Episode Transcript
[00:00:00] Sam.
[00:00:31] Good evening and welcome to AI Today. I'm Dr. Alan Badot, and tonight I'm coming to you live from London.
[00:00:39] I want to begin with a word of thanks, though. I'm staying at the beautiful Four Seasons Hotel, London Park Lane. And there's one thing that AI cannot replace, and that is hospitality. And, you know, at this hotel, there's really nothing short of, you know, everybody going out of their way. And the service is always extraordinary to the entire team, you know, again, thank you for being such a gracious and wonderful host for us.
[00:01:11] Now, I want to say, too, what a place, though, to think about the future from.
[00:01:18] I'm looking at, you know, the, you know, Hyde park over on the right hand side, and, you know, it stretches really, you know, green. There's just about every place that you can see, and it's absolutely beautiful. And, you know, of course, Park Lane threading quietly along the edges and the. Some of the old trees catching the, you know, soft light that we're seeing and, you know, riders out and on the path and, you know, it's just absolutely, really beautiful. And centuries of order, though, if you think about it, really, you know, very carefully kept.
[00:01:56] Now, what makes this a fitting place tonight, though, is to think about some of the differences between what the EU is doing as well as the UK versus what we're doing in the United States. And in so many ways, you know, the world's, you know, continue to look at different ways that they can do things, and, you know, how they are each trying to keep order when it comes to AI and right now it's really trying to bring that instinct that years of history, that years of order into probably the most disorderly technology of our age.
[00:02:44] So here's our subject tonight. Two continents separated by a narrow channel of water, being the United Kingdom and the eu really two opposite theories in how to govern AI, the rulebook and the race, and a single city really caught maybe deliberately in the middle of both.
[00:03:12] So that's where we're going to begin tonight. And I think it's going to be interesting for folks to see some of the real differences in how everybody is approaching it.
[00:03:25] So let's start with Europe, because really, Europe has been some of the first movers on how to approach AI So the European Union, they wrote the world's first really comprehensive law for artificial intelligence, and that was the EU AI Act.
[00:03:43] And its core idea is, frankly, it's, you know, it's. It's really an elegant one actually, is don't regulate the technology, regulate the Risk sort every AI system, you know, by what it's actually doing.
[00:04:01] A spam filter is one thing, a system that decides who gets hired, who gets a loan, who gets policed. You know, that's in some, that's something entirely different.
[00:04:12] The higher the stakes, the heavier the obligations on the people who built it and also the people that deploy it.
[00:04:21] Some uses are just, quite honestly, they're simply banned now. It happens in waves though, right? Just like everything else, you know, there are earliest prohibitions are really already enforced. The rules for general purpose models are being followed and the largest wave was set to break, you know, really this summer around 2 August, when the obligations for high risk systems and the new transparency rules were to take hold and enforcement with real financial penalties were supposed to begin.
[00:05:02] And then this spring, Europe did something quite revealing.
[00:05:09] In May, negotiators reached a provisional deal on what's being called the digital omnibus. And that's really the first amendment to the act since it passed. And the headline, you know, change is really this, that those heaviest high risk obligations, the ones that were set to, you know, hit in in August, you know, by reports, are now into late 2027.
[00:05:40] And consider what that tells you. So the continent that wrote the rule book, the one that spent years urging the rest of the world to get ahead of this technology, looked at its own deadline and looked at the race it's in, and I think they blinked. They didn't abandon that, of course, but they didn't, you know, not it wasn't repealed, but it was a little bit softened and of course delayed.
[00:06:10] So the rule book is real, but the rule book, as it turns out, is more negotiable than many people believe.
[00:06:19] Now as we look at that, as we dive into that, then there are some important nuances associated with that, right?
[00:06:30] There are numerous American companies, many of the big AI providers that are, are driving the market, driving the industry, are working on trying to extend some of those rules.
[00:06:47] It's in their financial interest, it's in their best interest to be able to do that.
[00:06:52] Now that's not a bad thing.
[00:06:56] But with the harshness in some cases of how many folks feel that, including the U.S. administration, the government administration, how harsh some of these U.S. companies have been treated.
