Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Stocastic parrots are going to stochasticate.

The author also cited a few human assisted efforts as ML only.

The fact that the author also is surprised that GPT is better at falsifying user input while it struggles at new ideas demonstrates the fact that those who are hyping LLMs as getting us closer to strong AI don't know or ate ignoring the know limitations of problems like automated theorem solving.

I think generative AI is powerful and useful. But the AGI is near camp is starting to make it a hard sell because the general public is discovering the limits and people are trying to force it into inappropriate domains.

Over parameterization and double decent is great at expanding what it can do, but I haven't seen anything that justifies the AGI hype yet.



Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: