Jagged AGI: o3, Gemini 2.5, and everything after

The capabilities of AI post gpt3 have become extraordinary and clearly in many cases superhuman.

However (as the article admits) there is still no general agreement of what AGI is, or how we (or even if we can) get there from here.

What there is is a growing and often naïve excitement that anticipates it as coming into view, and unfortunately that will be accompanied by the hype-merchants desperate to be first to "call it".

This article seems reasonable in some ways but unfortunately falls into the latter category with its title and sloganeering.

"AGI" in the title of any article should be seen as a cautionary flag. On HN - if anywhere - we need to be on the alert for this.

The exact definition of AGI is pretty much the least interesting thing about AGI. It's basically bike-shedding at this point: arguing about something easy to understand instead of tackling the really hard questions like "how competent can AI get before it's too dangerous to be in the hands of flakey tech companies?"

> how competent can AI get before it's too dangerous to be in the hands of flakey tech companies?

Ever heard of Pandora's Box? Yeah. That ship has sailed. No moratorium you could enact would, at this point, stop the innovation from happening, possibly even independently by multiple teams globally. Economic incentives are stacked in such a way that flakey tech companies will prioritise shareholder value over anything else. Whatever comes next will come, and all we can do is lean back and enjoy the show.

Given the gigantic amount of compute power and storage needed to train and run LLMs, this is certainly not true. It is absolutely feasible for government to check every data center capable of advancing the state of the art in AI to ensure that no such research is taking place.

Of course, the political will to do so doesn't exist to even a tiny extent. But if such a will existed, it would be far easier to enforce than the prevention of human cloning, and that one has been successfully implemented for decades now.

If you just bow out of the AI race you are handing it to other countries where practices might not be subjected to the same standards. It's suicide to do this.

That's only if countries can't agree on this. I just gave the example of human cloning, which has been banned globally. You can also look at nuclear non-proliferation, which has been largely successful (though not completely) despite huge incentives for any country to defy it.

Now that’s a big if.

Given the current state of the world, do you really think the USA, China, India, Iran, Brazil, North Korea, and Russia, would all have the same opinion on the danger of AI systems and would—despite very obvious and tangible strategic advantages—all halt development for humanity’s sake?

Human cloning is an issue that is mostly academic in nature, but I’d bet everything I have that bioengineers all over the world secretly are researching this on government programmes, and nuclear non-proliferation is a joke. It was essentially about stripping Russia of its nukes, but all global powers still have them, and countries like Iran, North Korea, and India measure their development on the possession of nuclear weapons. It was successful only if by success you mean the USA didn’t maintain their minuteman program.