Latest / CXOTalk / AI Regulation & Innovation: Insights from the UK House of Lords | CXOTalk #875
Transcript
- 0:00Welcome to CXO Talk episode 875. Today we're honored to host 2
- 0:07distinguished members of the UK House of Lords, Lord Chris
- 0:12Holmes and Lord Tim Clement Jones.
- 0:15We'll explore how policy makers at the highest level balance
- 0:19innovation, regulation, and public trust and technology,
- 0:24specifically related to AI, digital assets, and open data.
- 0:31Let's get into it, gentlemen. Welcome to CXO talk.
- 0:34It's great to see you both. Hi, Michael, be here.
- 0:38Lord Holmes, tell us, why do we care about this topic of AI
- 0:43regulation and governance in this way and and especially at
- 0:47this time? They're incredibly powerful
- 0:49technologies, but they're in our human hands.
- 0:52And if we're going to optimize their potential, what's being
- 0:56cognizant of their challenges and their risks, It seems
- 1:00entirely logical as we have done through centuries with other
- 1:05advancements and technologies that we consider the right
- 1:09regulatory framework. So we balance all of the
- 1:12competing needs and we put in what I would describe as right
- 1:16sized regulation, right sized regulation, good for innovation,
- 1:21good for investor, good for citizen, good for creative, good
- 1:26for consumer. Tim, why is all of this so
- 1:29beneficial and how do we avoid interfering, presenting
- 1:33obstacles to innovation, as Chris just said?
- 1:38Both Chris and I believe that actually regulation can help
- 1:42innovation. And if you basically get it
- 1:45right in terms of clarity, clarity, consistency and
- 1:50certainty for business, for consumers, for developers, that
- 1:57will actually lead to a much better and safer form of
- 2:03innovation at the end of the day.
- 2:05And that's what we want and we want to see it adoption.
- 2:07We want to see roll out of new technologies, but they've got to
- 2:11be for our benefit. And you know, as you and I have
- 2:15discussed over the years, Michael, this technology has
- 2:18been advancing at an incredibly fast rate.
- 2:21You know, back in November 22, for instance, suddenly ChatGPT
- 2:27burst on the world and people have been struggling to catch up
- 2:32ever since. And now, of course, this year
- 2:34people are talking about agentic AI, which is, you know, much
- 2:38more autonomous. So you've got this mixture of
- 2:41large language models and autonomy.
- 2:44And so you've absolutely got to have some kind of guardrails
- 2:47around that. Chris, when we talk about the
- 2:51the guardrails and the the necessity for protection, how do
- 2:58you implement that type of governance, but at the same time
- 3:01ensure that there is there are no obstacles or or minimal
- 3:05obstacles to innovation Going back to your initial comment.
- 3:09We know what we need to know to make a success of this.
- 3:15All of history suggests to me that we know how to do right
- 3:19size regulation. We all know bad regulation.
- 3:24But that's bad regulation. That doesn't mean that as a
- 3:27consequence regulation is bad. I think if we take a principles
- 3:33based, outcomes focused and inputs understood approach, we
- 3:39give ourselves the best chance of making a success of this and
- 3:43to those principles. Trust and transparency,
- 3:46inclusion and innovation, interoperability and
- 3:49international perspective, accountability, assurance and
- 3:53accessibility. Good principles I would suggest
- 3:57for AI regulation. Good principles, all regulation.
- 4:02And in the UK, the previous government in its white paper
- 4:07set out principles but set them out on a voluntary basis.
- 4:12But they're good principles. But if they're good principles,
- 4:15it seems logical that one would want them on a statutory basis
- 4:20to give that opportunity for whichever part of society,
- 4:25whichever part of the public or private sector, whichever part
- 4:28of the economy you find yourself in, you're likely to experience
- 4:35the same approach to AI, the same guardrails around AI.
- 4:41If it was Tim said. Whether you're an individual,
- 4:44whether you're a consumer, whether you're a business,
- 4:47whether you're an innovator, what ultimately you want,
- 4:51certainty, clarity, consistency. Because from those three CS
- 4:57comes what you really need as a human to enable action
- 5:01confidence. How do you get there?
- 5:04But I think by putting it in in place, a regulatory architecture
- 5:07which is rooted in those principles, crucially is
- 5:12horizontally focused. So it's cross sector, it goes
- 5:15right across the economy. By doing that, you're able to
- 5:18identify areas where there is minimal or indeed no regulatory
- 5:25cover or regulator in the UK. Recruitment being an obvious
- 5:28example to bring to bear on that.
- 5:30And if you if you have that framework, it has the principles
- 5:36at its core, but it has the agility and the dynamism to
- 5:40develop, particularly in our common law jurisdiction in the
- 5:44United Kingdom. It can develop, it can be agile.
- 5:48And for a comparator, if one looks at other more prescriptive
- 5:53jurisdictions, there's no suggestion necessarily that
- 5:56prescription is wrong of itself, but it is necessarily different.
- 6:02And prescription necessarily tends towards trying to capture
- 6:07every last element in that legislation, which always has a
- 6:12danger of trapping that legislation that point in time
- 6:16or ending up being over controlling through that over
- 6:21prescription. So I think by having agility,
- 6:24flexibility, it's entirely possible.
- 6:27It's what all governments, it's what all regulators should be
- 6:30seeking to hold at the same time, the needs of the citizen,
- 6:34the needs of the innovator, the needs of the investor.
- 6:38It makes it more complex, but that is completely what needs to
- 6:42be if you're going to have this right size regulatory approach.
- 6:45I think any form of regulation in the world of AI does have to
- 6:49be risk based. So and that gets you to a kind
- 6:52of proportionate approach to regulation.
- 6:55And the other point and Chris used the word interoperability
- 6:58earlier and that I think in EU KS context particularly, but I
- 7:03think also worldwide, if we're going to see AI applications
- 7:08thrive across the world, not just sitting one jurisdiction or
- 7:13another is the adoption of international standards.
