Latest / CXOTalk / Responsible AI, Public Policy, and Social Impact: A Conversation from the House of Lords
Transcript
- 0:00On episode #797 of CXO Talk, we're discussing a I and Public
- 0:09Policy with Lord Tim Clement Jones of the House of Lords.
- 0:15My guest cohost is Q Harrison Terry, the Chief Growth Officer
- 0:21for the Mark Cuban Companies. Yo, what's up?
- 0:24I'm glad to be here. And yeah, it's gonna be an
- 0:27exciting show. I'm in Israel today, so I've
- 0:30only got 30 minutes, but I'm excited to actually have a few
- 0:35conversations with Lord Tim Clement Jones and with that,
- 0:39Lord Tim Clement Jones. Welcome back to CXO Talk.
- 0:43Really good to be back, Michael, and good to be with you and Q.
- 0:47And very nice to meet Q as well. Tim.
- 0:50Tell us about your focus at the House of Lords as a member of
- 0:55the House of Lords. It's something that has really
- 0:58grown because I started off by being a digital.
- 1:02Or the digital spokesperson for my party in the House of Lords a
- 1:06few years ago. But now, as you know, everything
- 1:10has become digital. I've now become and the
- 1:13government have created a special department for science,
- 1:16innovation and technology. What I try and have to do is
- 1:20cover the full range of innovation and technology.
- 1:24So that is not just a I and all the tech around that, it's also
- 1:30life sciences, the. The application of new
- 1:34technology to health and other aspects, so it's a broad brief.
- 1:39You're working on bills that are regulating really ambiguous
- 1:45technologies related to our future, right?
- 1:49Like, I mean, AI is a big deal. I mean, you've done it.
- 1:51We've obviously played a part you've seen with the.
- 1:55I think GDPR was really good because it taught marketers like
- 1:58myself what we could and couldn't do and where to place
- 2:01data, and it gave us some standards.
- 2:04But marketing had existed for that.
- 2:07I don't know, like decades. We've been doing online
- 2:10marketing for decades. It just hadn't been as big as it
- 2:12is today. With AI, we're literally
- 2:15governing something that is in its infancy and would you give
- 2:20your kid training wheels before it could even get out the womb?
- 2:23And I feel like, you know, I'm seeing that argument being
- 2:26brought up a ton, especially here in the US My question for
- 2:29you is how do you even conceptualize something and and
- 2:32and and the impact that it's going to have on society in the
- 2:35next few decades? What you have to do is start
- 2:38from the risk aspect. You we don't really know and So
- 2:43what you have to do is try and get some guardrails in place
- 2:47where you do have to assess the risk, you know.
- 2:50Otherwise if you don't start now, when by the time we get to
- 2:54artificial general intelligence, it'll be too late.
- 2:58So what we have to do is try and work out what is appropriate.
- 3:01And, you know, one of our big debates at the moment is what is
- 3:04proportionate. In all of this, and you know a
- 3:08lot of people, there are people who believe that any form of
- 3:11regulation stifles innovation. I'm a believer in regulation.
- 3:17Proportionate regulation that makes sure that developers have
- 3:20some certainty around what they do.
- 3:22So actually that stimulates innovation.
- 3:25And I think there's a bit of a philosophical divide, but
- 3:30increasingly I think when people talk about a I, they do think
- 3:34that some guardrails are appropriate.
- 3:36I want to advocate for some of the companies that are being
- 3:40picked on or challenged in the landscape.
- 3:42Is this regulation just for that or for them?
- 3:46Because it's like if I'm meta and I say I'm getting an AI and
- 3:49then the UK parliament comes in and says, well, we've got an AI
- 3:52bill, it's almost like you're putting handcuffs on me.
- 3:55And it's not just meta, it's Google and Microsoft etc.
- 3:59Is that what this really is, or? If I may say is probably the
- 4:03attitude of a few years ago, the kind of move fast and break
- 4:06things type of philosophy. And we all know that that got us
- 4:10some places, but it also got us some detriment as well.
- 4:14And you know, I'm no doubt we're going to be talking about online
- 4:17safety and so on. And we don't want the same to
- 4:19happen with AI. We want to anticipate some of
- 4:22the issues. That are going to be associated
- 4:25with AI. So it is, it is right that we
- 4:27should be looking at this and I think increasingly those big
- 4:31tech companies now recognize that some form of guardrail,
- 4:36some form of regulation is appropriate.
- 4:38You know, if you hear people like Brad Smith, you, you listen
- 4:43to Sam Altman, you know all of those guys, they're beginning to
- 4:47to understand the need for it. And you know now when you talk
- 4:50to Meta. They're also concerned to make
- 4:53sure that there is some kind of regulation.
- 4:57The big issue is how far do you go in terms of the smaller
- 5:00companies? What do you regulate?