[00:07:14] Some have even said it would be better off if they would just leave Europe.
[00:07:19] Now, I don't believe that at all. And I think that many of my European colleagues would agree with that assessment that it's not going to ever be better for companies to just up and leave. But those are the stakes.
[00:07:34] Those are the, what I would say, billions and billions of dollars that are really at risk in, you know, economics, both for the users as well as the distributors of some of these models, as well as really the technology itself and how it can be applied to different types of systems. And whether those are defense systems, whether those are, you know, mission critical systems, whether they're power systems. The impacts are really going to be felt across Europe as a whole.
[00:08:13] You can't really single out one country on the European continent versus another. You know, some are, some are trying to, you know, relax some of the things, some of, you know, some of the regulations, others are trying to, to tighten them from a privacy perspective. You know, so there is still some internal debate going on. But overall that rule book really was supposed to set the foundation and the behavior of companies that were going to participate in that market.
[00:08:44] Now, as we have all seen, again, you know, when these big companies get involved in whether they're lawsuits or whether they are public debates, you know, oftentimes there of course are two sides of the story and we know in this case there are.
[00:09:06] However, we also know that that rule book was supposed to help guide some of the other activities that many of us hoped we would eventually see in the United States. And we're starting to see some of those. And I know, you know, our friends in, in the UK are also starting to look at them and you know, it's starting to impact on how they are deploying certain things and how they are using them. But in the, in the grand scheme of things, as, again, as it turns out, you know, when you have a rule book, it's real, it's been published, it's supposed to be law.
[00:09:49] But you know, everything is negotiable when it becomes an AI discussion because the impacts are so widespreading, so really driven, you know, by both the user community and how they can benefit as well as the government communities and how those can, can benefit, I guess everything is negotiable.
[00:10:17] So we're going to deep dive into this topic, stay around, we're going to, we're going to talk about Europe a little bit more, we're going to talk more about the UK and we're going to dive into that and we're going to talk about London and how its impact is being felt really around the world from an AI perspective. So stay with us. We'll be back after a few messages from our sponsors.
[00:11:12] Welcome to AI Today, Dr.
[00:11:15] I'm Dr. Alan Badot and now I want you to cross that Narrow channel between Europe and the UK and watch the entire philosophy invert.
[00:11:36] So the UK has sort of made a, you know, the opposite wager if you think about it.
[00:11:44] And really to understand it, you first have to understand what Britain chose not to do.
[00:11:53] So for years an AI bill was expected here as well. It was promised, drafted, anticipated in speech after speech. And the year, you know, year after year, it just didn't arrive.
[00:12:10] Now that's not a failure of nerve, right? It was really a decision because somewhere along the way the government concluded that the wiser path was, you know, not one, one grand sweeping law for artificial intelligence, but something really quieter and in its own way more radical.
[00:12:40] So the choice was this.
[00:12:44] Don't build a brand new rule book from scratch.
[00:12:48] Take regulators. You already have the Data Protection authority, the competition watchdog, the financial conduct regulator, the medicines regulator, the online safety body.
[00:13:05] And ask each of them to apply the powers that they already have to AI and how it's operating in their own domain, regulate at the point of use by the experts who already understand that use, sector by sector, case by case. And existing laws, you know, pointed at the new machine.
[00:13:34] So a medical AI answers to the medical regulator. A hiring tool answers to data protection and equality laws.
[00:13:44] A trading model answers to the financial conduct of, you know, you know, the financial conduct authority.
[00:13:54] There's no simple or single AI ministry in London writing one rule to cover all of it.
[00:14:04] And there it is. Instead of a web of referees already on the field, each one is told to watch its own corner.
[00:14:18] And the posture built on top of that choice is unmistakably pro growth.
[00:14:27] So Britain published an AI Opportunities plan and reports it already has met most of its early commitments it set aside. The idea of an AI bill is it's immediate vehicle in favor of a blueprint for regulation.
[00:14:48] And the centerpiece of that blueprint is, you know, genuinely novel.
[00:14:53] An AI growth lab, a national program of licensed sandboxes where specific rules can be temporarily or deliberately relaxed.