- 7:16And I think that is really the big challenge that we've got.
- 7:20So whatever form of regulation in any jurisdiction, because the
- 7:25US is going to be different from Europe, is going to be different
- 7:28from the UK. What we need to do is make sure
- 7:31that developers and businesses and adopters and so on aren't
- 7:35trammelled by regulation too much.
- 7:37They really can understand what standards they're meant to
- 7:41adhere to, whether they're of safety or ethics or you know,
- 7:44how however you might describe it.
- 7:46And those are the standards which are currently in the
- 7:49offing and are being developed. And I think that's one of the
- 7:52most positive aspects at the moment.
- 7:55I think that's right, Tim. And building on that standards
- 7:58point, that sense of bringing the world together and having
- 8:03that international, collaborative, relational
- 8:08approach has to be right anytime, but particularly at the
- 8:10moment, it would seem extraordinarily pressing because
- 8:15to go to the US, people would be mistaken.
- 8:19But it's easy to be mistaken to think in the AI space, there's
- 8:23no legislation, there's no regulation.
- 8:25But when you go into more detail and you see what's happening at
- 8:28a state level, different legislative and regulatory
- 8:32instruments being brought to bear, it's far more interesting
- 8:36and far more complex than that. And it demonstrates again a need
- 8:40to bring some interoperable approach to it.
- 8:44They'll need to be at some stage some federal activity in the
- 8:49space for sure. For that.
- 8:50I'm really keen and brought into the all of the legislation I've
- 8:57done in this space, this sense of really looking at that
- 9:01international piece and it and the standard is a key part of
- 9:05it. But also, we could potentially
- 9:07do nothing short of potentially reinvigorate so many of those
- 9:13international organizations set up in the wake of the horrors of
- 9:16the Second World War. And if we could get a form of
- 9:21engagements and approach from all of those international
- 9:23organizations, that could be extraordinary powerful for them
- 9:26and indeed for the entire planet.
- 9:29I just want to tell everybody that right now you can ask your
- 9:35questions on Twitter using the hashtag CXO Talk.
- 9:39And if you're watching on LinkedIn, just pop your
- 9:42questions into the LinkedIn chat.
- 9:44I urge you to ask questions because when else can you ask to
- 9:51members of the House of Lords pretty much whatever you want on
- 9:54these topics. So ask your questions.
- 9:57Now. You've both been talking about
- 10:01international collaboration. Seems like a a very apartment
- 10:07discussion today with what's going on.
- 10:11So what are your thoughts in this environment where we have
- 10:18tariffs, where the world economy is being upended because
- 10:24collaboration doesn't seem to be happening.
- 10:26So what's the impact there on AI and and the the development of
- 10:34these tools that we all want to develop properly?
- 10:37People, you know, like us who work in the digital world find
- 10:40it very ironic that all the debate coming from the States is
- 10:45about the, the world of physical goods, the trade in physical
- 10:50goods, the deficit in physical goods from the states in terms
- 10:56of, you know, the states feeling it has to impose tariffs on
- 10:59those imports. Whereas actually the, the United
- 11:04States now is, you know, the world's most powerful country in
- 11:09terms of digital services. And, you know, the big tech
- 11:13companies, Microsoft, Google, Amazon, they are immensely
- 11:18powerful. And I I think that is the
- 11:21countervailing power in the economy of the US, which is not
- 11:26really being taken account of. And it was what was the digital
- 11:30services tax was going to try to redress.
- 11:33But of course, now that he's up for grabs and, you know, there
- 11:37could well be some kind of trade off from what we see in the
- 11:41media about what our government's up to.
- 11:43And it may be what Europe may have to trade off to some
- 11:47extent. But it is ironic because
- 11:50actually you should really take the two things together and say,
- 11:54well, well, look, you know, we need, if we're going to come to
- 11:57agreement, we have to take account of the fact that the US
- 12:01companies, big tech companies are hugely dominant.
- 12:05It seems extraordinary analogue in a digital world or
- 12:09extraordinarily automotive in an algorithmic world.
- 12:13The automotive industry is obviously incredibly important
- 12:18and significant worldwide, but to have something which looks at
- 12:25this through a lens doesn't look so much on services, on the
- 12:30digital economy, if you will. And also I was under the
- 12:36impression that quite a number of centuries ago we'd resolved
- 12:41this issue of international trade.
- 12:45I very much remember reading Adam Smith.
- 12:48It's only makes sense to have international trade and all of
- 12:54the economic and indeed social benefits that flow from that.
- 12:59Of course there are questions around dumping, there are
- 13:03questions around other forms of protectionism.
- 13:07But as a cardinal principle, largely, we can win as a
- 13:14connected relating human society if we try and establish and
- 13:19enable that route to free trade, as opposed to going an
- 13:24alternative route where we're only beginning to start to see
- 13:28the more than significant consequences.
- 13:32Tim doesn't seem to me like right at the moment, the idea of
- 13:36a connected world society is among the primary goals, let's
- 13:40just say, at least in the US. No, that's very true.
- 13:43In fact, it's almost, it's a disrupted society.
- 13:47And and it seems that there is some goal to actually disrupt.
- 13:54Now, you know, we're told that actually this is going to lead
- 13:57to, you know, and certainly U.S. citizens are being told this is
- 14:01going to lead to a powerful, even more powerful U.S. economy,
- 14:06higher living standards and so on.
- 14:08But you know, I've been brought up in the world and Chris has
- 14:11said so himself just now. You know, we've lived in a world
- 14:15where free trade was seen as something to be desired and the
- 14:19more we could encourage that and enable that, the better.
- 14:23And indeed, the whole digital services area has been built on
- 14:28that. The Internet has been built on
- 14:30the essence of free trade effectively only mitigated in
- 14:35terms of online safety. A whole of digital markets,
- 14:40competition, antitrust policy has been predicated on that.
