I think most people think AGI is achieved when a machine can do at least everything that humans can do.
Like not necessarily physical things, but mental or digital things.
Humans will create a better LLM, (say GPT-5) than all the other LLMs that currently exist.
If you tasked any current LLM with creating a GPT-5 LLM that is better than itself, can it do it? If not then it's probably not AGI and has some shortcomings making it not general or intelligent enough.
Like not necessarily physical things, but mental or digital things.
Humans will create a better LLM, (say GPT-5) than all the other LLMs that currently exist.
If you tasked any current LLM with creating a GPT-5 LLM that is better than itself, can it do it? If not then it's probably not AGI and has some shortcomings making it not general or intelligent enough.