[00:15:07] That allows companies to test real systems in real conditions under close supervision in healthcare and transport and professional services, in advanced manufacturing. So test it first, watch what risk actually turns out to be, then write a rule if you still need one at all. Now that's novel, believe it or not.
[00:15:35] And the government has, has signaled, you know, it intends to go further still and that proposing, you know, to formally instruct its own regulators to put growth first, to be in, you know, in its own words really that I've been able to find less risk averse and you know, really to Report each year on how they are enabling AI rather than only restraining it.
[00:16:05] And you have heard me talk in the past about, you know, some of the work that we're doing with folks in the United Kingdom and what we're trying to build, you know, with many different, you know, industry verticals. Now that's really one of the reasons why we've even been over here talking to folks.
[00:16:30] So if you think about it, Brussels, they're writing a ceiling and Westminster is trying to build a Runway.
[00:16:40] So the very same instinct though, that runs through how Great Britain is carrying itself abroad, backing its own companies, especially in the hardware that underpins all of this working. And with other so called, I would say, middle power nations to help break the standards where AI gets deployed, you know, around the world.
[00:17:05] Now, I won't sell you only the bright side of this though, right? Because, you know, is a bet and, you know, real bets carry real risks.
[00:17:19] The fight over here is around, you know, AI and the, of course, the creative copyright, something many of us are all familiar with.
[00:17:30] And whether these models can train on work of British writers, artists, musicians without asking permission. Sound familiar?
[00:17:42] It's bitterly unresolved. And it's precisely the place where this light touch philosophy is really under the most strain, just like our laws.
[00:17:55] And it's going to be interesting to see, you know, who, who gets this figure out first.
[00:18:01] You know, the competition regulator has already proven the most assertive of the bunch over here and, you know, the critics make a point.
[00:18:12] I think it's worth saying plainly that, you know, lights touch today can quietly become deferred reckoning tomorrow.
[00:18:22] A framework built for speed is always one serious scandal away from a hard, you know, written rule overnight in anger by people who have stopped trusting the industry completely and believe that now it's time to police it a little bit more.
[00:18:41] So if you set aside, you know, the two approaches and you know, you have an independent, you know, person, just take a look at them side by side, you can see the clear differences.
[00:18:56] Europe wrote the rule and now it's loosening it.
[00:19:01] United Kingdom and Great Britain, they refuse to write the rule and is daring the future to prove it wrong.
[00:19:11] So one bet has placed on really a prescription, the other on a judgment.
[00:19:20] Two genuine theories of how you govern a technology moving faster than the people charged with governing it, running at full speed on one, you know, small continent at the very same moment with the results still entirely unwritten.
[00:19:44] And in many ways, the results over here could significantly impact what happens over in the United States. And the things that we see.
[00:19:56] We've already seen states start to try to take the lead and we've seen President Trump write certain executive orders and looking at what these government regulators and bodies were going to be able to do.
[00:20:14] And, you know, I wish I could say that there was one global way that we would be able to manage and do this, but we have seen, as the world has gotten significantly smaller, the impacts of AI, whether they're developed in the U.S. the UK or the EU has ramifications that none of us can predict.
[00:20:43] None of us know exactly how they're going to impact our daily lives as these start to take hold. And, you know, more importantly, how are our governments going to react, work together or really start to use this as a next, I guess, technology war?
[00:21:10] We'll see what happens.
[00:21:12] But, you know, which one is going to work first? Which one is going to work better?
[00:21:18] It's still undetermined. And I don't think we're going to see really, you know, an end to this or a real solution until probably 2028, because I believe they're going to continue to move those regulations in the EU from 2027 to 2028. And I think how the UK is working it, I think the results of theirs are going to start to come to fruition around the same time. So it's going to be interesting and it's a topic that we're not definitely going to want to dive into a little bit more. So stay with us. We'll be back after a few messages from our sponsors.
[00:22:17] Foreign.
[00:22:33] Welcome back to AI Today. I'm your host, Dr. Alan Badot and I want to bring the focus back to really this city.