- 14:44But you only interfere with digital services where there's
- 14:48clearly an abuse of a dominant position, for instance.
- 14:52So you know the sense are we having to rewire our brains in
- 14:57order to adjust to all this? I don't think it's going to
- 15:00change our attitude to thinking that it's much more desirable to
- 15:05have an open economy, both digital and analog, But, you
- 15:11know, we may have to adjust to a new reality.
- 15:13Now would be an excellent time to subscribe to the CXO Talk
- 15:19newsletter, just go to CXO Talk. But we we have a a number of
- 15:23questions that are coming in. Why don't we jump to some
- 15:25audience questions right now? And this is from this first
- 15:30question is from Arsalan Khan on Twitter.
- 15:35And he says going back to the discussion earlier about the
- 15:41guardrails and risks, he said who he says whoever makes the
- 15:45guardrails becomes the gatekeeper.
- 15:49How do you ensure that there are no unintended biases that creep
- 15:55in as you're undertaking regulatory efforts?
- 15:59That's the same truth when one is regulating or legislating for
- 16:05anything, and it's A at the right point to raise.
- 16:10The reality is, if you understand and are always
- 16:14conscious of the values, the social, the economic context in
- 16:21which you're seeking to bring about regulations and
- 16:25guardrails, and crucially, you have meaningful, sustained
- 16:31public engagement, that gives you the best opportunity of not
- 16:36just bringing to bear that right size regulation, but right size
- 16:40regulation which is really rooting that social and economic
- 16:44context and thus has the ability to thrive as it develops over
- 16:48time. If you like legislators and
- 16:52governments and so on, that's their responsibility to set the
- 16:55rules for the regulators. So it's not as if these guys are
- 16:58sort of freestanding. You know, they're there to put
- 17:02into practice the principles basically, and they should be
- 17:06regular oversight of them. I mean, one of our complaints in
- 17:09the UK is that we're not active enough as legislators in
- 17:13overseeing how our regulators are doing.
- 17:16And you know, as a result, if you're not careful, you get
- 17:19accused of bureaucracy, red tape, regulators blocking and so
- 17:25on. Whereas actually, for the most
- 17:26part, whether it's protecting consumers or farmers or those
- 17:32who are subject to the, the, the weight of technology, if you
- 17:36like it's, it's a, it's meant to be beneficial and it's perfectly
- 17:40possible to make sure that it delivers.
- 17:43And this is from Anthony Scribignano on Twitter.
- 17:48He's been a guest on CXO Talk a number of times.
- 17:50And I, I believe Tim, you know, Anthony, and Anthony says this,
- 17:55he says, please share thoughts on the ongoing difficulty of
- 18:00protecting intellectual property across borders.
- 18:04That is in the category of innovations and artificial
- 18:08intelligence, he says. As an inventor and innovator, it
- 18:13seems antithetical to him that it is so difficult to protect.
- 18:17We could talk about AI, particularly an intellectual
- 18:21property till the cows come home and especially given the
- 18:24difficulty of intellectual property copyright, for
- 18:28instance, used for training large language models being used
- 18:33in different jurisdictions, but being subject to different forms
- 18:37of copyright protection. I mean, the state's actually,
- 18:40it's it's very, very interesting case where you've got the need
- 18:44to register copyright and yet there is the fair use exemption,
- 18:50but that is being tested. And actually for in respect of
- 18:54people like yourself creators, I'm actually quite optimistic
- 18:58that in different jurisdictions, whether it's Northern California
- 19:02or Delaware or wherever it may be, judges are beginning to come
- 19:06to the conclusion that fair use does not cover the use of
- 19:11copyright content for training purposes.
- 19:13And in the UK we have our own debate taking place where our
- 19:18government is proposing to align itself with the EU and that is
- 19:23going to be causing difficulties because they're proposal is to
- 19:27have an exception for the training of large language
- 19:30models that requires creators to opt out and say, sorry, I don't
- 19:36want you to use my material when it's going to be incredibly
- 19:39difficult both technologically and in practical terms for
- 19:44creators to opt out. So it is a really live issue.
- 19:48And the big question is, you know, should those large
- 19:51language developers have a free pass when, you know, creative
- 19:57rights are still very, very important, whether it's authors
- 20:01or musicians or visual artists or film makers?
- 20:06It seems extraordinary that we're in a an era now just by
- 20:11dint of having a new technology that that then necessitates
- 20:17tearing up centuries of well understood IP and copyright
- 20:22jurisprudence. Absolutely extraordinary.
- 20:26As I said at the outset, principles based outcomes
- 20:30focused, inputs understood and to those inputs, inputs
- 20:34understood, respected, remunerated where appropriate
- 20:38and consented. And we're a long way from that
- 20:41in any jurisdiction. I think the Supreme Court
- 20:44decision was extraordinary really to come to that
- 20:48conclusion. As Tim says, there's more hope
- 20:51in some of the state and judicial decisions and others
- 20:56coming down the track in the UK. It's just as live an issue.
- 21:02I published a report just at the beginning of March to try and
- 21:07bring more focus on to a lot of these areas about where AI is
- 21:12currently impacting people's lives and thus the need for
- 21:16cross sector cross cutting AI legislation.
- 21:18I drew out eight realities, 8 billion reasons to regulate, and
- 21:24those realities were people at the sharp end of this.
- 21:26And one of those archetypes was indeed the creative who finds
- 21:30herself or himself with their work being taken with no
- 21:35consent, no respect and no remuneration.
- 21:38It's why in the UKA bunch of our greatest artists, music artists,
- 21:44released an album and it was an album of silence to make that
- 21:47very point. This is where we're heading.
- 21:48If we except, which I don't believe for one instant we
- 21:52should, I don't believe one instant we need to, that this
- 21:56work can just be taken, then that's what we'll be left with.
- 22:00Silence. Where otherwise we'd have those
- 22:03sweet sounds that musicians bring us, blank canvases, a want
- 22:08of artists bringing this stuff to bear.