- 5:03How do you regulate? And in the case of online
- 5:06safety, you know what impact there is on freedom of speech,
- 5:09for instance. One of the real questions that I
- 5:13have about this is what is unique about AI?
- 5:17Because historically it would be almost unheard of that
- 5:22technologists would approach the government and say, hey, you
- 5:27need to regulate us. But with AI, as Tim, as you
- 5:32pointed out, that's beginning to happen.
- 5:35So what is causing this very unusual dynamic where folks are
- 5:42saying, hey government, you need to you need to regulate us?
- 5:46I absolutely understand why there is a little bit of
- 5:49bafflement if you don't know what the potential of AI is.
- 5:53It's partly about the black box nature of AI, the neural
- 5:57networks, the fact that they can make decisions and predictions
- 6:02which impact on you without reasons being completely
- 6:05understood. And that's a machine making
- 6:08those decisions, and you can't interrogate it in a in an
- 6:11appropriate kind of a way unless you design it ahead of the game.
- 6:16And it's the autonomous nature of AI too, where you may not
- 6:20have full control over what the AI is actually doing in terms of
- 6:25what what it's deciding, what data it's actually ingesting.
- 6:29And you know, now we've got these large language models that
- 6:33has become particularly acute because we don't actually know.
- 6:37What all the data that has been ingested by these large language
- 6:41models is, and you know, as I said a little bit earlier, what
- 6:45we're trying to do is anticipate some of the problems where AGI,
- 6:49artificial and general intelligence would have a, you
- 6:52know, greater degree of autonomy, a greater power.
- 6:56In terms of prediction and decision making and if you like
- 7:01just content creation than anything we've ever seen before.
- 7:07And so we do have to make sure that there is some form of
- 7:12ethical set of guardrails around that.
- 7:16Please subscribe to our YouTube channel, hit the subscribe
- 7:20button at the bottom of our website so that we can send you
- 7:24our newsletter and you will be notified of upcoming shows.
- 7:28What is driving this request for regulation from technologists?
- 7:36Is it is it fear or is it fear that the government may take
- 7:40control? Is it fear of their own
- 7:41creation, the A I creation? I know the narrative is somewhat
- 7:45polarized about AI, but you know you've seen 1000 technologists
- 7:49right in you've seen people like open AI and anthropic and so on.
- 7:56You know all join in with the the the view that we should
- 8:02even, I mean you know I don't actually accept that but a lot
- 8:06of technologists were asking for a moratorium on further
- 8:09development ending regulations. Now I think you know, I maybe I
- 8:13got a greater faith in. Politicians ability to move fast
- 8:17and regulate rather than break things.
- 8:19But you know, I do believe that there is a possibility of fairly
- 8:23quickly moving to at least standards so that we can get
- 8:28sets of standards for testing, sets of standards for risk and
- 8:32impact assessment, testing, standards for continuous
- 8:37monitoring. And even on an international
- 8:39basis and so get that in place whether or not it's a big
- 8:44victory, make it part of our technology culture and both
- 8:49sides of the Atlantic and I think we, you know we would have
- 8:52a great achievement there. One of the things that I think
- 8:55about, if we jump just a few years into the future, there's
- 8:58going to be a case where you're going to have to litigate AI.
- 9:01And the reason that I believe a lot of these entrepreneurs are
- 9:04raising their hands and they're saying hey challenges
- 9:06challenges, challenges is because they're lawyers are
- 9:09telling them hey, we need to see why a all of this and one way to
- 9:13do that is to go and not lobby not not don't send your lawyers
- 9:17in saying this is what we suggest but you as the
- 9:20entrepreneur say I don't know what's going on if something
- 9:23happens. I basically created some basis
- 9:28of identification from the jump. And so you can't, when you have
- 9:32to go litigate the AI, your founders are removed from that
- 9:37because they're gonna say, hey, I told you I didn't know what
- 9:39was gonna happen, and this could have massive deadly
- 9:42implications. I love that the the US approach
- 9:45to this kind of thing is litigation which establishes
- 9:49standards, is establishes norms. And you know when you have
- 9:53people like the National Institute for Standards and
- 9:56Technology setting standards as they are doing for risk
- 9:59management and so on, that becomes a norm.
- 10:02And therefore it's people are going to litigate over this and
- 10:05say to the technologies, look guys, you didn't conform to the
- 10:09risk management framework set out by the National Institute.
- 10:13And so they're going to be liable in in the court.
- 10:17Now, that's not the way Europeans tend to do it.
- 10:20We tend to try and regulate in advance of this kind of
- 10:24litigious behavior, if you like. But they're both valid.
- 10:28And they both lead to the creation of standards which can
- 10:32be common. And I'm very great believer in
- 10:36trying to converge as much as possible on these standards so
- 10:39that then technologists can get on with their job.
