1
00:00:02,040 --> 00:00:05,080
Speaker 1: Hey, Malcolm Glawell Here, I'm back in your feed today

2
00:00:05,120 --> 00:00:07,960
Speaker 1: because we are re releasing an episode of Smart Talks

3
00:00:07,960 --> 00:00:13,039
Speaker 1: with IBM on a very timely topic, AI governance and

4
00:00:13,119 --> 00:00:17,880
Speaker 1: why regulation is critical to building responsible and accountable AI.

5
00:00:18,480 --> 00:00:24,320
Speaker 1: I hope you enjoy it. Hello, Hello, Welcome to Smart

6
00:00:24,360 --> 00:00:30,000
Speaker 1: Talks with IBM, a podcast from Pushkin Industries, iHeartRadio and IBM.

7
00:00:30,320 --> 00:00:34,720
Speaker 1: I'm Malcolm Glabwell. This season, we're continuing our conversation with

8
00:00:34,880 --> 00:00:39,360
Speaker 1: new creators visionaries who are creatively applying technology in business

9
00:00:39,360 --> 00:00:43,080
Speaker 1: to drive change, but with a focus on the transformative

10
00:00:43,120 --> 00:00:47,040
Speaker 1: power of artificial intelligence and what it means to leverage

11
00:00:47,080 --> 00:00:51,959
Speaker 1: AI as a game changing multiplier for your business. Our

12
00:00:51,960 --> 00:00:57,160
Speaker 1: guest today is Christina Montgomery, IBM's Chief Privacy and Trust Officer.

13
00:00:57,760 --> 00:01:01,760
Speaker 1: She's also chair of IBM's AI FAS Export In addition

14
00:01:01,840 --> 00:01:06,200
Speaker 1: to overseeing IBM's privacy policy, a core part of Christina's

15
00:01:06,280 --> 00:01:10,640
Speaker 1: job involves AI governance, making sure the way AI is

16
00:01:10,760 --> 00:01:17,039
Speaker 1: used complies with the international legal regulations customized for each industry.

17
00:01:17,720 --> 00:01:22,920
Speaker 1: In today's episode, Christina will explain why businesses need foundational

18
00:01:22,959 --> 00:01:27,520
Speaker 1: principles when it comes to using technology, why AI regulation

19
00:01:27,840 --> 00:01:32,160
Speaker 1: should focus on specific use cases over the technology itself,

20
00:01:32,680 --> 00:01:36,840
Speaker 1: and share a little bit about her landmark congressional testimony.

21
00:01:37,240 --> 00:01:41,440
Speaker 1: Last May, Christina spoke with doctor Lori Santos, host of

22
00:01:41,480 --> 00:01:46,240
Speaker 1: the Pushkin podcast The Happiness Lab, a cognitive scientist and

23
00:01:46,440 --> 00:01:50,720
Speaker 1: psychology professor at Yale University. Laurie is an expert on

24
00:01:50,960 --> 00:01:56,040
Speaker 1: human happiness and cognition. Okay, let's get to the interview.

25
00:01:58,600 --> 00:02:00,680
Speaker 2: So, Christina, I'm so excited to talk to you today.

26
00:02:00,840 --> 00:02:03,040
Speaker 2: So let's start by talking a little bit about your

27
00:02:03,120 --> 00:02:05,840
Speaker 2: role at IBM. What does a Chief Privacy and Trust

28
00:02:05,840 --> 00:02:06,880
Speaker 2: Officer actually do.

29
00:02:07,680 --> 00:02:11,440
Speaker 3: It's a really dynamic profession and it's not a new profession,

30
00:02:12,000 --> 00:02:14,880
Speaker 3: but the role has really changed. I mean, my role

31
00:02:15,000 --> 00:02:18,680
Speaker 3: today is broader than just helping to ensure compliance with

32
00:02:18,840 --> 00:02:23,480
Speaker 3: data protection laws globally. I'm also responsible for AI governance.

33
00:02:23,520 --> 00:02:26,359
Speaker 3: I co chair or AI Ethics Board here at IBM,

34
00:02:26,800 --> 00:02:29,840
Speaker 3: and for data clearance and data governance as well for

35
00:02:29,919 --> 00:02:33,480
Speaker 3: the company. So I have both a compliance aspect to

36
00:02:33,520 --> 00:02:36,200
Speaker 3: my role, really important on a global basis, but also

37
00:02:36,800 --> 00:02:41,679
Speaker 3: help the business to competitively differentiate because really trust is

38
00:02:41,680 --> 00:02:45,400
Speaker 3: a strategic advantage for IBM and a competitive differentiator as

39
00:02:45,400 --> 00:02:49,360
Speaker 3: a company that's been responsibly managing the most sensitive data

40
00:02:49,360 --> 00:02:51,960
Speaker 3: for our clients for more than a century now and

41
00:02:52,120 --> 00:02:54,840
Speaker 3: helping to usher new technologies into the world with trust

42
00:02:54,840 --> 00:02:58,240
Speaker 3: and transparency. And so that's also a key aspect of

43
00:02:58,280 --> 00:02:58,720
Speaker 3: my role.

44
00:02:59,200 --> 00:03:01,480
Speaker 2: And so joined us here on smart Talks back in

45
00:03:01,520 --> 00:03:04,079
Speaker 2: twenty twenty one, and you chatted with us about IBM's

46
00:03:04,080 --> 00:03:07,560
Speaker 2: approach of building trust and transparency with AI, and that

47
00:03:07,680 --> 00:03:09,600
Speaker 2: was only two years ago. But it almost feels like

48
00:03:09,639 --> 00:03:12,480
Speaker 2: an eternity has happened in the field of AI since then,

49
00:03:13,000 --> 00:03:15,360
Speaker 2: and so I'm curious how much has changed since you

50
00:03:15,400 --> 00:03:17,960
Speaker 2: were here last time. Were the things you told us

51
00:03:18,000 --> 00:03:21,080
Speaker 2: before you are they still true? How are things changing?