- 22:10And then the additional kicker on this, it is where you have AI
- 22:15generated content competing with those artists as well.
- 22:19So we need to be conscious of that, you know, double impact on
- 22:23our creative community who we should be rightly standing up
- 22:28for their IP, their copyright, their rights.
- 22:32Tim Oliver P on LinkedIn raises a similar issue.
- 22:37He says there is not just a need for guardrails for AI regulation
- 22:42is definitely required. However, the impact of
- 22:45regulation hits many other areas.
- 22:48He gives the example of AI image manipulation and he says AI in
- 22:53financial services is another key area.
- 22:56And here's the crux of his question, Tim.
- 22:59He says how will regulation help de risk AI LED decision making
- 23:06for customers. He's saying customers, but I
- 23:08think that's really for for all of us in general.
- 23:12Chris and I are both fans of putting forward private members
- 23:16bills, and Chris has had a terrific private members bill on
- 23:21AI regulation in general. I've taken a, a narrower
- 23:26approach, which is about taking decisions by automated algorithm
- 23:32in the public sector. And that is also a very big
- 23:37issue, issue in the private sector.
- 23:38Now the, the increasing use of AI models in the private sector
- 23:45to take automated decisions is going to be a, a bigger and
- 23:50bigger issue as time goes on. And, you know, we have to have
- 23:54a, a risk assessment, an impact assessment to start with.
- 23:58We have to have transparency about the use of these models.
- 24:01We have to have the ability of, you know, the citizen or the
- 24:05consumer to be able to make a complaint, to understand what
- 24:09the decision has been made about them, then make a complaint and
- 24:12get redress. So there's a whole series of
- 24:16steps which, you know, most governments haven't yet really
- 24:19thought about. But, you know, we can't have a
- 24:22situation where decisions are being made by machine, where the
- 24:27citizen, the human doesn't really have any agency at all.
- 24:31And I think many of us are worried that governments, in
- 24:34their enthusiasm, you know, to get a bit of red tape, to be
- 24:38more productive and so on and so forth, are increasingly going to
- 24:41adopt these models. And we're not going to have any
- 24:45ability or any insight into what's happening.
- 24:49And of course, you know, the same must be true of the private
- 24:52sector. They've got the same drive
- 24:54towards productivity themselves. And you know, people now talk
- 24:58about workplace impact assessments to make sure that AI
- 25:03is genuinely going to be used for the benefit of employees in
- 25:07terms of augmenting their working experience rather than
- 25:10simply throwing them out of a job.
- 25:13Chris thoughts on this notion of AI as the decision making
- 25:19overlord. Very much so.
- 25:21And as Tim says, it's why his bill in the in the public sector
- 25:25use of Adm is so important because this covers the whole of
- 25:31society and the economy. And back to the principles,
- 25:35there should always be human in the loop.
- 25:37Potentially you can move to a position of human over the loop.
- 25:42And there should always be the right to a public explanation of
- 25:46any decision. There must always be the right
- 25:48to know that you're subject to an automated decision.
- 25:52We say, for example, you apply for a bank loan and you get
- 25:56turned down and is turned down by an automated decision and you
- 26:01don't even know that that was in the AI was in the mix.
- 26:05That's bad enough. But imagine where you're in that
- 26:09state context where the state, understandably, in exchange for
- 26:15our protection, the state has granted extraordinary powers,
- 26:18unique powers in any open society.
- 26:21So imagine you have your benefits suspended on the back
- 26:26of an automated decision and you don't even know that AI was in
- 26:30the mix. This is happening right now.
- 26:33You don't have to imagine it. It's happening right now.
- 26:36And it takes us right back, Michael, to your initial
- 26:39question, really the sense of, well, why?
- 26:42Why do we need legislation? Why do we need regulation?
- 26:45Because all of these instances be you, a benefit claimant, a
- 26:51job seeker, the creatives, a teacher in education, transplant
- 26:57patients, a voter. This is not hypothetical.
- 27:02This is not something for the future.
- 27:03We're not having to contemplate legislating or regulating for a
- 27:07thing that is coming down the track.
- 27:10This is happening to citizens. This is happening to individuals
- 27:15right now. So this next question is from
- 27:18Greg Walters. And he says, do you feel the UK
- 27:23is better off building a shared AI alliance with the US based on
- 27:29innovation, agile regulation and safety alignment rather than
- 27:35trying to conform with the EU slower, more centralized
- 27:40regulatory framework? So EU versus US in terms of
- 27:46international alliances, who wants to grab that one?
- 27:49I don't think the EU got it completely right on their AI act
- 27:53it it's too disjointed really in terms of how it operates because
- 27:59it's got a sort of almost like a special break out for large
- 28:02language models. Whereas actually the risk
- 28:04framework should apply to every form of AI and and the risk
- 28:09assessment should apply accordingly.
- 28:12But I don't, I think it's slightly a false dichotomy to
- 28:15say, does the EU need to, does the UK need to follow the EU or
- 28:20the, the US? Because currently, I mean, as
- 28:23Chris earlier said, the US has a whole series of states which is
- 28:28beginning to line up different forms of regulation.
- 28:31California, Colorado, you know, the list goes on of states that
- 28:35are beginning to first of all deal with the whole issue of
- 28:38transparency, but also the question of, of what AI does and
- 28:43the risk and this kind of guardrails that are appropriate.
- 28:46And there has been a an awful lot of bills trying to get
- 28:50through Congress over the years as well without success.
- 28:54But there's clearly going to have to be a point where, you
- 28:57know, different in in order to, if you like get it right across
- 29:02the states instead of having it between different states, you're
- 29:05going to have to have some kind of federal legislation.
- 29:08So, and we don't know what that that's going to be.
- 29:11So I think the UK is rightly, you know, following its own
- 29:15path. But what we have to have, we do
- 29:18have to have guardrails. They will have to be risk based.