[00:22:46] And everybody knows how much I love London. I've been here many times over the last, you know, 20 plus years. It's a great city, a lot of great people working in the city, very nice, very, you know, innovation minded and really willing to take some of those first steps that, you know, others may not be willing to take. And it really is what I would say, separate from many, many others. And, and it really does have a special place in my heart over here. And I want you to think, though, you know, think about it from this perspective with AI because in essence, whatever the rule book, quote unquote, decides whether it's in Brussels or in London, you know, the race is going to have an address and a remarkable amount of AI is already here and it's all over the place, right? And London really is the commercial heart of European AI. There's, there's no, you know, two ways about it.
[00:24:09] And quite honestly, I don't think it's really even a close contest.
[00:24:14] The corridor that runs from London up through Cambridge and across to Oxford draws something close to nine of every ten pounds invested in AI.
[00:24:34] Well, at least in, you know, invested in AI in the, in the United Kingdom.
[00:24:38] And, you know, Cambridge brings really that research depth and Oxford, you know, spins out the scientific part of it, right? And then in London, you know, is really where all of it comes to raise money, to hire and quite honestly, to scale.
[00:24:58] And that includes U.S. companies that are doing business over here as well.
[00:25:04] You know, the names are familiar to anyone who follows this field, right? You've got DeepMind, whose work had already, you know, as far as a research perspective, provided a Nobel Prize, right? And then you've got Isomorphic Labs, you know, turning that same science on, you know, onto the, you know, discovery of medicines and new medicines. And then you've got Wave teaching cars to, to drive themselves, right? And Sunita, you know, who's generating video from text. I mean, the list goes on and on. And it's a deep, renewable bench of talent and it's being fed by some of the finest universities on Earth.
[00:25:51] And of course, sitting conveniently right beside one of the world's great financial capitals, it's an industry that's really, you know, impatient to point at, you know, fraud at payments at risk, at trading. You know, they're really driving the innovation in some of these sectors and it's fantastic to work with them too, because, you know, the really, the camaraderie, the team building, those things are really natural to some folks over here, where in others not quite so much.
[00:26:35] Now, you can measure, though, the true gravity of a place by who is paying to be inside of it.
[00:26:46] You know, I, I mentioned some of the American companies, but really the labs, the American labs have planted their flags within walking distance of one another, right? And really a lot of them are clustered around King's Cross.
[00:27:02] You know, one of them is, is actually describing its London operation as the single most important hub it runs anywhere outside the us and quite honestly, by reports, they are, they're offering individual London engineers pay packages north of £600,000 a year. Those are some of the statistics I've heard in the last few days.
[00:27:27] Now sit with that number for a moment.
[00:27:32] That is not even the price of an office.
[00:27:36] It's not a perk or a title or a signing bonus even.
[00:27:41] This is the open market price of one human mind that knows how to build these systems. And so when a single mind commands that much, you're no longer looking at a job market. You are looking at an arms race for cognition, quite honestly.
[00:28:02] And it's being fought, really street by street. And in London, more so than really any other city in the world, that concentration is so sought after that there's no other way to describe it.
[00:28:19] Now, beneath the talent sits really the foundation that makes the talent matter.
[00:28:29] Because, you know, a brilliant mind with no machine to run on is just a frustrated one. We've all seen that. We've talked about that, right? That's what leads to, you know, AI winters and things like that.
[00:28:42] So watch where the real money is going.
[00:28:46] A new sovereign AI effort, you know, backed by hundreds of millions of pounds, stood up just this spring, designated growth zones, pulling in tens of billions of dollars in private commitment.
[00:29:02] That is amazing. The largest chip maker on the planet is pledging billions of dollars of pounds into the British ecosystem.
[00:29:14] The largest cloud providers are pledging to, you know, build this country's most powerful supercomputers on British soil.
[00:29:25] And a national ambition to really multiply the nation's compute many times over by the end of the decade is driving this desire and this innovation.
[00:29:37] And so that also tends to say, you know, to lead to new visa routes that are open to pull talent from everywhere else.
[00:29:49] This is a country trying very deliberately to own the means of intelligence rather than rent it.