- 10:42You know, they can feel quite confident that this is what they
- 10:46need to do to avoid a black box or this is what they need to do
- 10:49to make sure they're not developing high risk AI systems,
- 10:53you know, so we've got. A lot of potential there.
- 10:56So to that extent I'm actually quite optimistic.
- 10:59Can you elaborate on the phrase you just used high risk AI
- 11:04systems? What qualifies?
- 11:07What pushes the application of this technology into that
- 11:12category? Well, let me take a US example,
- 11:15Michael. You know a lot of cities in the
- 11:18US, Oakland, Portland, Chicago, have banned the use of.
- 11:23Live facial recognition cameras. AI powered live live facial
- 11:28recognition cameras and OR systems and rightly so in my
- 11:34view pending a proper. A kind of overview for an
- 11:40ethical framework for the deployment of this kind of
- 11:43technology because it's seen as continuous surveillance
- 11:47basically, of our citizens. Now for me that is high risk and
- 11:52what you need to do before you install those cameras, before
- 11:55the police install them or the security services, what we need
- 11:59to do is have an assessment of the impact and how much people
- 12:04need to consent to that being deployed.
- 12:08The biases which may be inherent in the processing of the the
- 12:12system, you know, for instance, the racial prejudice that might
- 12:16be inherent in the data sets when they, you know, in a sense
- 12:20pick out particular people from the crowd, all that kind of
- 12:24thing. That, for me, is high risk and
- 12:27requires a greater degree of impact assessment.
- 12:30Before you deploy and that is the kind of regulation that I
- 12:35think is very appetite and it's of course what the EU are
- 12:38adopting currently. I wanna shift a little bit
- 12:41towards something that is just as important and it's the online
- 12:45safety. Bill.
- 12:46My question here is. I feel like I'm being over
- 12:50regulated. I'm being regulated on AI.
- 12:52I'm being regulated on all of my terms of service on all the
- 12:56websites that I now maintain that I have to enforce.
- 12:58Got to make the Internet a safer place.
- 13:00I have to control the users and control my content.
- 13:04I mean, what do we expect these tech companies to really become
- 13:08because it seems like a lot of the legislation that you're
- 13:12you're proposing is. Is just telling me that I have
- 13:16to follow the rules and I can't think outside the box.
- 13:19Thus I can't bring you the devices or the services that I
- 13:23want to. I have to follow the line and
- 13:26and pretty much we're all going to look.
- 13:28We're all going to be using the same devices, the same websites,
- 13:31because it's going to cost too much to to to think outside the
- 13:35box. Having spent over the last two
- 13:37or three years working on the online safety bill, which
- 13:42hopefully will become an act, you know.
- 13:44September, because we're very, very nearly there.
- 13:47I think we've managed to work our way through without over
- 13:53regulating and without impinging too much on freedom of
- 13:56expression. Because what we're really doing
- 13:58now, we're really regulating for harms against children.
- 14:02And everybody would agree that what you don't want is minors
- 14:06suffering harms. So you've got to define those
- 14:09harms and make sure the platforms adhere to that so they
- 14:13don't expose those. Children to harms and the second
- 14:16thing is illegal behavior. And our principle has been what
- 14:21is illegal offline should be illegal online.
- 14:24And that's been the principle we've tried to establish
- 14:27everything else. Is OK as long as everybody knows
- 14:31when they get on a platform, this is what the terms of
- 14:34service provides. So if you're in a sort of a very
- 14:38free speech environment and that's the terms of service,
- 14:42like parts of Reddit for instance, the community that you
- 14:46know is quite explicit about the things they talk about for
- 14:49adults, that's fine. We've also arranged it so that
- 14:53it's obligatory to have certain user.
- 14:56Empowerment tools. So if you don't want to see
- 14:58particular content, you have the right to press a button and say
- 15:03that's the kind of content I don't want to see.
- 15:05And that's freedom of choice for the citizen, basically.
- 15:08So we've tried to work our way through in all of this without
- 15:14being too negative about, you know, new innovations and
- 15:18platforms and also without impinging too much on the
- 15:21smaller platforms. Now, you know, we have a debate
- 15:24about whether you should. Regulate those who are risky as
- 15:30opposed to those who are large. And we haven't yet quite
- 15:33resolved that. You know, at the moment we're
- 15:35regulating the larger platforms, not the risky platforms.
- 15:39Now I believe that even smaller platforms can be quite risky,
- 15:44but that's a debate to be had continuously.
- 15:46You know, we'll see how the bill works out in future if we don't
- 15:51manage to establish risk as the basis for regulation.