- 29:22And we do have to adopt pretty much the same standards that
- 29:25NIST is advocating that other standard setters like the
- 29:32International Standards Organization, I triple E the EU,
- 29:36are all espousing. Indeed, the OECD is trying to
- 29:40put it all together and get convergence between the
- 29:43standards. So I actually think that if only
- 29:46we thought rather clearly and maybe adopted what Chris has put
- 29:50forward in his bill, that we'd actually find ourselves in a
- 29:54much better place. I, I, I think this argument
- 29:57about whether or not we're going down AUS or an EU track is not
- 30:02going to be particularly helpful.
- 30:05No, I agree with that. I, I don't think as Tim
- 30:08identifies, there isn't AUS track and there isn't really,
- 30:13though it would appear so there isn't an actual EU track either.
- 30:16When it, when it comes to it in the UK, what do we have?
- 30:20We have an opportunity because of our common law tradition, to
- 30:26reach out to our friends in the EU and the US and right around
- 30:32the common law jurisdictions of the world, to put in place
- 30:36principles based legislation which can develop over time and
- 30:42can deliver for innovator and investor, consumer and creative,
- 30:48and ultimately for citizen. It's why I brought my AI
- 30:52Regulation Bill to bear in November 2023.
- 30:56It was in good shape. I got it through all stages of
- 30:59the House of Lords. It was set to go in the House of
- 31:01Commons until somebody thought it was a good idea to call the
- 31:04general election. Well, we all know how that ended
- 31:07up, but I brought it back just over a month ago and I still
- 31:12believe that we have, if not a unique opportunity in the UK.
- 31:17We have an important opportunity to play amongst friends to bring
- 31:22into bear cross sector legislation which can give a
- 31:27coherence, A clarity and a consistency of approach.
- 31:33Well, that's what you want if you're a citizen or if you're an
- 31:37investor. Totally possible for us to do
- 31:41that. It's unfortunate right now that
- 31:44the UK Government are, shall we say, increasingly reluctant to
- 31:49take such an approach. Let's jump to another question.
- 31:54You can see I really prioritize the questions from the audience
- 31:58over my own. And this is from Funke, Abimbola
- 32:02and Chris. I'll direct this one to you to
- 32:05start. And she says a risk based
- 32:09approach is the best way to go factoring in proportionality.
- 32:13We have numerous examples of such regulations in place and
- 32:17should adopt A similar approach in the UK.
- 32:20And I find this notion of proportionality to be
- 32:24particularly interesting because right now on the world stage
- 32:29stage, there's a lot of discussion of proportionality
- 32:33and it seems the entire definition of proportionality
- 32:36has completely been blown up. So thoughts on this, Chris?
- 32:40My bill to get back to that very much has this conception of risk
- 32:44space, but proportionality running through.
- 32:47Let me give one example to give the sense of why proportionality
- 32:51is such a useful, legal and indeed underpinning that social
- 32:58construct. So I suggest one of the clause
- 33:01in my bill is all organizations developing, deploying or using
- 33:07AI should have an AI responsible officer.
- 33:10Now before anybody thinks overly burdens compliance booth a dead
- 33:16hand on any innovator, any business, any scale, any growth
- 33:20because of the proportionality clause running through the bill,
- 33:23don't think of individual, don't think of group, don't think of
- 33:27team, think about role, think about function.
- 33:30So for those micro businesses just starting off, they will
- 33:34obviously have a proportionately different approach to satisfying
- 33:40that AI responsible officer than say you know, a business of
- 33:4420,000 employees with multiple sites right around the UK,
- 33:49nevermind internationally. So it's an it's an important
- 33:53principle. I think we can bring it to life.
- 33:56There's another principle which appears a lot of other
- 34:00legislation in the UK of reasonableness.
- 34:02And again, people sometimes when you first come to it, struggle
- 34:06with that concept being, well, what's reasonable, what isn't.
- 34:08But that's the joy of drafting in that flexible, agile,
- 34:15developmental way, because quite right, what is considered
- 34:21proportionate today may be very different to what's considered
- 34:25proportion in 10 years time, but quite right too.
- 34:27So thus, the statute or The Reg has the potential to have that
- 34:31developmental nature within it because of the joy of that
- 34:36agility and flexibility of English common law.
- 34:39We have a question from Vanity Osmani, who says how will the
- 34:44need for a sovereign AI affect the possibility of a global
- 34:48agreement on AI regulation? Sovereign AI is actually
- 34:53important because I think we need capacity to create our own
- 34:58models. But I don't think it's going to
- 35:01be very helpful for us simply to expect to be doing large
- 35:04language models because we don't have the AXA compute, we don't
- 35:08have the the chip, the powerful chips.
- 35:12We don't have, except in the case of health data, we don't
- 35:15have the huge data sets that have been accumulated by some of
- 35:19the AI developers. Do think that in terms of open
- 35:25source development, we can really be, you know, pretty high
- 35:32up there in terms of world rankings.
- 35:34And I think sovereign AI to that extent is important.
- 35:37But for my money, more important even than sovereign AI, where of
- 35:42course our major universities have got, you know, great skills
- 35:46and there are spin outs and start-ups and so on, is
- 35:49sovereign cloud. And what we lack is sovereign
- 35:52cloud in the UK. And cloud is the platform for so
- 35:56much that goes on and we haven't got that.
- 36:00And it's dominated by two or three major US big tech
- 36:05companies. And again, that's another
- 36:07example, as we talked about earlier, of American dominance
- 36:11in this area. Now would be an excellent time
- 36:15to subscribe to the CXO Talk newsletter.
- 36:18Just go to CXO Talk and subscribe.
- 36:21We have incredible shows coming up and just this really amazing
- 36:25library of discussions just like this.
- 36:27OK, we have a question from Isabelle Doran and she is CEO of
- 36:32the Association of Photographers.
