1. Better algorithms, more efficient, faster, more capable. 2. Better hardware, for example bespoke chips for each particular common computation. Could figure out molecular computing as Drexler predicted. 3. Recruit more computer capacity. Every networked computer could be used by a well-funded AI willing to pay enough. 4. Applying AI to exponentially increasing quantity of tasks by progressive decomposition of tasks into subtasks. Each worked in parallel as needed. 5. AGI would interact with the physical world with robotics that it designs, manufactures and operates.
If this is correct, then we could expect what AI-safety folks name "FOOM". Link: https://www.lesswrong.com/posts/t8dtuYXPhwt93PstB/requiremen...
If encouraged, as opposed to restrained, FOOM could happen days after AGI has the means to improve in the above dimensions. The hardware dimension would be limited at first by the human supply chain, and those limits would be gone when AGI runs its own supply chain - perhaps within a year after AGI.
https://en.wikipedia.org/wiki/Supermind_(novel)
which like a lot of science fiction confuses high intelligence with E.S.P. You could increase the number of coefficients in Chat-GPT 1000 times and it won't be able to tell you who won a sports game that happened last weekend if it was trained a month ago unless somebody told it. (Bing Chat succeeds at this because it gets fed information on demand from Bing.) To do otherwise would be "E.S.P."
I think there are various practical limits on intelligence and I'm not sure if a higher intelligence could really do much more than we can. For one thing there are various results such as
https://en.wikipedia.org/wiki/G%C3%B6del%27s_incompleteness_...
https://en.wikipedia.org/wiki/Halting_problem
and just because your model of computation is based on neurons instead of opcodes and registers does not make those limitations go away.
If people are not making progress on a "theory of everything" or other tough problems in fundamental physics such as "What is dark matter?" it is not because they aren't smart enough to work out a theory. There are many theories of quantum gravity today and many ideas about what dark matter could be. To prove one of them would take a difficult experiment that involves manipulating large amounts of matter and a higher intelligence still has to do that experiment.
We see a lot of highly effective people that are experts in two domains. I'm not sure we have examples to look to, but I think an intelligence that can run in parallel and is an expert in N domains is going to be a Big Deal.
I do like your thoughts on it not knowing the future, or having omnipresence. I hope those remain powerful barriers for AI.
It's a strange approach. You don't see people building biological weapons because they want them regulated. But ChatGPT was out there to prove that AI is a real thing and not some interns in a box.
AGI: ETOOMANYTOKENS
ASI: ETOOMANYPEOPLE