52
00:03:21,760 --> 00:03:25,239
Speaker 3: You're absolutely right, it feels like the world has changed

53
00:03:25,320 --> 00:03:29,000
Speaker 3: really in the last two years. But the same fundamental

54
00:03:29,040 --> 00:03:33,960
Speaker 3: principles and the same overall governance applied to IBM's program

55
00:03:34,520 --> 00:03:38,880
Speaker 3: for data protection and responsible AI that we talked about

56
00:03:38,920 --> 00:03:41,600
Speaker 3: two years ago, and not much has changed there from

57
00:03:41,640 --> 00:03:44,800
Speaker 3: our perspective. And the good thing is we've put these

58
00:03:44,840 --> 00:03:49,320
Speaker 3: practices and this governance approach into place, and we've have

59
00:03:49,520 --> 00:03:53,200
Speaker 3: an established way of looking at these emerging technologies. As

60
00:03:53,200 --> 00:03:56,640
Speaker 3: the technology evolves, the tech is more powerful, for sure.

61
00:03:56,760 --> 00:04:00,800
Speaker 3: Foundation models are vastly larger and more capable and are

62
00:04:00,880 --> 00:04:04,320
Speaker 3: creating in some respects new issues. But that just makes

63
00:04:04,320 --> 00:04:06,120
Speaker 3: it all the more urgent to do what we've been

64
00:04:06,160 --> 00:04:09,760
Speaker 3: doing and to put trust and transparency into place across

65
00:04:09,800 --> 00:04:12,440
Speaker 3: the business to be accountable to those principles.

66
00:04:13,400 --> 00:04:15,560
Speaker 2: And so our conversation today is really centered around this

67
00:04:15,680 --> 00:04:18,919
Speaker 2: need for new AI regulation and part of that regulation

68
00:04:19,040 --> 00:04:21,920
Speaker 2: involves the mitigation of bias. And this is something I

69
00:04:21,960 --> 00:04:24,320
Speaker 2: think about a ton as a psychologist, right, you know,

70
00:04:24,360 --> 00:04:27,080
Speaker 2: I know, like my students and everyone who's interacting with

71
00:04:27,160 --> 00:04:30,279
Speaker 2: AI is assuming that the kind of knowledge that they're

72
00:04:30,279 --> 00:04:33,320
Speaker 2: getting from this kind of learning is accurate, right, But

73
00:04:33,360 --> 00:04:35,400
Speaker 2: of course AI is only as good as the knowledge

74
00:04:35,400 --> 00:04:37,720
Speaker 2: that's going in. And so talk to me a little

75
00:04:37,720 --> 00:04:40,960
Speaker 2: bit about like why bias occurs in AI and the

76
00:04:41,080 --> 00:04:42,960
Speaker 2: level of the problem that we're really dealing with.

77
00:04:44,040 --> 00:04:47,320
Speaker 3: Yeah, Well, obviously AI is based on data, right, It's

78
00:04:47,440 --> 00:04:52,440
Speaker 3: trained with data, and that data could be biased in

79
00:04:52,440 --> 00:04:54,920
Speaker 3: and of itself, and that's where issues could come up.

80
00:04:54,960 --> 00:04:56,840
Speaker 3: They come up in the data, they could also come

81
00:04:56,920 --> 00:05:00,479
Speaker 3: up in the output of the models themselves. So it's

82
00:05:00,600 --> 00:05:05,320
Speaker 3: really important that you build bias consideration and bias testing

83
00:05:05,440 --> 00:05:08,840
Speaker 3: into your product development cycle. And so what we've been

84
00:05:08,880 --> 00:05:11,440
Speaker 3: thinking about here at IBM and doing we had some

85
00:05:11,520 --> 00:05:14,479
Speaker 3: of our research teams delivered some of the very first

86
00:05:14,480 --> 00:05:17,839
Speaker 3: toolkits to help detect bias years ago now right and

87
00:05:17,880 --> 00:05:21,520
Speaker 3: deployed them to open source, and we have put into

88
00:05:21,560 --> 00:05:24,880
Speaker 3: place for our developers here at IBM and Ethics by

89
00:05:24,920 --> 00:05:28,120
Speaker 3: Design playbook that's sort of a step by step approach

90
00:05:28,880 --> 00:05:35,360
Speaker 3: which also addresses very fully bias considerations, and we provide

91
00:05:35,480 --> 00:05:38,640
Speaker 3: not only like here's a point when you should test

92
00:05:38,680 --> 00:05:41,040
Speaker 3: for it and you consider it in the data, you

93
00:05:41,120 --> 00:05:42,839
Speaker 3: have to measure it both at the data and the

94
00:05:42,880 --> 00:05:46,400
Speaker 3: model level or the outcome level, and we provide guidance

95
00:05:46,440 --> 00:05:49,080
Speaker 3: with respect to what tools can best be used to

96
00:05:49,120 --> 00:05:52,599
Speaker 3: accomplish that. So it's a really important issue. It's one

97
00:05:52,640 --> 00:05:55,600
Speaker 3: you can't just talk about. You have to provide essentially

98
00:05:55,720 --> 00:05:58,960
Speaker 3: the technology and the capabilities and the guidance to enable

99
00:05:59,000 --> 00:05:59,960
Speaker 3: people to test for.

100
00:06:00,600 --> 00:06:03,280
Speaker 2: Recently, you had this wonderful opportunity to head to Congress

101
00:06:03,320 --> 00:06:06,359
Speaker 2: to talk about AI, and in your testimony before Congress,

102
00:06:06,400 --> 00:06:09,400
Speaker 2: you mentioned that it's often said that innovation moves too

103
00:06:09,440 --> 00:06:12,040
Speaker 2: fast for government to keep up, and this is something

104
00:06:12,040 --> 00:06:14,200
Speaker 2: that I also worry about as a psychologist, right our

105
00:06:14,240 --> 00:06:17,160
Speaker 2: policy makers really understanding the issues that they're dealing with,

106
00:06:17,600 --> 00:06:19,880
Speaker 2: And so I'm curious how you're approaching this challenge of

107
00:06:19,920 --> 00:06:22,359
Speaker 2: adapting AI policies to keep up with the sort of

108
00:06:22,440 --> 00:06:24,880
Speaker 2: rapid pace of all the advancements we're seeing in the

109
00:06:24,920 --> 00:06:26,320
Speaker 2: AI technology itself.

