Anthropic is at the most important moment in its history and has a warning: we must lift the AI ​​accelerator

Anthropic and OpenAI are competing in two parallel races. On the one hand, the race to have the best model that continues to push cutting-edge AI. On the other hand, going public to become a public company. The winner of the first is debatable, but in the second competition it is Anthropic that has the advantage.

This Monday he confirmed that he had registered his application for the long-awaited IPO, which could become the largest operation of this type in history which, as my colleague Javier Pastor commented, is reminiscent of Netscape going public and marking the beginning of the dotcom fever. And, precisely in this scenario, the company has launched an interesting notice: a global pause must be made in the development of frontier AI.

And it makes all the sense in the world at the same time that it is something that cannot happen taking into account the situation between the two powers that are in the current technological war.

Taking your foot off the accelerator vs the technology race

Anthropic has been using something curious for some time: the fear tactic. It is something that their main rival has thrown in their face (to automatically do exactly the same), but the truth is that Dario Amodei's people from time to time drop some 'bombshell' about how humans run the risk of losing control over this technology, with Amodei ensuring that there are possibilities that "things will go very, very bad."

In a recent post, the company behind Claude noted that they would be happy to take their foot off the gas on cutting-edge AI development only, and this is important, if they were confident that others would do the same. Again, they compare it to war development, as if it were an explosion in the number of nuclear weapons, noting that they believe it would be a positive pause when comparing the increase in increasingly capable AIs with an “arms control problem.”

When they talk about cutting-edge AI they refer to various models, but they focus on those that develop themselves. In a document shared this Thursday, two heavyweights from the company's research team stated that AI technology is approaching the point where systems can self-develop. That is to say: AI writing itself to improve.

But, if they are so close (or so they say, since they have been talking about the same thing for months), why stop? Well, not for the technology sector, but for everything else. Anthropic believes that a slowdown would allow other actors - countries and regulators, for example - to catch up with technology.

They comment that “it would be good for the world to have the option to temporarily slow down or pause the development of frontier AI to allow social structures and research to keep up.” The problem is coordination between countries because there is no trust at all that countries tell each other the truth.

As we read in The Telegraph, the company noted that “a significant slowdown or pause would require multiple laboratories with sufficient resources at or near the border AI agreeing to stop under the same conditions”, but the problem is that “it would also require each one to be able to verify what the others are doing and make sure that they have really stopped.”

This has obviously brought criticism, such as that Anthropic is overestimating its capabilities (because of the AI ​​that writes itself) or exaggerating the abilities of AI so that regulations are introduced that 'harm' its competitors.

In any case, this stoppage is not going to occur in a context in which the United States and China cannot allow themselves to take their foot off the accelerator as they find themselves in a contest in which China is clear that it wants to be the leading technological power in the short term and the United States, evidently, does not want to allow it.