be nice if

115
00:05:39,600 --> 00:05:41,800
Speaker 1: it's your own money, um and you could afford to

116
00:05:41,800 --> 00:05:43,799
Speaker 1: lose it. But I'm not so sure that that's really

117
00:05:44,320 --> 00:05:47,440
Speaker 1: always ready for prime time unless there's a human they're saying,

118
00:05:47,440 --> 00:05:49,920
Speaker 1: wait a second before we hit the buy button, let's

119
00:05:49,920 --> 00:05:51,839
Speaker 1: figure out what's going on. Something to stay tuned for.

120
00:05:51,960 --> 00:05:55,600
Speaker 1: Stephen Chabinski, chair of the Global Data Privacy and Cybersecurity

121
00:05:55,640 --> 00:05:58,279
Speaker 1: Practice at whiteen Case, right here on Bloomberg Radio