- 15:54But nevertheless, that's the. About the last remaining debate
- 15:58that we've got, because what we're regulating is the
- 16:01functions of the systems. We're not regulating content so
- 16:06much. It's how the amplification of
- 16:09the algorithm works and so on. Really important stuff.
- 16:12Again, we're back to the black box.
- 16:15We mustn't have those black boxes, but the algorithms are
- 16:19inextricably linked with the content.
- 16:24So how do you regulate the container without getting into
- 16:30the messy set of judgment calls for the content that is being
- 16:36delivered? It's all about transparency as
- 16:39as indeed, and this is the common link with AI regulation
- 16:43as well. It's all about transparency.
- 16:45What we ask is for the platforms to show their risk assessments
- 16:52of their platforms and the technology and the algorithms
- 16:54they're using. And then the regulator basically
- 16:59sees. What they say about the impact,
- 17:02because that's what the platforms do.
- 17:04So it's not as if the it's not as if except in extreme cases
- 17:10where the regulator steps in and actually looks inside the
- 17:13algorithm or gets a technological report on the
- 17:18algorithm. By and large it's a lot about
- 17:20self reporting. It's a lot about saying these
- 17:23are the harms we discovered and also we want to see researcher
- 17:27access as well. So there is some independence.
- 17:30Third party validation, but it's not a very heavy hand.
- 17:33Actually a lot of it is. This is being done by codes of
- 17:36practice and so you just expect people to follow the codes and
- 17:41and and it's, you know, it's mainly good, good practice.
- 17:46I mean, we've got some pretty big companies already starting
- 17:50to follow the codes. Just the other day, Meta
- 17:53announced that they're open sourcing one of their large
- 17:56language models, LAMA. And I'm curious your take on
- 18:00that, like is that what you want to see from the industry or is
- 18:03that still just another press release?
- 18:05I think that's a slightly different thing, strangely
- 18:07enough, because you you could say that an open source a a
- 18:12powerful open source a I is like opening a Pandora's box on, you
- 18:18know, a very powerful creature. And you know, I, I I feel quite
- 18:25nervous about that. I mean the other the chat.
- 18:28GPTG, PT4 and and Ember, you know which is the platform on
- 18:34which Bard is based. They're not open source And so,
- 18:40you know, being able to utilize an open source like Lama, I, I I
- 18:46have my doubts about, I mean that to me is quite high risk
- 18:51and I think what I would like to see is a full impact assessment
- 18:56from Meta. About that, and I think, I think
- 19:00we'd all be greatly reassured if we could see that.
- 19:03But I, I, I, I'm quite nervous about the the ability to build,
- 19:10you know, quite malign systems off the back of an open source
- 19:14AI as powerful as Lama. I mean, let's face it, the only
- 19:18people who can build these massive generative AI systems
- 19:24are those who a have, you know, extra compute power and massive.
- 19:28Data sets. And so you know, this is only a
- 19:31limited number of big tech companies who can do that and
- 19:35therefore being able to piggyback on an open source
- 19:40system like that is quite something for any tech player.
- 19:44So at the end of the day, what is the the problem with these
- 19:52mega players and their AI? Are you concerned that there is
- 19:57a an over consolidation of power and control and therefore impact
- 20:02on society? In terms of ethical regulation,
- 20:07I mean, you know, the big tech companies, they've for quite a
- 20:10long time they have responded. You know, we've talked about
- 20:14responsible AI. They have dialogues all the time
- 20:20with politicians like myself. So I think there's a general
- 20:24acceptance that there is a need for ethical behavior.
- 20:28And indeed, now there's a growing acceptance of the need
- 20:31for regulation, or at least for standards.
- 20:36I think what is really going to be in dispute is whether or not
- 20:41there should be greater control over the competition aspect and
- 20:45what by that I mean opening up competition.
- 20:48Because, you know, as I said earlier, with the exocompu power
- 20:53and the huge data sets that are needed currently in the current
- 20:57form of machine learning, it may change in the future.
- 21:00But at the moment, you know, it means that you're concentrating
- 21:04power in the hands of Microsoft and meta, Apple, Google and so
- 21:11on. And, you know, we need to make
- 21:13sure that they actually don't behave in any kind of
- 21:17monopolistic way. I mean, I'm afraid I'm a science
- 21:21fiction reader. And um, you know, when I look
- 21:24back at UH, William Gibson and UH, things like Neuromancer, UH,
- 21:28they were competing AI systems, but very limited AI systems.
- 21:33And I don't want to get to a situation where, for instance,
- 21:37we have the same situation we have in app stores, you know,
- 21:42where there's a very limited choice of app stores.
- 21:45I don't want to have a very limited choice of AI systems
- 21:49simply because there are only four or five the fangs, only
- 21:54four or five companies that can actually build these things.