110
00:06:27,560 --> 00:06:30,760
Speaker 3: I think it's really critically important that you have foundational

111
00:06:30,800 --> 00:06:36,240
Speaker 3: principles that applied to not only how you use technology,

112
00:06:36,279 --> 00:06:37,800
Speaker 3: but whether you're going to use it in the first

113
00:06:37,800 --> 00:06:39,440
Speaker 3: place and where you're going to use and apply it

114
00:06:39,440 --> 00:06:43,560
Speaker 3: across your company. And then your program from a governance perspective,

115
00:06:43,600 --> 00:06:46,120
Speaker 3: has to be agile. It has to be able to

116
00:06:46,200 --> 00:06:51,920
Speaker 3: address emerging capabilities, new training methods, etc. And part of

117
00:06:51,960 --> 00:06:56,479
Speaker 3: that involves helping to educate and instill and empower a

118
00:06:56,520 --> 00:07:00,320
Speaker 3: trustworthy culture at a company so you can spot those

119
00:07:00,360 --> 00:07:02,640
Speaker 3: issues so you can ask the right questions at the

120
00:07:02,680 --> 00:07:05,880
Speaker 3: right time if you try. We talked about during the

121
00:07:05,920 --> 00:07:10,160
Speaker 3: Senate hearing, and IBM's been talking for years about regulating

122
00:07:10,480 --> 00:07:13,600
Speaker 3: the use, not the technology itself, because if you try

123
00:07:13,600 --> 00:07:17,440
Speaker 3: to regulate technology, you're very quickly going to find out

124
00:07:17,840 --> 00:07:20,760
Speaker 3: regulation will absolutely never keep up with that.

125
00:07:21,000 --> 00:07:23,240
Speaker 2: And so in your testimony to Congress, you also talked

126
00:07:23,240 --> 00:07:26,480
Speaker 2: about this idea of a precision regulation approach for AI.

127
00:07:27,080 --> 00:07:29,280
Speaker 2: Tell me more about this. What is a precision regulation

128
00:07:29,360 --> 00:07:31,720
Speaker 2: approach and why could that be so important.

129
00:07:31,920 --> 00:07:34,880
Speaker 3: It's funny because I was able to share with Congress

130
00:07:35,520 --> 00:07:39,320
Speaker 3: our precision regulation point of view in twenty twenty three,

131
00:07:39,680 --> 00:07:42,480
Speaker 3: but that precision regulation point of view was published by

132
00:07:42,600 --> 00:07:47,280
Speaker 3: IBM in twenty twenty. So we have not changed our

133
00:07:47,400 --> 00:07:52,320
Speaker 3: position that you should apply the tightest controls, the strictest

134
00:07:52,360 --> 00:07:56,600
Speaker 3: regulatory requirements to the technology where the end use and

135
00:07:56,720 --> 00:08:00,720
Speaker 3: risk of societal harm is the greatest. So that's essentially

136
00:08:00,720 --> 00:08:03,680
Speaker 3: what it is. There's lots of AI technology that's used

137
00:08:03,680 --> 00:08:07,320
Speaker 3: today that doesn't touch people, that's very low risk in nature.

138
00:08:07,840 --> 00:08:11,280
Speaker 3: And even when you think about AI that delivers a

139
00:08:11,400 --> 00:08:16,840
Speaker 3: movie recommendation versus AI that is used to diagnose cancer, right,

140
00:08:16,880 --> 00:08:20,360
Speaker 3: there's very different implications associated with those two uses of

141
00:08:20,400 --> 00:08:25,240
Speaker 3: the technology. And so essentially what precision regulation is is

142
00:08:25,240 --> 00:08:28,640
Speaker 3: apply different rules to different risks, right, more stringent regulation

143
00:08:28,920 --> 00:08:32,040
Speaker 3: to the use cases with the greatest risk. And then

144
00:08:32,200 --> 00:08:36,360
Speaker 3: also we build that out calling for things like transparency

145
00:08:36,960 --> 00:08:40,240
Speaker 3: you see it today with content right, misinformation and the

146
00:08:40,440 --> 00:08:44,480
Speaker 3: like we believe that consumers should always know when they're

147
00:08:44,520 --> 00:08:47,679
Speaker 3: interacting with an AI system, So be transparent, don't hydro

148
00:08:47,679 --> 00:08:52,120
Speaker 3: your AI. Clearly define the risks. So as a country,

149
00:08:52,440 --> 00:08:55,319
Speaker 3: we need to have some clear guidance right in globally

150
00:08:55,360 --> 00:08:59,079
Speaker 3: as well in terms of which uses of AI or

151
00:08:59,160 --> 00:09:03,880
Speaker 3: higher risk role apply higher and stricter regulation and have

152
00:09:04,000 --> 00:09:06,600
Speaker 3: sort of a common understanding of what those high risk

153
00:09:06,840 --> 00:09:10,600
Speaker 3: uses are and then demonstrate the impact in the cases

154
00:09:10,640 --> 00:09:15,160
Speaker 3: of those higher risk uses. So companies who are using

155
00:09:15,200 --> 00:09:19,280
Speaker 3: AI in spaces where they can impact people's legal rights,

156
00:09:19,320 --> 00:09:23,679
Speaker 3: for example, should have to conduct an impact assessment that

157
00:09:23,800 --> 00:09:27,560
Speaker 3: demonstrates that the technology isn't biased. So we've been pretty

158
00:09:27,559 --> 00:09:31,680
Speaker 3: clear about apply the most stringent regulation to the highest

159
00:09:31,760 --> 00:09:32,720
Speaker 3: risk uses of AI.

160
00:09:34,160 --> 00:09:36,439
Speaker 2: And so so far we've been talking about your congressional

161
00:09:36,480 --> 00:09:39,320
Speaker 2: testimony in terms of, you know, the specific content that

162
00:09:39,360 --> 00:09:41,840
Speaker 2: you talked about, But I'm just curious on a personal level,

163
00:09:42,160 --> 00:09:44,200
Speaker 2: you know, what was that like right like right now

164
00:09:44,280 --> 00:09:46,520
Speaker 2: it feels like at a policy level, like there's a

165
00:09:46,640 --> 00:09:48,960
Speaker 2: kind of fever pitch going on with AI right now.

166
00:09:49,200 --> 00:09:50,640
Speaker 2: You know what did that feel like to kind of

167
00:09:50,760 --> 00:09:53,160
Speaker 2: really have the opportunity to talk to policy makers and

168
00:09:53,200 --> 00:09:56,200
Speaker 2: sort of influence what they're thinking about AI technologies like

169
00:09:56,280 --> 00:09:57,240
Speaker 2: in the coming century.

