Thoughts about AI

Recently, Dario Amodei, CEO of Anthropic, published an essay about pacing frontier AI Development. This is all fine of course, but when all the other CEOs agreed with him immediately, which made me wonder:

Why all these people who sometimes even refuse to shake hands, came to an aggreement all of a sudden?

The conversation seems to revolve around self-improving AI and the question is that what if these models start to improve themselves? Like a 3D printer that prints new parts to improve itself automatically. Although sounds far-fetched to someone working with these models daily, one couldn’t easily eliminate the probability of it being true.

On the other hand, the marketing departments of frontier AI companies don’t stop talking about how their new model is “the best thing ever since sliced bread”. It seems to me that we are hearing about a new model every week at this point which begs the question:

Why release newer, more capable models every day, if they are too dangerous?

I don’t think releasing a new model is automatically the opposite of pacing AI development. It’s about giving people something real to evaluate: how it behaves, what it can do, and where its limits are. But that only helps if evaluation and safeguards keep up with the release.

All of this doesn’t change the fact that I feel tired of speaking and hearing about AI all the time. Some people praise it, some people hate it and I’m just sitting here hoping, someday this whole ruse will end one way or the other.

Stay positive.