- 21:58We've got very interesting question from a friend of both
- 22:01Mike and myself. Jav is over at Harvard Business
- 22:06School and he's asking the question, is there concern that
- 22:09the regulation in the UK and the EU will slow down innovation
- 22:12within the jurisdiction. So already a I development
- 22:16within the US and China is much further ahead and regulation in
- 22:20the UK should make it worse, right?
- 22:23Actually not. I think that it's going to
- 22:25create a certain confidence in the ability to do this stuff.
- 22:30Don't forget that if you add the UK to Europe, you know you're
- 22:33well over 450 million consumers. And if you then establish norms
- 22:39and you will have the same norms in the States, believe me, the
- 22:43standards will be virtually identical.
- 22:45At the end of the day, that's what developers will want.
- 22:48That's the market that developers will want.
- 22:50They don't. They will.
- 22:51They will be. Very silly if they start trying
- 22:54to develop AI which are effectively black box because
- 22:58they'll find it's not when business, when business and
- 23:02government and so on procure AI. They won't procure black box
- 23:07systems. If you don't get kite marked or
- 23:10certificated as being a conforming AI system for a
- 23:14consumer, for instance, you won't be able to sell your
- 23:17product to the consumer. And the same is going to be true
- 23:20of $0.10 and Alibaba and all the Chinese companies as well.
- 23:23And in fact, of course, the Chinese government of adopted
- 23:27standards or are adopting standards for generative AI
- 23:31systems themselves. They may not apply to their own
- 23:34government systems, but that's not the commercial aspect.
- 23:38The commercial aspect is both in data and in AI, private Chinese
- 23:43companies, such as they are, will need to conform, and when
- 23:47they export or sell their services or systems abroad, they
- 23:52will have to conform. So we have another really
- 23:55interesting question from Twitter, and this is from
- 23:59Elizabeth Shaw, who says how well would using the UK harm
- 24:06spill as a work as a template for regulating the use of AI.
- 24:12I think it's going to be a very, very different story for a I
- 24:17regulation, because it won't involve anything like the
- 24:20difficulties that we've had over freedom of expression.
- 24:23Because what we've absolutely had to do is try and make sure
- 24:28that we don't close down free speech on our social media
- 24:33platforms. And therefore, we've had to be
- 24:35very, very particular and quite complicated in the way we do it.
- 24:40I think that AI regulation is going to be much more
- 24:44straightforward than that. It'll be simply look before you
- 24:47institute AI systems or procure them or adopt them or develop
- 24:53them and depending on you know what stage you get to, you've
- 24:57got to carry out. It's rather like the data
- 25:00protection impact assessment. You've got to carry out that
- 25:03assessment if it turns out that it's high risk.
- 25:06These are the kinds of things you've got to do in terms of
- 25:08transparency, further work on bias elimination, further work
- 25:15on explanation and being very clear about liability in these
- 25:21circumstances. So you know this is going to be
- 25:24a set of requirements. These are the standards you
- 25:26would stopped. They could be those from the
- 25:28International standards organization, they could be
- 25:31those that NIST has put forward or IEEE so on.
- 25:36Those are all coming together. It's going to be much more
- 25:38straight, it's going to be much more like product liability.
- 25:41I mean, you know, we already have product liability laws in
- 25:45the UK which are pretty clear. We have health and safety laws
- 25:50which are even clearer. Um, uh, and so does Europe.
- 25:54In fact, ours are based on uh those that we had when we were
- 25:58in the EU. And so that's much more like
- 26:01product liability in a much more straightforward kind of a way.
- 26:05But we've got to make sure that in the way that Hugh was talking
- 26:09earlier, we can't have developers say no, no, that was
- 26:12an AI system, nothing to do with me.
- 26:15You know it's been running around on its own.
- 26:17I don't take liability for that, you know, So it's that's a
- 26:21really important bit of liability establishment that we
- 26:24need. The whole concept of AI
- 26:28regulation today is something that I feel like is very narrow
- 26:31to where we are today. Like as a future thinker, one of
- 26:35the biggest skill sets that we all possess is being able to
- 26:37separate from today, from tomorrow and then from tomorrow
- 26:41to like 10 years. So let's just say for the sake
- 26:44of this, this, this, this discussion you every everything
- 26:49you're working on goes exactly as as as planned.
- 26:52What does the vision of tomorrow and even 10 years into the
- 26:57future look like if all these things are running as they are
- 27:01being designed? One of the things that's most
- 27:04difficult is to future proof our regulation and.
- 27:07But that's exactly what we need to do both in online safety and
- 27:11in terms of the general A I approach.
- 27:14And I believe that actually at the end of the day what we need
- 27:17to do is have a system where A I is our servant, not our master.
- 27:21And in 10 years time that's got to be the case.
- 27:24And that's what I'm working towards.
- 27:26So that it's a tool that we use in our employment.