170
00:09:57,280 --> 00:10:00,719
Speaker 3: Perhaps I was really an honor to able to do

171
00:10:00,760 --> 00:10:03,640
Speaker 3: that and to be one of the first set of

172
00:10:03,679 --> 00:10:07,839
Speaker 3: invitees to the first hearing. And what I learned from

173
00:10:07,880 --> 00:10:11,200
Speaker 3: it essentially is, you know, really two things. The first

174
00:10:11,200 --> 00:10:15,120
Speaker 3: is really the value of authenticity. So both as an

175
00:10:15,120 --> 00:10:19,160
Speaker 3: individual and as a company, I was able to talk

176
00:10:19,200 --> 00:10:21,679
Speaker 3: about what I do. You know, I need a lot

177
00:10:21,679 --> 00:10:25,520
Speaker 3: of advanced prep right. I talked about what my job is,

178
00:10:26,480 --> 00:10:29,200
Speaker 3: what IBM has been putting in place for years now.

179
00:10:29,800 --> 00:10:33,640
Speaker 3: So this isn't about creating something. This was just about

180
00:10:33,679 --> 00:10:36,000
Speaker 3: showing up and being authentic. And we were invited for

181
00:10:36,040 --> 00:10:38,880
Speaker 3: a reason. We were invited because we were one of

182
00:10:38,880 --> 00:10:43,679
Speaker 3: the earliest companies in the AI technology space. We're the

183
00:10:43,679 --> 00:10:49,120
Speaker 3: oldest technology company and we are trusted and that's an honor.

184
00:10:49,280 --> 00:10:51,520
Speaker 3: And then the second thing I came away with was

185
00:10:51,559 --> 00:10:53,920
Speaker 3: really how important this issue is to society. I don't

186
00:10:53,920 --> 00:10:59,439
Speaker 3: think I appreciated it as much until following that experience.

187
00:11:00,120 --> 00:11:03,840
Speaker 3: I had outreached from colleagues I hadn't worked with for years.

188
00:11:03,920 --> 00:11:06,000
Speaker 3: I had an outreach from family members who heard me

189
00:11:06,040 --> 00:11:09,120
Speaker 3: on the radio, you know, my mother and my mother

190
00:11:09,200 --> 00:11:12,319
Speaker 3: in law, and my nieces and nephews and my friends

191
00:11:12,320 --> 00:11:14,160
Speaker 3: of my kids were all like, Oh, I get it.

192
00:11:14,240 --> 00:11:16,520
Speaker 3: I get what you do. Now, Wow, that's pretty cool,

193
00:11:16,760 --> 00:11:19,560
Speaker 3: you know. So that was really probably the best and

194
00:11:19,640 --> 00:11:21,720
Speaker 3: most impactful takeaway that I had.

195
00:11:22,000 --> 00:11:25,600
Speaker 1: The mass adoption of generative AI, happening at breakneck speed,

196
00:11:26,000 --> 00:11:30,000
Speaker 1: has spurred societies and governments around the world to get

197
00:11:30,080 --> 00:11:36,400
Speaker 1: serious about regulating AI. For businesses, compliance is complex enough already,

198
00:11:36,559 --> 00:11:39,880
Speaker 1: but throw anever involving technology like AI into the mix,

199
00:11:40,280 --> 00:11:46,520
Speaker 1: and compliance itself becomes an exercise in adaptability. As regulators

200
00:11:46,559 --> 00:11:51,120
Speaker 1: seek greater accountability in how AI is used, businesses need

201
00:11:51,200 --> 00:11:56,120
Speaker 1: help creating governance processes comprehensive enough to comply with the law,

202
00:11:56,360 --> 00:11:59,440
Speaker 1: but agile enough to keep up with the rapid rate

203
00:11:59,480 --> 00:12:04,840
Speaker 1: of change in AI development. Regulatory scrutiny isn't the only

204
00:12:04,960 --> 00:12:10,079
Speaker 1: consideration either responsible AI governance. A business's ability to prove

205
00:12:10,160 --> 00:12:14,960
Speaker 1: its AI models are transparent and explainable is also key

206
00:12:15,040 --> 00:12:20,040
Speaker 1: to building trust with customers, regardless of industry. In the

207
00:12:20,080 --> 00:12:24,319
Speaker 1: next part of their conversation, Laurie asked Christina what businesses

208
00:12:24,320 --> 00:12:28,720
Speaker 1: should consider when approaching AI governance. Let's listen.

209
00:12:29,600 --> 00:12:32,000
Speaker 2: So it's a particular role that businesses are playing in

210
00:12:32,080 --> 00:12:34,880
Speaker 2: AI governance, Like why is it so critical for businesses

211
00:12:34,920 --> 00:12:35,840
Speaker 2: to be part of this?

212
00:12:36,720 --> 00:12:41,240
Speaker 3: So I think it's really critically important that businesses understand

213
00:12:41,720 --> 00:12:44,640
Speaker 3: the impacts that technology can have, both in making them

214
00:12:44,640 --> 00:12:48,480
Speaker 3: better businesses, but the impacts that those technologies can have

215
00:12:48,960 --> 00:12:54,079
Speaker 3: on the consumers that they are supporting. You know, businesses

216
00:12:54,200 --> 00:12:58,800
Speaker 3: need to be deploying AI technology that is in alignment

217
00:12:58,960 --> 00:13:00,840
Speaker 3: with the goals that they set for it, and that

218
00:13:00,960 --> 00:13:03,840
Speaker 3: can be trusted. I think for us and for our clients,

219
00:13:04,160 --> 00:13:07,080
Speaker 3: a lot of this comes back to trust in tech.

220
00:13:07,520 --> 00:13:13,200
Speaker 3: If you deploy something that doesn't work, that hallucinates, that discriminates,

221
00:13:13,679 --> 00:13:17,839
Speaker 3: that isn't transparent, where decisions can't be explained, then you

222
00:13:17,960 --> 00:13:21,600
Speaker 3: are going to very rapidly erode the trust at best

223
00:13:21,760 --> 00:13:25,079
Speaker 3: right of your clients and at worst for yourself. You're

224
00:13:25,120 --> 00:13:27,800
Speaker 3: going to create legal and regulatory issues for yourself as well.