- 36:35And Chris, she says, do you think that the importance of
- 36:39ensuring UK digital sovereignty, we should be pursuing our own
- 36:45agenda without the US and EU, given the EU appears to be
- 36:50rolling back on guardrails for its citizens, is it possible?
- 36:55Yes, I think there's a real opportunity for the UK to say
- 37:00how we want things to be in an interconnected, in an
- 37:03interoperable, in an internationally connected way.
- 37:07But yes, to say these are the principles on which we base this
- 37:12legislation because these are the shared principles on which
- 37:16we've based our society and our economy.
- 37:19And if we don't, we'll end up potentially on one hand being a
- 37:25rule taker from across the pond. Well, taking something else as
- 37:32well at the same time, neither beneficial for UK creatives.
- 37:37And we have a way to use the technologies to solve for the
- 37:43technologies. If we think about what we can do
- 37:45with a combination, say for example of the metadata
- 37:50watermarking and fingerprint, for example, in combination,
- 37:53there's a real opportunity there to offer real effective
- 37:57sustainable technology proof solutions for our photographers
- 38:02and all of our great creatives. And bear in mind folks, our
- 38:06creative industries in the UK and it's a picture mirrored in
- 38:11many other jurisdictions, are growing at twice the rate of the
- 38:15rest of our economy, 126 billion.
- 38:18It's worth thinking hard about what we do to support all those
- 38:20great creatives who not only do so much for our economy, but do
- 38:26so much, frankly, to nourish our souls and make the life the
- 38:32beautiful thing that it truly can be.
- 38:34I could just add to that what we mustn't do, and I totally agree
- 38:38with you, Chris, what we mustn't do is sacrifice all those hard
- 38:41won creative rights on the altar of some kind of growth strategy.
- 38:46You know this, this. That would be utterly futile at
- 38:50the end of the day, and counterproductive.
- 38:53Let me toss out a question to either one of you, digital
- 38:55assets and tokenization. Why is legal clarity around
- 39:00digital assets, crypto currencies, NFTS, tokenized
- 39:04property important for innovation and market
- 39:07confidence? Either one of you very quickly
- 39:09please. If you bear in mind, whichever
- 39:11stat you take from various consultancies, it's likely that
- 39:15around 80% of all value is going to be exchanged by tokens by
- 39:2220-30. So this is an extraordinary
- 39:25opportunity for any economy, for any jurisdiction.
- 39:30And again, what do we need? That clarity, that certainty,
- 39:34that consistency, What delivers then the confidence for people
- 39:37to invest in, for people to innovate around, for people to
- 39:40develop platforms, for people to develop tokens of themselves.
- 39:44So in brief, in the UK, we're just bringing forth the new
- 39:47property digital assets bill. Tim and I both sat on the
- 39:50special bill committee for that. We've got report stage in about
- 39:533 weeks. What that seeks to do is give
- 39:56that clarity that digital assets can be considered as property.
- 40:02So really delivering to that certainty point, which enables
- 40:05investment, enables innovation and again, to the growth point
- 40:09that Tim rightly raised, the government here and around the
- 40:12world talks about growth. Well, who wouldn't?
- 40:15But if you really want to consider.
- 40:17Where growth is most likely to positively come from in the
- 40:21shortest time in a sustainable, effective and emancipatory
- 40:25place. Where it's going to come from
- 40:26these new technologies and it's going to come from everything
- 40:29around digital assets and tokens.
- 40:31There's a great deal of impatience in the industry about
- 40:36the fact that government really isn't picking up the ball
- 40:39quickly enough. We may be identifying the need
- 40:41to identify exactly what is a digital asset and what is not
- 40:45and the legal definition. But, you know, as as an economy,
- 40:50we're not really moving fast enough on this.
- 40:53Tim, can you give us one sentence on open banking, open
- 40:57finance and initiatives, why they benefit consumers and what
- 41:03should be done there and literally just very, very
- 41:06quickly, please? We've got legislation going
- 41:08through on really trying to roll out the concept of open banking
- 41:13in terms of joining up for the benefit of the consumer.
- 41:18A lot of the data that is held on them by different service
- 41:21providers, which happens in banking.
- 41:23And the idea is to try and make that available to a a wider
- 41:28group of businesses is in terms of the consumer and so on and
- 41:32services provided. And that many people think is a
- 41:38great way forward. I have my I think most
- 41:42legislators do have concerns to make sure that the data is
- 41:47firmly secure. Because, you know, if you
- 41:50imagine the idea of open banking where different financial
- 41:54institutions have your data and it's all designed to be for your
- 41:58benefit in a personalized this kind of way, you know, your four
- 42:01O 1 or whatever it might be, Michael, but you know, this is
- 42:06going to be an important set of services.
- 42:10And you know, This is why we need legislation to make sure
- 42:14that there are ground rules around this, that it is secure
- 42:18and that the consumer really does.
- 42:20And if it comes to the citizen, it it the citizen is also
- 42:23benefited. Tim, just literally one sentence
- 42:27on the importance of open source LLM models in this broad scheme.
- 42:33Very quickly please. I'm really keen on open source
- 42:36models because you know, even if it's off the back of LAMA or
- 42:41DeepSeek or whatever, it gives the smaller developer the chance
- 42:46to use the power of those big models to and using their own
- 42:50particular data set to really move forward in a whole bunch of
- 42:55places, particularly healthcare, where this could be game
- 42:59changing in many ways. Chris, let me jump to you.
- 43:03And this is a question from doctor Carolina Sanchez
- 43:06Hernandez. And she says from a governance
- 43:10and assurance percepts perspective, how do we agree on
- 43:14best practices within this tech race?
- 43:17And if you can answer this literally in one or two
- 43:19sentences, we are going to run out of time.
- 43:22By the approach that we should always take to legislation,
- 43:25considering the context which we're legislating, ensuring
- 43:28meaningful public engagements. Clause 7 in my private members
- 43:34Bill is all about public engagements.
- 43:36The most important clause. Getting that really good.