- 27:29It helps us do our jobs. It doesn't substitute for us all
- 27:34the time. It's something that improves the
- 27:37quality of our lives in everything we do.
- 27:40Now because I'm an optimist about the about AI, you know, I
- 27:44think it's like electricity or you know, the printing press,
- 27:49you know, it's a it's a beneficial technology that we've
- 27:52got to tab with both hands, but we've got to do it very clearly
- 27:56and knowing what the risks are. Lord Tim, you've got some big
- 28:00responsibilities. I'm going to be looking toward
- 28:02you in 10 years with AI has replaced me and it's taken my
- 28:07job and I'm trying to figure out, you know, what I'm going to
- 28:10do with society. But because you're here,
- 28:12hopefully that won't be a scenario that I'm living in.
- 28:16Unfortunately, chunks of you being substituted.
- 28:19Yeah, I love that. I'm going to take that in, put
- 28:23that on my resume, if you don't mind.
- 28:26Likewise. Thank you all.
- 28:28It's been a pleasure for today. I do.
- 28:30I do have to run, unfortunately, but this is a great
- 28:33conversation. Thanks you.
- 28:35Take care. Bye, bye.
- 28:37Tim. When it comes to this kind of
- 28:42regulation, what are the biggest?
- 28:45Challenges in balancing the competing interests, for
- 28:51example, technological project versus innovation which you
- 28:56alluded to earlier. I think it's really approach to
- 28:59risk and I think that you know there's going to be inevitably a
- 29:04difference of approach in different parts of the world as
- 29:09to what is considered high risk. And I think that's going to be
- 29:14where some of the debate and the argument is going to be what it
- 29:17what do I need to do if you know you determine that something is
- 29:22high risk as I might have a different opinion about that and
- 29:27you know I think we've got to get some, some view about that
- 29:30now. It's quite interesting.
- 29:31I think in the States you have, you know, very strong views
- 29:34about civil liberties, about surveillance and the rights of
- 29:39privacy and so on. It's more about data in our case
- 29:43because obviously your social media companies in a sense make
- 29:46their money from data. You know, it's that kind of
- 29:52approach that's been traditional, um, so, but we
- 29:55we've pushed back against that quite heavily with GDPR.
- 29:59So there are going to be cultural differences on this and
- 30:01I think that's going to be the big challenge.
- 30:04Um, uh, uh, you know? And then of course on weaponry,
- 30:08which we haven't touched on yet, That is where there's going to
- 30:12be even bigger debate because you know, lethal autonomous
- 30:15weapons for a start, you have to define what an autonomous weapon
- 30:18is and you have to decide whether or not that's going to
- 30:22put you at a competitive disadvantage internationally
- 30:26unless you, if you do decide to limit the use of them.
- 30:30But at the end of the day, and and also the whole business
- 30:33about whether it's offensive or defensive and whether it's
- 30:37appropriate to use it defensively rather than
- 30:39offensively. But you know, and then again, if
- 30:43you use an autonomous weapon in relation to nuclear weapons,
- 30:47that's even more high risk than you can possibly imagine.
- 30:51So there are a lot of questions there which I think we, you
- 30:55know, we need to resolve, but we do need to start resolving them.
- 30:59Honestly, how practical is this, even when you talk about lethal
- 31:04autonomous weapons? And the ability to get countries
- 31:10to agree on standards? Obviously, that's far more
- 31:13complex than regulating Google, Facebook or any other company.
- 31:18Yes, it is. But of course, we already have
- 31:20international humanitarian law, which applies across the
- 31:24battlefield. And So what you need to do is
- 31:29make sure that international human rights law actually is fit
- 31:35for purpose and that it does cover those kinds of autonomous
- 31:39weapons already. Things like drones, for
- 31:41instance, that are that are on the battlefield that may have
- 31:45quite a high degree of autonomy. So we've got to make sure that
- 31:48those are covered and that the and that we know when it's
- 31:52appropriate to have human in the loop.
- 31:54You know, it may be that certain weapons must have a human in the
- 31:58loop, otherwise there is a breach of international
- 32:01humanitarian law. So we've got some of the tools
- 32:04for this. You know, I'm we know we're not
- 32:07in Wild West country here. We can, we can move towards
- 32:10agreement on that. We have an interesting question
- 32:14from Twitter, and this is from Chris Peterson, who's asks he
- 32:19says, has Brexit complicated the regulatory picture or hurt the
- 32:25UK's ability to attract people and companies in the AI space?
- 32:30And I think in the medium term that is already happening.
- 32:34Basically it is much more difficult now to attract and
- 32:40move people around, you know, even the high skilled ones.