225
00:13:27,840 --> 00:13:31,760
Speaker 3: So trusted technology is really important, and I think there's

226
00:13:31,800 --> 00:13:34,000
Speaker 3: a lot of pressure on businesses today to move very

227
00:13:34,080 --> 00:13:36,720
Speaker 3: rapidly and adopt technology. But if you do it without

228
00:13:36,840 --> 00:13:40,319
Speaker 3: having a program of governance in place, you're really risking

229
00:13:40,520 --> 00:13:42,000
Speaker 3: eroding that trust.

230
00:13:41,960 --> 00:13:44,000
Speaker 2: And so this is really where I think a strong

231
00:13:44,040 --> 00:13:47,600
Speaker 2: AI governance comes in. Talk about from your perspective, how

232
00:13:47,840 --> 00:13:51,200
Speaker 2: this really contributes to maintaining the trust that customers and

233
00:13:51,240 --> 00:13:53,120
Speaker 2: stakeholders have in these technologies.

234
00:13:53,320 --> 00:13:55,880
Speaker 3: Yeah. Absolutely, I mean you need to have a governance

235
00:13:55,920 --> 00:13:59,880
Speaker 3: program because you need to understand that the technology, particularly

236
00:13:59,880 --> 00:14:04,520
Speaker 3: in the AI space, that you are deploying, is explainable.

237
00:14:04,600 --> 00:14:08,920
Speaker 3: You need to understand why it's making decisions and recommendations

238
00:14:09,000 --> 00:14:10,440
Speaker 3: that it's making, and you need to be able to

239
00:14:10,440 --> 00:14:12,840
Speaker 3: explain that to your consumers. I mean, you can't do

240
00:14:12,920 --> 00:14:15,000
Speaker 3: that if you don't know where your data is coming from,

241
00:14:15,000 --> 00:14:17,480
Speaker 3: what data are you using to train those models, if

242
00:14:17,480 --> 00:14:21,280
Speaker 3: you don't have a program that manages the alignment of

243
00:14:21,320 --> 00:14:24,880
Speaker 3: your AI models over time to make sure as AI

244
00:14:25,120 --> 00:14:29,520
Speaker 3: learns and evolves over uses, which is in large part

245
00:14:30,120 --> 00:14:33,920
Speaker 3: what makes it so beneficial that it stays in alignment

246
00:14:33,960 --> 00:14:37,440
Speaker 3: with the objectives that you set for the technology over time.

247
00:14:38,080 --> 00:14:41,920
Speaker 3: So you can't do that without a robust governance process

248
00:14:41,960 --> 00:14:45,400
Speaker 3: in place. So we work with clients to share our

249
00:14:45,440 --> 00:14:47,920
Speaker 3: own story here at IBM in terms of how we

250
00:14:47,960 --> 00:14:51,560
Speaker 3: put that in place, but also in our consulting practice

251
00:14:52,320 --> 00:14:56,840
Speaker 3: to help clients work with these new generative capabilities and

252
00:14:56,840 --> 00:14:59,720
Speaker 3: foundation models and the like in order to put them

253
00:14:59,760 --> 00:15:01,840
Speaker 3: to work for their business in a way that's going

254
00:15:01,880 --> 00:15:04,560
Speaker 3: to be impactful to that business, but at the same

255
00:15:04,600 --> 00:15:05,640
Speaker 3: time be trusted.

256
00:15:05,840 --> 00:15:07,640
Speaker 2: So now I wanted to turn a little bit towards

257
00:15:07,680 --> 00:15:11,520
Speaker 2: Watson X governance, and so IBM recently announced their AI platform,

258
00:15:11,600 --> 00:15:15,320
Speaker 2: Watson X, which will include a governance component. Could you

259
00:15:15,320 --> 00:15:17,960
Speaker 2: tell us a little more about watsonx dot governance.

260
00:15:18,560 --> 00:15:20,800
Speaker 3: Yeah, I mean before I do that, I'll just back

261
00:15:20,880 --> 00:15:24,480
Speaker 3: up and talk about the full platform and then lean

262
00:15:24,520 --> 00:15:27,400
Speaker 3: into Watson X because I think it's important to understand

263
00:15:27,680 --> 00:15:33,520
Speaker 3: the delivery of a full suite of capabilities, to get data,

264
00:15:33,720 --> 00:15:36,600
Speaker 3: to train models, and then to govern them over their

265
00:15:36,640 --> 00:15:42,120
Speaker 3: life cycle. All of these things are really important. From

266
00:15:42,200 --> 00:15:45,040
Speaker 3: the onset you need to make sure that you have.

267
00:15:46,000 --> 00:15:50,520
Speaker 3: For our watsonex dot AI for example, that's the studio

268
00:15:50,640 --> 00:15:55,080
Speaker 3: to train new foundation models and generative AI and machine

269
00:15:55,120 --> 00:16:00,880
Speaker 3: learning capabilities, and we are populating that studio with some

270
00:16:01,400 --> 00:16:06,920
Speaker 3: IBM trained foundation models, which we're curating and tailoring more

271
00:16:06,920 --> 00:16:10,200
Speaker 3: specifically for enterprises. So that's really important. It comes back

272
00:16:10,200 --> 00:16:13,360
Speaker 3: to the point I made earlier about business trust and

273
00:16:13,440 --> 00:16:19,840
Speaker 3: the need to have enterprise ready technologies in the AI space,

274
00:16:20,160 --> 00:16:23,680
Speaker 3: and then the watsonex dot data is a fit for

275
00:16:23,800 --> 00:16:27,320
Speaker 3: purpose data store or a data lake, and then watsonex

276
00:16:27,320 --> 00:16:31,960
Speaker 3: dot gov. So that's a particular component of the platform

277
00:16:32,440 --> 00:16:36,160
Speaker 3: that my team and the AI Ethics Board has really

278
00:16:36,240 --> 00:16:39,440
Speaker 3: worked closely with the product team on developing, and we're

279
00:16:39,600 --> 00:16:42,440
Speaker 3: using it internally here in the Chief Privacy Office as

280
00:16:42,480 --> 00:16:46,840
Speaker 3: well to help us govern our own uses of AI

281
00:16:47,000 --> 00:16:52,920
Speaker 3: technology and our compliance program here. And it essentially helps

282
00:16:53,000 --> 00:16:57,640
Speaker 3: to notify you if a model becomes biased or gets

283
00:16:57,640 --> 00:17:00,760
Speaker 3: out of alignment as you're using it over time. So

284
00:17:00,880 --> 00:17:03,480
Speaker 3: companies are going to need these capabilities. I mean they

285
00:17:03,560 --> 00:17:07,760
Speaker 3: need them today to deliver technologies with trust. They'll need

286
00:17:07,800 --> 00:17:11,960
Speaker 3: them tomorrow to comply with regulation which is on the horizon.

