> All it takes is one developer to show one project that was delivered in less time than they'd have been able to do otherwise and bingo-bango: AI tools as they exist today—flawed as they might be—can do something a human couldn't have done otherwise
Author forgot one problem with this. If AI creates one good project (or even image/audio for that matter) it does not automatically mean it can produce anything else in that space. On the contrary, I think this IS the main obstacle for LLMs to achieve any semblance of AGI. Right it's more about finding what subspaces AI is good at, and even then the output will still be incosistent.
Author forgot one problem with this. If AI creates one good project (or even image/audio for that matter) it does not automatically mean it can produce anything else in that space. On the contrary, I think this IS the main obstacle for LLMs to achieve any semblance of AGI. Right it's more about finding what subspaces AI is good at, and even then the output will still be incosistent.