- 43:39And again, technology enables such an opportunity to transform
- 43:43how we engage with the citizenry, how we engage with
- 43:45the public. So having that, having all of
- 43:48that in place gives us the best opportunity to come up with the
- 43:50right results. But again, underlining on a
- 43:52cracked record, but doing it in the legal code in which we
- 43:56legislate gives the agility, gives a dynamism so we can
- 43:59develop over time. Tim, we have a question
- 44:02specifically directed to you from Bridget E, who says even if
- 44:07the UK regulations are eventually risk spaced risk
- 44:11based like the EU AI Act, will they keep U with rapid changes
- 44:17in technology? Chris and I are great believers
- 44:20in agile regulation. And if it's sufficiently agile
- 44:24and sufficiently principal and outcome based, actually you can,
- 44:29you can do that. You don't have to regulate for
- 44:32every possible form of AI or every, you know, possible neural
- 44:37network or however or algorithm. However you might define it.
- 44:42If you define it by the risk and the outcomes, then I think
- 44:46you're in a much better place. And you know, we're used to to
- 44:49outcome based regulation in the UK, principal based regulation.
- 44:53This is not, you know, a foreign concept for us.
- 44:57And it's worked perfectly well in a number of different areas,
- 45:00particularly financial services. And add to that the question of
- 45:04sandboxing, which Chris and I are both very keen on.
- 45:08That then gives you the ability to safely develop, to innovate
- 45:14under the DDI of the regulator without transgressing the
- 45:19regulations. And it means that you can
- 45:21innovate safely. And then when you've got to your
- 45:24beta point, your product that is fit for purpose, that's the
- 45:29point when you can take off and you know that you're meeting the
- 45:33regulatory requirements. So I I, you know, I'm a fan of
- 45:37proportionate regulation and I do believe that we have the
- 45:42makings of it. Let's move on to Michaela
- 45:45D'mello. And Michaela says she's seeing a
- 45:50fragile relationship between regulation of AI and the
- 45:54baseline data ethics issues that go a step beyond regulation and
- 46:00key and are key for building public trust.
- 46:04What are your thoughts about this relationship?
- 46:08As regulations evolve so quickly, how can we encourage
- 46:12organizations of all sizes, innovators, etcetera, to choose
- 46:16to be more ethical in their use of AI?
- 46:19So Chris, thoughts on the the ethical use of AI and relation
- 46:25to innovation. Please, ethical approach, not
- 46:29only is the right approach, but ultimately it will always make
- 46:33good economic sense because of the certainty, the reliability,
- 46:37the solidity that comes from that approach.
- 46:39So even if you only do it for economic reasons, would you
- 46:43know, people would do it for broader reasons that even if you
- 46:45do it for economic reasons, it's a right approach.
- 46:47Moving beyond that, you can talk about responsible AI, which I
- 46:51think is a very helpful conception.
- 46:54And to the the earlier question, I think that sense of the role
- 46:58of the professional, I think we can have professional standards
- 47:03and the professionalization of the data science, the AI world
- 47:09where people who work incredibly hard come with great concepts,
- 47:13great models in that world, it would really benefit from having
- 47:17some professional standards, some qualifications recognized
- 47:20around that. All of that would help.
- 47:22And to the the Hong Kong point, I think I was down in Hong Kong
- 47:26last spring. It's really interesting what's
- 47:29happening anyway with the Hong Kong Monetary Authority in terms
- 47:33of sandboxing around potential AI models in the financial
- 47:39services arena. They're really using Hong Kong
- 47:44as a Petri dish, not just for AI.
- 47:46But back to our early discussion around crypto and tokenization.
- 47:49It's fascinating what's happening there, but all of it
- 47:52comes back to that sense of another thread running through
- 47:55all of this professionalization, ethical approach, responsible
- 48:00AI, What ultimately we're saying there, we're saying that's about
- 48:05standards. This is absolutely fundamental,
- 48:08a public trust. If you haven't got public trust,
- 48:11and we have had enough difficulty getting public trust
- 48:15for the use of the data sharing and access in the so many
- 48:19different ways. We've got to make sure that we
- 48:22don't make the same mistake with artificial intelligence.
- 48:25But we gain that public trust because that's what gives
- 48:29everyone licensed to use it and to make our lives better.
- 48:33Why, Right, Tim? I think it's why I always say it
- 48:36is the most important clause in my bill because exactly it's the
- 48:40point you made that if, if the public don't trust AI, then
- 48:44they're not going to most likely avail themselves of the benefits
- 48:47and opportunities and simultaneously they're likely to
- 48:50suffer the sharp end and the potential burdens of it.
- 48:53And that would be utterly tragic.
- 48:56But the great news is, if we so choose, it's utterly avoidable.
- 49:00Here's a simple, very important question from Dawn Davidson, who
- 49:06says given the unique insights of employees working directly
- 49:11with AI, do you agree that stronger whistle blower
- 49:15protections are urgently needed in the UK to ensure concerns can
- 49:19be raised safely, particularly particularly as this technology
- 49:24rapidly evolves? Chris, you want to jump into
- 49:27that really quickly, please. Yes, I think whistle blowing has
- 49:31an important role to play, better relations around all of
- 49:34this stuff in the workplace. I think there are some real
- 49:37opportunities with the Employment Rights Bill, which is
- 49:40come to the Lords right now. So I know Tim and I are going to
- 49:43be working a lot on this and the numerous issues around that
- 49:47relationship between AI and the employee and the employer
- 49:50indeed. This is going to be one of the
- 49:52big things in the future as AI increasingly is adopted by
- 49:56employers. It's going to be AI in the
- 49:59workplace and whistle blowing is a crucial part of that.
- 50:02Tim, our salon Khan asks another very important question.
- 50:07He says what about the digital divide?