287
00:17:11,560 --> 00:17:14,359
Speaker 2: And I think compliance becomes even more complex when you

288
00:17:14,400 --> 00:17:18,560
Speaker 2: consider international data protection laws and regulations. Honestly, I don't

289
00:17:18,560 --> 00:17:21,120
Speaker 2: know how anyone on any company's legal team is keeping

290
00:17:21,200 --> 00:17:23,439
Speaker 2: up with us these days. But my question for you

291
00:17:23,560 --> 00:17:27,160
Speaker 2: is really how can businesses develop a strategy to maintain

292
00:17:27,240 --> 00:17:30,240
Speaker 2: compliance and to deal with it in this ever changing landscape.

293
00:17:30,320 --> 00:17:34,280
Speaker 3: It's increasingly more challenging. In fact, I saw statistic just

294
00:17:34,359 --> 00:17:38,960
Speaker 3: this morning that the regulatory obligations on companies have increased

295
00:17:38,960 --> 00:17:42,840
Speaker 3: something like seven hundred times in the last twenty years.

296
00:17:42,640 --> 00:17:47,240
Speaker 3: So it really is a huge focus area for companies.

297
00:17:47,400 --> 00:17:50,280
Speaker 3: You have to have a process in place in order

298
00:17:50,320 --> 00:17:52,840
Speaker 3: to do that, and it's not easy, particularly for a

299
00:17:52,920 --> 00:17:56,800
Speaker 3: company like IBM that it has a presence in over

300
00:17:56,800 --> 00:18:00,000
Speaker 3: one hundred and seventy countries around the world. There's more

301
00:18:00,000 --> 00:18:04,760
Speaker 3: more than one hundred and fifty comprehensive privacy regulations, there

302
00:18:04,800 --> 00:18:09,320
Speaker 3: are regulations of non personal data, there are AI regulations emerging,

303
00:18:10,320 --> 00:18:14,359
Speaker 3: so you really need an operational approach to it in

304
00:18:14,480 --> 00:18:16,520
Speaker 3: order to stay compliant. But one of the things we

305
00:18:16,560 --> 00:18:18,720
Speaker 3: do is we set a baseline, and a lot of

306
00:18:18,720 --> 00:18:22,159
Speaker 3: companies do this as well. So we define a privacy baseline,

307
00:18:22,200 --> 00:18:27,040
Speaker 3: we define an AI baseline, and we ensure then as

308
00:18:27,040 --> 00:18:29,440
Speaker 3: a result of that that there are very few deviances

309
00:18:29,520 --> 00:18:32,600
Speaker 3: because it incorporates in that baseline. So that's one of

310
00:18:32,640 --> 00:18:34,840
Speaker 3: the ways we do it. Other companies, I think are

311
00:18:34,880 --> 00:18:40,080
Speaker 3: similarly situated in terms of doing that. But again, it

312
00:18:40,240 --> 00:18:42,919
Speaker 3: is a real challenge for global companies. It's one of

313
00:18:42,920 --> 00:18:46,880
Speaker 3: the reasons why we advocate for as much alignment as

314
00:18:46,960 --> 00:18:52,320
Speaker 3: possible on the international realm as well as nationally here

315
00:18:52,359 --> 00:18:56,160
Speaker 3: in the US, as much alignment as possible to make

316
00:18:56,359 --> 00:19:01,320
Speaker 3: compliance easier for easier and not just because companies want

317
00:19:01,359 --> 00:19:04,480
Speaker 3: an easy way to comply. But the harder it is,

318
00:19:04,760 --> 00:19:08,639
Speaker 3: the less likely there will be compliance. And it's not

319
00:19:08,760 --> 00:19:14,840
Speaker 3: the objective of anybody, governments, companies, consumers to have to

320
00:19:15,000 --> 00:19:18,520
Speaker 3: set legal obligations that companies simply can't meet.

321
00:19:18,960 --> 00:19:21,040
Speaker 2: So what advice would you give to other companies who

322
00:19:21,040 --> 00:19:24,000
Speaker 2: are looking to rethink or strengthen their approach to AI government.

323
00:19:24,160 --> 00:19:27,280
Speaker 3: You need to start with, as we did, foundational principles,

324
00:19:27,920 --> 00:19:31,359
Speaker 3: and you need to start making decisions about what technology

325
00:19:31,359 --> 00:19:33,720
Speaker 3: you're going to deploy and what technology you're not, what

326
00:19:33,720 --> 00:19:34,840
Speaker 3: are you going to use it for, and what aren't

327
00:19:34,840 --> 00:19:36,240
Speaker 3: you going to use it for? And then when you

328
00:19:36,280 --> 00:19:40,640
Speaker 3: do use it, align to those principles. That's really important.

329
00:19:40,720 --> 00:19:45,240
Speaker 3: Formalize a program, have someone within the organization, whether it's

330
00:19:45,280 --> 00:19:49,560
Speaker 3: the chief Privacy officer, whether it's some other role, a

331
00:19:49,640 --> 00:19:54,080
Speaker 3: chief AI ethics officer, but have an accountable individual and

332
00:19:54,160 --> 00:19:58,399
Speaker 3: accountable organization. Do a maturity assessment, figure out where you

333
00:19:58,400 --> 00:20:01,159
Speaker 3: are and where you need to be, and really start

334
00:20:01,359 --> 00:20:05,520
Speaker 3: you know, putting it into place today. Don't wait for

335
00:20:05,920 --> 00:20:08,960
Speaker 3: regulation to apply directly to your business, because it'll be

336
00:20:09,040 --> 00:20:09,440
Speaker 3: too late.

337
00:20:10,400 --> 00:20:13,800
Speaker 2: So Smart Talks features new creators, these visionaries like yourself

338
00:20:13,800 --> 00:20:17,360
Speaker 2: who are creatively applying technology in business to drive change.

