Except that the bubble's money is not being invested into cutting-edge ML research, but only into LLMs. And it has been obvious from the start to anyone half-competent about the topic that LLMs are not the path to AGI (if such a thing ever happens anyway).

I don't think it's that obvious, in fact the 'bitter lesson' teaches us that simple scale leads to qualitative, not just quantitative improvement.

It does look like this is now topping out, but it's still not sure.

It seems to me a couple of simple innovations, like the transformer, could quite possibly lead to AGI, and the infrastructure would 'light up' like all that overinvested dark fiber in the 90s.

> But what if someone reaches autonomous AGI with this push?

What if Jesus turns up again? Seems a little optimistic, especially with several leading AI voices suggesting that AGI is at least a lot further away than just parameter expansion.

It seems rather more likely to me, even if it's millennia way, that we get a semblance of an autonomous agentic AI, than what you suggest.

It might be impossible, or just need some innovations (eg, transformer), but my point is the investments are non-linear.

They are not investing X to get a return of Y.

If someone reaches AGI, current business models, ROI etc will be meaningless.

> If someone reaches AGI, current business models, ROI etc will be meaningless.

sure, but its still a moonshot, compared to our current tech. I think such hope leaves us vulnerable to cognitive biases such as sunk cost fallacies. If Jesus comes back that really would change everything, that's the clarion call of many cults that end in tragedy.

I imagine there is fruit that is considerably lower hanging, that has more obvious ROI but is just considerably less sexy than AGI.

Probably the most reliable person I can think of to estimate that would be Hassabis at Deepmind and he's saying like 5 years give or take a factor of two. (for AGI, not Jesus)

head guy at Meta stated in a talk that current techniques ain't gonna be enough and they at least need to start to be more creative.

Yeah that seems the consensus. There are lots of people working on the stuff though.

> But what if someone reaches autonomous AGI with this push?

What is "autonomous AGI"? How do we know when we've reached it?

When you can use AI as though it's an employee, instead of repeatedly 'prompting' it with small problems and tasks.

It will have agency, it will perform the role. A part of that is that it will have to maintain a running context, and learn as it goes, which seem to be the missing pieces in current llms.

I suppose we'll know, when we start rating AI by 'performance review', like employees, instead of the current 'solve problem' scorecards.