- 32:43We do have exceptions for high skilled technologists, but the
- 32:48visa charges are eye watering compared to many other
- 32:52countries. And of course between European
- 32:55members of the European Union. You can travel, these are free,
- 32:58you can work anywhere within the 27, you know.
- 33:01So it's a real, it's a real problem for us.
- 33:04And of course, I would much prefer us to be regulating on
- 33:08the basis of 28 rather than 27. But, you know, that's history
- 33:12now, I'm afraid. And what we have to do is try
- 33:15and come to agreement with the EU about making sure that we do
- 33:21have quite a high degree of convergence in our regulatory
- 33:25systems. It would be crazy for us as a
- 33:28country, no more than 70 million, to start trying to
- 33:32establish completely different rules from the EU, who now have
- 33:36begun to really establish a suite of rules relating to a I
- 33:41social media, digital markets and so on.
- 33:45We mustn't diverge too much otherwise, you know, why should
- 33:48American companies, for instance, come and establish
- 33:51here when they they went to Paris or Berlin, They'd have a
- 33:55bigger market. They could attract, you know,
- 33:58all the postgraduates that they would need from universities
- 34:02across the EU far more easily. So you know, we've got to make
- 34:06sure these barriers to skills and and jobs and so on are much
- 34:12lower than they are at the moment.
- 34:13I know that you said earlier that your focus is not on the
- 34:19content so much as on things like the algorithms, but when we
- 34:25start to talk about misinformation, disinformation,
- 34:31and even in the case of large language models, hallucinations
- 34:37where there's no ill intent. But the machine simply gives you
- 34:42an incorrect answer with absolute 100% confidence.
- 34:46How does how does all of that factor into the regulation, the
- 34:51regulate regulatory kind of framework?
- 34:53You can't is the answer. All you can do is there are two
- 34:57things you can do. You can be very advertent to how
- 35:01the system amplifies information, misinformation,
- 35:06disinformation. And therefore what you've got to
- 35:08try and do is make sure that the media, the social media
- 35:12platforms are very conscious about that and that they have
- 35:17provenance checking systems, fact checking systems and so on.
- 35:21I wouldn't make that compulsory, but I think it should be part of
- 35:24your terms of service that you know, Where possible, you try
- 35:28and check the provenance of the sources.
- 35:33If you're a social media platform putting out, say,
- 35:36public health information, that's misleading and so on.
- 35:38So you've got you've got that. And the other is, you've got to
- 35:42make sure your population is equipped that they're
- 35:45sufficiently media and digitally literate, and to make sure that
- 35:50they've got at least some idea of what is being fed to them
- 35:54now, because I don't think there's any easy solution to
- 35:58disinformation and misinformation.
- 36:00It's something we're going to have to live with.
- 36:01You know, OK, we might recognize whether there's a bot being
- 36:06pushing out, you know, whether it's an inauthentic, what they
- 36:09call an inauthentic account that's, you know, different
- 36:13story. But the ordinary pushing out of
- 36:16misleading information, I mean that is something where as
- 36:21citizens we have to have a, you know, we have to be a bit more
- 36:24sophisticated than we are now. But, you know, let's face it,
- 36:28there are still many people in your country who believe that
- 36:31the election was stolen. And I'm afraid, you know, how do
- 36:35you, how do you treat that? Is that disinformation,
- 36:37misinformation? Many people believe that to be
- 36:40true. We have another interesting
- 36:43question on Twitter from Chris Peterson, who comes back and on
- 36:48the subject of an educated populist.
- 36:51And I think this also gets directly to you, to the point
- 36:53you were just making, he says this.
- 36:55Listening to Lord Clement Jones makes me wish pretty much anyone
- 37:00in quotes. He says our upper house, the
- 37:04Senate, had his not your knowledge, understanding and
- 37:08perspective in the US. And I guess is a rhetorical
- 37:12question. How can we expect the population
- 37:16to be educated when frankly our own politicians seem to know
- 37:20very little about the subjects in which they're discussing?
- 37:24I am very heartened by the fact that people like Chuck Schumer,
- 37:27for instance, the Majority Leader and the Senate is very
- 37:31interested in this subject. And he's got a working group.
- 37:35And I think that the, you know, senior Democrats in in Congress,
- 37:39they want to see legislation along these lines.
- 37:42And it could easily be bipartisan because quite
- 37:44frankly, you know a I which is black box affects all the
- 37:49population and I I believe there's a great room for parties
- 37:55and agreement on that kind of thing.
- 37:57And of course the White House itself, we've seen the blueprint
- 38:00for a digital A I Bill of Rights.
- 38:05We've seen the White House even today I believe who've got all
- 38:10the big tech CE O's in today to talk to them about testing and
- 38:16getting their commitment to particular testing of a I
- 38:21systems before they're put into use.
- 38:23So I think there's quite a lot happening there.