339
00:20:17,600 --> 00:20:20,280
Speaker 2: I'm curious if you see yourself as creative.

340
00:20:20,720 --> 00:20:24,040
Speaker 3: You know, I definitely do. I mean you need to

341
00:20:24,119 --> 00:20:28,760
Speaker 3: be creative when you're working in an industry that evolves

342
00:20:28,800 --> 00:20:33,320
Speaker 3: so very quickly. So you know, I started with IBM

343
00:20:33,560 --> 00:20:36,680
Speaker 3: when we were primarily a hardware company, right and we've

344
00:20:36,800 --> 00:20:40,199
Speaker 3: changed our business so significantly over the years, and the

345
00:20:40,280 --> 00:20:44,720
Speaker 3: issues that are raised with respect to each new technology,

346
00:20:44,760 --> 00:20:48,400
Speaker 3: whether it be cloud, whether it be AI now where

347
00:20:48,400 --> 00:20:50,000
Speaker 3: we're seeing a ton of issues, or you look at

348
00:20:50,000 --> 00:20:54,159
Speaker 3: emergent issues in the space of things like neurotechnologies and

349
00:20:54,240 --> 00:21:00,840
Speaker 3: quantum computers. You have to be strategic and you have

350
00:21:00,960 --> 00:21:04,200
Speaker 3: to be creative and thinking about how you can adapt

351
00:21:04,840 --> 00:21:10,080
Speaker 3: agilely quickly a company to an environment that is changing

352
00:21:10,119 --> 00:21:11,880
Speaker 3: so quickly and.

353
00:21:11,880 --> 00:21:14,919
Speaker 2: With this transformation happening at such a rapid pace. Do

354
00:21:14,960 --> 00:21:17,040
Speaker 2: you think creativity plays a role in how you think

355
00:21:17,040 --> 00:21:20,359
Speaker 2: about and implement, specifically a trustworthy AI strategy.

356
00:21:22,840 --> 00:21:26,880
Speaker 3: Yeah, I absolutely think it does because again, it comes

357
00:21:26,920 --> 00:21:30,080
Speaker 3: back to these capabilities, and there are ways. I guess

358
00:21:30,280 --> 00:21:34,040
Speaker 3: how you define creativity could be different, right, but I'm

359
00:21:34,080 --> 00:21:37,280
Speaker 3: thinking of creativity in the sense of sort of agility

360
00:21:37,320 --> 00:21:41,400
Speaker 3: and strategic vision and creative problem solving. I think that's

361
00:21:41,680 --> 00:21:44,800
Speaker 3: really important in the world that we're in right now,

362
00:21:44,840 --> 00:21:49,399
Speaker 3: being able to creatively problem solve with new issues that

363
00:21:49,480 --> 00:21:52,560
Speaker 3: are rising sort of every day.

364
00:21:52,960 --> 00:21:54,520
Speaker 2: And so, how do you see the role of chief

365
00:21:54,520 --> 00:21:58,160
Speaker 2: privacy officer evolving in the future as AI technology continues

366
00:21:58,200 --> 00:22:01,080
Speaker 2: to advance, Like what stuff should CPOs take to stay

367
00:22:01,119 --> 00:22:03,040
Speaker 2: ahead of all these changes that are come in their way?

368
00:22:04,560 --> 00:22:08,480
Speaker 3: So the role is evolving in most companies, I would

369
00:22:08,520 --> 00:22:13,480
Speaker 3: say pretty rapidly. Many companies are looking to chief privacy

370
00:22:13,520 --> 00:22:17,040
Speaker 3: officers who are ready understand the data that's being used

371
00:22:17,040 --> 00:22:20,560
Speaker 3: in the organization and have programs to ensure compliance with

372
00:22:20,680 --> 00:22:24,440
Speaker 3: laws that require you to manage that data in accordance

373
00:22:24,480 --> 00:22:27,120
Speaker 3: with data protection laws and the like. It's a natural

374
00:22:27,160 --> 00:22:33,160
Speaker 3: place and position for AI responsibility. And so I think

375
00:22:33,160 --> 00:22:35,840
Speaker 3: what's happening to a lot of chief privacy officers is

376
00:22:35,880 --> 00:22:39,400
Speaker 3: they're being asked to take on this AI governance responsibility

377
00:22:39,400 --> 00:22:42,920
Speaker 3: for companies and if not take it on at least

378
00:22:42,960 --> 00:22:46,399
Speaker 3: play a very key role working with other parts of

379
00:22:46,400 --> 00:22:50,040
Speaker 3: the business in AI governance. So that really is changing.

380
00:22:50,280 --> 00:22:54,280
Speaker 3: And if chief privacy officers are in companies who maybe

381
00:22:54,359 --> 00:22:58,119
Speaker 3: haven't started thinking about AI yet, they should, So I

382
00:22:58,160 --> 00:23:02,000
Speaker 3: would encourage them to look at different resources that are

383
00:23:02,040 --> 00:23:06,399
Speaker 3: available already in AI governance space. For example, the International

384
00:23:06,440 --> 00:23:10,880
Speaker 3: Association of Privacy Professionals, which is the seventy five thousand

385
00:23:10,920 --> 00:23:16,040
Speaker 3: member professional body for the profession of Chief Privacy officers,

386
00:23:16,119 --> 00:23:20,760
Speaker 3: just recently launched an AI Governance Initiative and an AI

387
00:23:20,800 --> 00:23:24,800
Speaker 3: Governance certification program. I sit on their advisory board. But

388
00:23:24,880 --> 00:23:27,560
Speaker 3: that's just emblematic of the fact that the field is

389
00:23:27,640 --> 00:23:29,080
Speaker 3: changing so rapidly.

390
00:23:30,320 --> 00:23:32,880
Speaker 2: And so, you know, speaking of rapid change, when you're

391
00:23:33,200 --> 00:23:35,560
Speaker 2: back here on smart Talks in twenty twenty one, you

392
00:23:35,600 --> 00:23:37,960
Speaker 2: said that the future of AI will be more transparent

393
00:23:37,960 --> 00:23:40,199
Speaker 2: and more trustworthy. You know, what do you see the

394
00:23:40,240 --> 00:23:42,280
Speaker 2: next five to ten years holding? You know, when you're

395
00:23:42,280 --> 00:23:45,200
Speaker 2: back on smart Talks in you know, twenty twenty six,

396
00:23:45,320 --> 00:23:47,000
Speaker 2: you know twenty thirty, you know what are we going

397
00:23:47,040 --> 00:23:49,320
Speaker 2: to be talking about when it comes to AI technology

398
00:23:49,359 --> 00:23:49,960
Speaker 2: and governance.