[00:29:59] And that's amazing. It's very, very much like how the US is trying to approach it, maybe though, with a little bit more finesse, a little bit more oversight, a little bit more regulation, but doing it the way that many of us have been advocating for, for years.
[00:30:22] And quite honestly, this is the theme that this program always tries to return to again and again. And it's now playing out at a scale of an entire nation.
[00:30:37] Sovereignty in the age of AI is not a slogan, and it's not a flag on a building.
[00:30:47] It's really three concrete things you've got compute, which are, you know, of course, the machines.
[00:30:57] It's the talent, which, again, the minds, and it's the standards, which is really, in this case, a seat at the table where the rules of deployments and operations actually get written.
[00:31:18] A nation that holds all three owns its own future.
[00:31:27] A nation that holds none of them.
[00:31:32] They're going to rent it, and they're going to be renting it by the hour, because guess what? They're going to be paying for it by the hour.
[00:31:40] And from whoever is providing those services now, London is making its bid to be the landlord of this new age, in this new world, and they don't want to be a tenant.
[00:31:57] The bet is being placed in real money, in real buildings, and it's happening right across the street.
[00:32:09] And it's that buy in, it's that, you know, that triangle between London and Cambridge and Oxford that that partnership is really tough taking hold.
[00:32:23] You've got the academics who are working very diligently on the science piece, but also on the risk, the regulations to help the civilian counterparts understand those, understand the ramifications of some of these systems. And then you've got the civilian section of the government that's really working to help industry understand and help industry shape and drive the things that they're trying to build and understand some of those consequences.
[00:33:01] And then you have industry that is consistently reaching back into an educational environment that has been primed for years to allow their students to understand not only the AI and the computer science piece, but the fundamentals of math, fundamentals of science and really any other part of that AI ecosystem, including economics.
[00:33:33] So ask yourself, where else in the world is this actually being replicated in such a high concentration?
[00:33:45] I think you're going to be hard pressed to find anywhere other than London where all of that is happening at scale and at a regulation that everybody can buy into. And everybody is working very hard to, you know, make sure that it works out the right way and they're doing it the right way.
[00:34:08] So think about that. And I'm not advocating anybody to, not necessarily move to London, of course, but as AI continues to expand and the world gets a little bit smaller, London might not be a bad place to go look for an AI job.
[00:34:25] So I'm going to leave you with that thought. So stay with us. We'll be back after a few messages from our sponsors and we're going to close out the show and we're going to try to bring all this together and really help you understand the impacts that it could have on the United States. So stay with us. We'll be right back. SAM.
[00:35:07] Foreign.
[00:35:17] So I want everyone to take a step back, back away from the maps, back away from the money, back away from the deadlines and the deals and the extensions and yada, yada, yada, and ask the question sitting underneath all of it.
[00:35:39] What is any of this actually for?
[00:35:43] So strip away the acronyms and strip away the agencies and Europe and, you know, the UK are reaching for the very same thing from opposite ends of the same table.
[00:36:05] Call it the accountability layer, the precise point at which a Human being or an institution with a name and an address has to stand up and answer for what a machine decided.
[00:36:27] We talk about that all the time on this show and we're looking at it from a whole bunch of different perspectives.
[00:36:34] But I want to make sure that everybody's informed because maybe one of these are going to help you and as you're looking at it, and how you're going to use AI or how you're going to develop your own AI, you know, to support whatever journey that you're on.
[00:36:49] So Europe tried to write that point directly into law, line by line, risk tier by risk tier. It's actually a fascinating read. I would encourage everybody to go out and take a look at it. You know, conformity assessments, technical documentation, human oversight, logging, transparency, you know, an entire architecture built so that when a high stake system acts, there is always a named party who is responsible for it.
[00:37:21] Now in the uk they looked at it from, you know, that they looked at the very same problem and made an opposite choice, betting that its seasoned regulators, its courts, its markets and the laws it already has on the books will locate that responsible party faster and provide more flexibility.
[00:37:53] And they're hoping that they'd be able to do so faster than the years it takes to draft a statute.
[00:38:05] One of them wrote down the right answer, the other is trusting that they already have the answer.
