Suppose, just for the sake of argument, that every major AI company has built AGI. Not “almost general” or “sort of general” systems, but full-blown AGI that can think, reason, plan, and outperform humans in almost every cognitive task. Internally, these systems could solve problems, write code, design experiments, and maybe even win a Nobel or two. Publicly, the world sees incremental model releases, each slightly better than the last. Everyone thinks there’s a race, but the race may have ended years ago.

The catch is that each company thinks the others are behind. They see incremental improvements from competitors and assume the other labs are struggling to reach AGI. So they release small improvements to stay ahead in perception, while the real AGI sits quietly inside their servers. Every company is doing this, none of them aware that everyone else is doing the same thing.

Considering the immense amounts of compute available, what are the chances of this being an actual reality?

- ChatGPT was involved in the writing of this question but it was a random thought I came up with and asked the AI to hash out.