399
00:23:50,800 --> 00:23:52,959
Speaker 3: So I try to be an optimist, right and I

400
00:23:53,000 --> 00:23:56,840
Speaker 3: said that two years ago, and I think we're seeing

401
00:23:56,880 --> 00:24:01,280
Speaker 3: it now come into fruition, and there will be requirements,

402
00:24:02,119 --> 00:24:04,919
Speaker 3: whether they're coming from the US, whether they're coming from Europe,

403
00:24:04,960 --> 00:24:08,560
Speaker 3: whether they're just coming from voluntary adoption by clients of

404
00:24:08,640 --> 00:24:13,359
Speaker 3: things like the NISS Risk Management Framework, really important voluntary frameworks.

405
00:24:14,119 --> 00:24:17,720
Speaker 3: You're going to have to adopt transparent and explainable practices

406
00:24:17,880 --> 00:24:20,560
Speaker 3: in your uses of AI. So I do see that happening.

407
00:24:20,560 --> 00:24:23,000
Speaker 3: And in the next five to ten years, boy, I

408
00:24:23,040 --> 00:24:28,200
Speaker 3: think we'll see more research into trust in and techniques

409
00:24:28,640 --> 00:24:32,960
Speaker 3: because we don't really know, for example, how to water mark.

410
00:24:33,720 --> 00:24:36,720
Speaker 3: We were calling for things like watermarking. There'll be more

411
00:24:36,800 --> 00:24:42,240
Speaker 3: research into how to do that. I think you'll see

412
00:24:42,359 --> 00:24:45,520
Speaker 3: you regulation that's specifically going to require those types of things.

413
00:24:45,600 --> 00:24:47,760
Speaker 3: So I think again, I think the regulation is going

414
00:24:47,800 --> 00:24:50,600
Speaker 3: to drive research. It's going to drive research into these

415
00:24:50,680 --> 00:24:55,840
Speaker 3: areas that will help ensure that we can deliver new capabilities,

416
00:24:55,920 --> 00:24:59,240
Speaker 3: generated capabilities and the like with trust and explainability.

417
00:24:59,440 --> 00:25:01,159
Speaker 2: Thank you so much Wi Christina for joining me on

418
00:25:01,200 --> 00:25:03,200
Speaker 2: smart Talks to talk about AI and governance.

419
00:25:04,160 --> 00:25:06,120
Speaker 3: Well, thank you very much for having me.

420
00:25:07,400 --> 00:25:12,119
Speaker 1: To unlock the transformative growth possible with artificial intelligence, businesses

421
00:25:12,200 --> 00:25:15,320
Speaker 1: need to know what they wish to grow into first.

422
00:25:16,280 --> 00:25:19,159
Speaker 1: Like Christina said, the best way forward in the AI

423
00:25:19,240 --> 00:25:23,280
Speaker 1: future is for businesses to figure out their own foundational

424
00:25:23,320 --> 00:25:27,960
Speaker 1: principles around using the technology, drawing on those principles to

425
00:25:28,000 --> 00:25:31,800
Speaker 1: apply AI in a way that's ethically consistent with their

426
00:25:31,840 --> 00:25:35,520
Speaker 1: mission and complies with the legal frameworks built to hold

427
00:25:35,560 --> 00:25:40,600
Speaker 1: the technology accountable. As AI adoption grows more and more widespread,

428
00:25:40,760 --> 00:25:45,080
Speaker 1: so too will the expectation from consumers and regulators that

429
00:25:45,200 --> 00:25:50,600
Speaker 1: businesses use it responsibly. Investing independable AI governance is a

430
00:25:50,640 --> 00:25:54,640
Speaker 1: way for businesses to lay the foundations for technology that

431
00:25:54,680 --> 00:25:58,560
Speaker 1: their customers can trust while rising to the challenge of

432
00:25:58,640 --> 00:26:04,480
Speaker 1: increasing regulators complexity. Though the emergence of AI does complicate

433
00:26:04,560 --> 00:26:09,119
Speaker 1: an already tough compliance landscape, businesses now face a creative

434
00:26:09,200 --> 00:26:13,840
Speaker 1: opportunity to set a precedent for what accountability in AI

435
00:26:13,960 --> 00:26:17,920
Speaker 1: looks like and rethink what it means to deploy trustworthy

436
00:26:18,359 --> 00:26:24,200
Speaker 1: artificial intelligence. I'm Malcolm Gladwell. This is a paid advertisement

437
00:26:24,600 --> 00:26:27,879
Speaker 1: from IBM. Smart Talks with IBM will be taking a

438
00:26:27,920 --> 00:26:32,040
Speaker 1: short hiatus, but look for new episodes in the coming weeks.

439
00:26:32,720 --> 00:26:36,080
Speaker 1: Smart Talks with IBM is produced by Matt Ramano, David

440
00:26:36,200 --> 00:26:41,240
Speaker 1: jaw nische Venkat and Royston Deserve with Jacob Goldstein. We're

441
00:26:41,359 --> 00:26:45,040
Speaker 1: edited by Lydia gene Kott. Our engineer is Jason Gambrel.

442
00:26:45,359 --> 00:26:50,320
Speaker 1: Theme song by Gramoscope. Special thanks to Carli Migliori, Andy Kelly,

443
00:26:50,720 --> 00:26:54,960
Speaker 1: Kathy Callahan and the eight Bar and IBM teams, as

444
00:26:54,960 --> 00:26:59,080
Speaker 1: well as the Pushkin marketing team. Smart Talks with IBM

445
00:26:59,359 --> 00:27:03,680
Speaker 1: is a production Pushkin Industries and Ruby Studio at iHeartMedia.

446
00:27:04,359 --> 00:27:08,480
Speaker 1: To find more Pushkin podcasts, listen on the iHeartRadio app,

447
00:27:08,760 --> 00:27:17,840
Speaker 1: Apple Podcasts, or wherever you listen to podcasts,