I have a friend who likes to take steroids to get big and ozempic to get thin and he is rather excited about it. One day he suggested your only options to get thin are ozempic or smoking crack. This is how people talk about ai these days.

Clickbait!

Easily refuted by the evidence of Anthropic's Claude Code in enterprises for the purpose of automating programming development activities.

Perhaps 10000x cheaper to perform tasks that the AI and humans can both competently perform.

Reading through Sutskever's deposition is a great example of where the approach to AI mentioned in this post, is failing. The way the world is setup means you need politicians who are hungry for power and willing to do the not so human thing.

Asking for consent to train content, make sure sure AI is "green".. these of idealogical fancies that don't match up to the real world and the winners the folks walking straight over these.

In both cases a smarter model should be help to help offset the greenhouse gas effects and then some when used to discover new materials etc. Hallucinations can be managed, detected and corrected in most cases when doing research and in many cases are just treated as failed attempts as it is easier and cheaper to just ask the model to try again.

The only idea worth chasing here is a community driven and open source approach.

We have this in may formats... my fave is Ben Goertzel's OpenCog .. but there is more. I think Emad Mustaq is doing something similar with decentralised compute/AI.

Again - all bleeding edge is for profit.. not even OpenAI's idealogical foundation was able to be saved from being turned in to a for profit gig.