00:00So the top line from that essay, that more than 3,000-word essay from the founder and CEO of
00:04Anthropic, as you said,
00:05was that this is a company that now wants to slow the pace of its frontier models, the developments of
00:11its frontier models.
00:12And he outlined three specific courses of action that he and the team at Anthropic would like to happen.
00:19There's one course of action that they are going to do, they say, which is having outside experts embedded into
00:25the company,
00:26given, they say, employee-level access to the work that they're doing to verify independently,
00:31they say, verify independently the work that they're doing and some of the risks that they're wrestling with.
00:36The other part of it is calling for the frontier labs in the U.S. to coordinate better.
00:41We're seeing some signals that that may happen, given what Sam Altman, the CEO, of course, of OpenAI, has been
00:47saying,
00:47broadly aligning with Dario. And you've heard similar supportive measures from Elon Musk as well.
00:54The third part is global coordination. And that seems to be the much more distant prospect.
01:01There's also a request from Amodi, once again, for the U.S. to regulate.
01:06And even though this is being picked up by, in some measures, by both Democrats and Republicans,
01:11it's becoming something of a bipartisan issue. There's probably not the momentum to get policy or get legislation on the
01:18books in the short term.
01:19So this seems like the more practical place in terms of where this lands is potentially getting those outside observers
01:26and maybe getting some of the big labs working closely and slowing the pace of the frontier, the developments of
01:32the frontier.
01:32And again, Sam Altman telling Forbes that there's going to be no IPO this year.
01:36And part of that is down to their desire to really focus on safety.
01:40And what about the Chinese AI companies and all of this as well?
01:45China's actually criticised these warnings about the risks of AI as well, calling it fear-mongering.
01:51Where does this leave the Chinese competitors in that rivalry?
01:54It's interesting. You've heard from two senior Chinese officials, one in the cybersecurity space,
02:00an article written there talking about the potential risks, but really with the lens focused on external actors,
02:06so foreign adversaries, so notably the US, quite frankly, and the risks that their technology poses
02:10in terms of potentially destabilising the Communist Party and its institutions of governance.
02:16And then you had officials coming out more recently in the last couple of hours saying that the risks around
02:23AI,
02:24with a nod to what we've been hearing from Dario and Sam Altman, OpenAI, is being overhyped.
02:29I think one way to read that is, let's remember that China is very much focused on open source
02:33or open weight models so that you can adjust and change.
02:38There is some kind of insight you can see into that.
02:40These are also models that are trained on chipsets that are generations behind the chipsets
02:45that are being used in the US, and yet they have advanced very, very quickly on some measures.
02:50They are just six months behind the US on this.
02:53There's a lot of distilling, which is basically where they use their own models,
02:56Chinese models, to kind of learn from, illicitly learn from,
03:00from Anthropik and DeepMind and Google and OpenAI.
03:04So there's that happening as well.
03:06This is, China would not be relaxed with the idea that you have rogue models
03:11and rogue cyber threats caused by frontier models,
03:15but they have a very different focus in terms of how they are building out AI,
03:19and they see it as equally as importantly as the Trump administration
03:23in terms of their economies and importantly for China as well, their military.
03:27Yeah, and in all of this, you know, you have to think also about there's investors,
03:34there's the US government and its wishes, and also I'll put in there the public,
03:41you know, the idea of consent of workers.
03:45What do they make of which jobs should stay and go of, of the kind of,
03:51the sucking in of knowledge that these AI apps have, have done globally, you know,
03:58the kind of knowledge of the whole of humankind that some would say.
04:02So there's all of those actors and stakeholders that you have to think about,
04:06but, but mainly on the US government, you know, does this mean a clash with Trump?
04:11There is a tension there because whatever, whatever the granularities of this top line is,
04:18whether or not we get legislation, there is a political pushback story now in the US.
04:23There are something like 70 different data centres that are stalled and upheld in the US
04:28because of local opposition.
04:30And that was prior to the, by the way, AI has a 10% chance of wiping us all out.
04:34And I'm paraphrasing some of the, some of the, some of the ex-researchers
04:37and current researchers who came out last week.
04:39So this is becoming increasingly a political issue.
04:43This is becoming increasingly an issue that is being felt by legislators on the doorstep
04:48that their constituents are concerned about in the US.
04:50You'd expect similar dynamics in Europe and in the UK as well.
04:55And so how politicians, how leaders grapple with that,
04:59Andy Burnham is suggesting that he doesn't want to see a moratorium
05:02and data centre build out here in the UK.
05:04How you make the case for good AI is going to be increasingly in,
05:08in the spotlight, I think.
05:09And, and we'll be incumbent on politicians to try to make that case
05:12if that's what they, they believe is important.
Comments