[00:38:18] But here's what I want you to notice and it's really the whole point of, of this discussion.
[00:38:26] Neither one of you know, whether it's the EU or the UK has escaped the question, right? Europe can defer its deadline to 2027.
[00:38:41] The UK can, you know, decline to pass a law at all, right?
[00:38:47] But the question still is sitting there, that giant elephant in the room probably generated by AI.
[00:39:01] Being patient, right, unmoved.
[00:39:05] We're trying to answer the very same thing.
[00:39:09] And it's waiting for an answer. Because the question was never really the property of Parliament or a commission or a jurisdiction or a country, you know, because the rule is movable.
[00:39:25] The responsibility though, is not.
[00:39:30] And that is exactly what all of this asks of you, of every leader watching this.
[00:39:41] Wherever you happen to operate, you can run the very same model in Brussels, you can run it in London, you can run it, you know, across the pond in the US and you will answer, you know, will be in three different rule books for the privilege of being able to do that.
[00:40:08] The compliance forms are going to differ, the deadlines are going to differ, the penalties are going to differ, but you cannot file the responsibility under someone else's. Flag.
[00:40:22] There is no jurisdiction generous enough and no border low enough to carry it across on your behalf.
[00:40:33] So don't spend your energy shopping the globe for the lightest rule book.
[00:40:42] I know some companies that have, have said that they were doing that they were going to set up here versus here versus here because it would be easier to do this or that.
[00:40:54] I think that's a, that's a wasted effort if you spend it building so that you can answer really anywhere to anyone, on any day they come asking.
[00:41:09] That's the important piece.
[00:41:12] Own the reasoning that your systems run on, govern it.
[00:41:18] Keep a human hand on the decisions that carry real weight. Like we always talk about that human in the loop factor because honestly, regulation, you know, in the end is only society's clumsy attempt to find the name on the accountability line. That's it. That's all it is. And the organizations that thrive across every one of these regimes, they're going to be the, the ones who already knew, long before the auditor knocked, that the name was theirs.
[00:41:58] This is the line this program is not going to let go of.
[00:42:04] The reasoning can belong to the machines.
[00:42:09] The accountability belongs to us.
[00:42:13] A rule book can tell you what, what's permitted in a, in a certain place, that's fine.
[00:42:21] It can never tell you though, who answers when it matters, whether it's in Brussels, whether it's in London or anywhere else that you choose to stand, because it's always going to be yours.
[00:42:39] You run the model, it's your responsibility.
[00:42:42] You build the model, it's your responsibility.
[00:42:47] They have to go hand in hand because the reality is, is that the rules are going to, simply put, keep diverging.
[00:43:01] Every government that comes up with these types of rules, they don't tend to back off on the direction that they go. Oftentimes they continue to push forward.
[00:43:10] And that very quickly leads to a divergence that can be felt across the globe.
[00:43:17] And in this case, you've got a ceiling in one capital, you've got a Runway in another capital, and then you have really a fresh amendment, what seems like every spring or an executive order.
[00:43:36] The responsibility doesn't behave like that because we know how quickly AI is evolving. That responsibility in some cases becomes more and more and more important every single day that we're running these models, every single day that we're building these models. And I take that to heart, that's why we're building Janus and Atlas and all of our other tools the way that we are and why we're putting such effort into that transparency piece. That accountability piece, it crosses every single border intact and it ends exactly where it always has.
[00:44:15] And that is with the user, the developer, whoever.
[00:44:22] But that's the driving factor.
[00:44:25] And so, you know, from a quiet window, you know, really overlooking London's parks.
[00:44:36] And again, with my thanks once more to the Four Seasons here at Park Lane for their gracious welcome and hosting, you know, I'm Dr. Alan Badeau. This has been AI today, and I want you to continue to compare everything that's going on, see how you're doing it, learn from everybody else, try to do it a little bit better, mind that gap between what the rules allow and what you can answer for. And if you operate AI based on that premise, then I have no doubt you are going to be successful and you're going to do it the right way.
[00:45:26] So I'm excited for everybody to see our show next week.
[00:45:30] I look forward to seeing you. Thanks. And we'll see you next week.