- 38:27We may not have the answers, but I think that both in Brussels
- 38:32and London and in DCI think there's a growing understanding
- 38:38of what we need to do and what the issues are.
- 38:40In that case, what advice do you have for policymakers and for
- 38:47politicians who want to address these issues in a substantive
- 38:53way? Our Prime Minister, having
- 38:56produced proposals which were very weak in the regulatory area
- 39:01earlier on this year in what we call a white paper now, has
- 39:05suggested we convene in London this December, November,
- 39:10December, an A I safety conference, an internationally
- 39:15global conference. Now I very, very much support
- 39:18that. So I would say, look, come
- 39:20together, bring, you know, let's involve the Chinese government
- 39:25as well. Let's bring in Chinese
- 39:29technologists and let's bring EU leaders together.
- 39:34Let's bring congressional leaders together as well.
- 39:38Come to London and let's try and converge on a particular set of
- 39:44solutions. Maybe we won't be able to agree
- 39:48on how much we need to regulate across the board, but what we
- 39:53can agree on is what kind of standards for safety, for
- 39:58freedom, from bias, for transparency, accountability and
- 40:04so on we need to put into effect.
- 40:07I think that's really. Where we should be going?
- 40:10What about advice for business people and technologists who are
- 40:17creating these technologies? I think what we do is we keep
- 40:21the dialogue open. I I talk a lot to technologists
- 40:25and I know that this idea that, you know, some governments have
- 40:29that innovation is the enemy of, that innovation is the enemy of
- 40:35regulation and vice versa is is wrong.
- 40:37I mean, most technologists I talk to are just keen for us to
- 40:41get on with it, to get some guardrails around this so that
- 40:44they can get on with, you know, understanding what they need to
- 40:48do to conform to regulation. And therefore their job as
- 40:54developers and adopters is made much easier.
- 40:57And I think they breathe a sigh of relief, and especially if we
- 41:01got some sort of international agreement on standards.
- 41:04As we finish up, Tim, any final thoughts that you would like to
- 41:07share? We were all a bit surprised by
- 41:09the speed at which CHAT, GPTG, PT4 and then Bard all came
- 41:15thundering down the track. Now we've had Llama in the last
- 41:18couple of days and the speed of development is unbelievable.
- 41:23But at the same time the narrative has been very, very
- 41:26confused. You know we've had 1000
- 41:28technologists on one side telling us that we're doomed
- 41:32more or less and the other side saying no, no it's great,
- 41:34another thousand the other day saying it's all great.
- 41:37So I think people might take, you know, might find it quite
- 41:42difficult to understand what the agenda is.
- 41:45But what it has done is made politicians think very, very
- 41:49carefully about what we need to do.
- 41:51It's heightened consciousness and it's made us think about the
- 41:54future, and it's made us think that things like artificial
- 41:58general intelligence are coming along much quicker than we ever
- 42:01thought. And that is only to the good,
- 42:04you know, it's really made sure that we're all, you know, we've
- 42:08got to speed up, we've got to, we've got to get going and make
- 42:12sure that we do the right thing for our our citizens.
- 42:16You sound very optimistic. I'm always optimistic until I
- 42:20fail because I believe that if you don't throw a lot of energy
- 42:25into this, you will fail because, you know, a lot of
- 42:28people say, no, we don't need this because you know, you'll
- 42:32stifle innovation. Oh, we can't have this because
- 42:35the Chinese will get ahead of us, you know, and so on and so
- 42:38forth. I think there are a lot of
- 42:39spurious arguments out there and I don't want them to prevail.
- 42:44So I spent quite a lot of energy trying to make sure that we
- 42:48adopt practical solutions, and there are a lot of practical
- 42:52solutions out there. And I think strangely enough,
- 42:54we've just demonstrated it with our online safety bill that it
- 42:58is possible even in a really complicated world, to do stuff
- 43:03to regulate and also to some extent future proof as well.
- 43:07Well, congratulations on moving your online safety bill forward.
- 43:13And with that, we're out of time.
- 43:16And I want to say a huge thank you to Lord Tim Clement Jones of
- 43:21the House of Lords for taking time to be with us today.
- 43:24Tim, thank you so much. Huge pleasure, Michael.
- 43:27And I hope you'll come back another time.
- 43:29Definitely you won't keep me away.
- 43:32And a huge thank you to our great audience.
- 43:36You guys ask the best questions, very smart, intelligent,
- 43:40sophisticated audience and we love that.
- 43:43Now before you go, please subscribe to our YouTube
- 43:48channel. Hit the subscribe button at the
- 43:50bottom of our website so that we can send you our newsletter and
- 43:55you will be notified of upcoming shows.
- 43:57Check out cxotalk.com. Have a great day everybody and
- 44:01we'll see you soon. Bye, bye.