- 50:10Should countries that are far ahead in AI even care about
- 50:14those who are far behind? Well, I think accessibility is
- 50:19going to be absolutely crucial. And the of course the
- 50:21interesting thing is that different jurisdictions have
- 50:25very different approaches. I mean, you know, we all thought
- 50:29that Africa was, you know, many sub-saharan African countries
- 50:32were behind in technology, but then they leapfrogged us in
- 50:36terms of payment systems, you know, using mobile payment
- 50:40systems well before many of us got there.
- 50:42So there is no Lion rule that says you can't adopt new
- 50:47technology even if you're a developing country and you don't
- 50:52have the actual access to, you know, every form of Internet and
- 50:55so on. You know, mobile communications
- 51:00are, you know, extremely sophisticated in quite a number
- 51:04of countries that you wouldn't think that they had very
- 51:07advanced economies. So.
- 51:09But I do believe that governments have a duty to try
- 51:12and make sure that their citizens have as much access or
- 51:16have a level playing field, if you like, for access to these
- 51:19technologies. And from my money, you know,
- 51:22we've just had a digital inclusion plan published by the
- 51:26UK government, which is broadly on the right track.
- 51:28But I think there should be an absolute right to Internet
- 51:32access, for instance, or at least to 5G.
- 51:35And that needs to be, you know, the bottom line, really, of the
- 51:39digital world. Vincent Cezara simplifies his
- 51:43question for me and he says are people ever going to get an
- 51:47apology for their social media chat, published artistic and
- 51:50creative works that were stolen from them, taken without
- 51:54informing them that it was to be used to train AIS?
- 51:57He wants apologies for his work and other people's work being
- 52:02sucked up into the LLM machine. Sadly, not from all the people
- 52:06who purloined their works. But never mind apology, there
- 52:09should be remuneration for all those works which made these
- 52:14particularly large foundational models.
- 52:16There should be remuneration, there should be respect and
- 52:19there should have been consent. And that needs to be a three
- 52:21pronged approach moving forward. Tim Lewis H says you.
- 52:27I don't know if he means you or the global you meaning both of
- 52:30you said that. You mentioned regulating
- 52:35automated decision making using bank loans as an example.
- 52:39But as you know, this kind of decision making is a constant
- 52:43across AI systems. Isn't there a danger that clever
- 52:47lawyers will find signing loopholes, much like accountants
- 52:50do with tax law? Wouldn't it make more sense to
- 52:53focus on regulating outcomes to ensure they're safe and
- 52:56transparent, rather than trying to constrict the machine?
- 53:00I don't think the two are incompatible, quite honestly.
- 53:04You know, I believe in outcome based regulation and you know,
- 53:07but on the other hand, I'm a lawyer, so you know, I, I, I do
- 53:11believe in the, in the, if you like the legal process.
- 53:14If there are flaws in regulation, then the regulator
- 53:18isn't doing their job or the regulated finance service
- 53:23provider is not doing their job, then we need to make sure we
- 53:26have the remedies. So, but you know, we're back to
- 53:30the proportion at regulation again, basically.
- 53:32And Chris Anthony Scribignano comes back because he wants you
- 53:37to know, he wants to make the point.
- 53:40And he was, he was the chief data scientist at Dunn and
- 53:44Bradstreet. So he's an expert in this.
- 53:45He says malfeasance in the context of cryptocurrency is
- 53:49often not covered by the same protections as Fiat based
- 53:53currency. The best frost fraudsters change
- 53:57their behavior faster than the best regulators.
- 54:00Well, I think there's fraud in crypto as there's fraud in Fiat
- 54:04and Fiat fraud still dwarfs crypto fraud without being
- 54:11unaware of the difficulties with crypto for getting the right
- 54:13size regulatory framework for crypto and tokens is far more
- 54:18interesting. As you know, Anthony, it's a far
- 54:19broader area than simply crypto so-called currencies.
- 54:23Getting the right regulatory framework in place, getting more
- 54:26public understanding of these technologies gives us the best
- 54:29chance to get the protections in place.
- 54:31And back to the early question Tim answered one word answers
- 54:33that labelling. So if somebody is going to be
- 54:37subject to AI in the machine or an automated decision, if it's
- 54:40labeled, they'll know that's the case and that will give them the
- 54:43best chance to decide whether they want that or not.
- 54:45Tim, to finish up, let me direct this first to you and then I'll
- 54:49direct the same question to Chris.
- 54:52What advice, Tim, do you have for policy makers regarding the
- 54:57building of public trust in AI, digital assets, and open data?
- 55:02Introduce proportionate regulation that basically make
- 55:07sure that the the high risk AI is regulated.
- 55:12That's be my advice, and that would create public trust and
- 55:17give us the license to keep innovating.
- 55:20Chris, same question. Advice for policy makers to
- 55:23build public trust in regards to technology and regulation.
- 55:27Engage, engage, engage. We have what we need to make a
- 55:31success of this because we understand critical thinking,
- 55:35values, ethics, responsible approaches, economics,
- 55:39philosophy. We need to be human.
- 55:42These are tools. They're incredibly powerful
- 55:44tools, but they're tools in our human hands.
- 55:47We decide. We determine ultimately our
- 55:50data, our decisions, our human LED digital futures.
- 55:56And with that, a huge thank you to Lord Chris Holmes and Lord
- 56:01Tim Clement Jones. Thank you both so much for being
- 56:04here. I'm grateful to you both.
- 56:06Thank you. Thank you, folks.
- 56:09Thank you for your questions and for watching and I hope Tim and
- 56:15Chris, I hope you'll come back. I, I, I feel like you're both so
- 56:19smart and the questions that I were, I was asking you don't,
- 56:25didn't challenge you enough. And so please come back again
- 56:29and let's do this one more time. Brilliant questions.
- 56:33Everybody, thank you for watching.
- 56:35We'll see you next time. We have amazing shows.
- 56:37Check out cxotalk.com. Be sure to subscribe to our
- 56:41newsletter because we want you as part of our community.
- 56:45Thank you so much everybody, and I hope you have a great day.