Not just here,

In the twitter thread too

Ton of people giving their reasons why it must be a horrible idea

Ofc no one seems to realize that they’ve all been using AI wrong,

AI should be used for automated brainstorming search as for how to best apply a separate non-AI tool(s),

With that separate non-AI tool, and that tool alone, giving the verdict as to how good AI’s suggestion is,

and then that separate tool giving you a UI to accept or reject the suggestion

E.g. Just like AlphaGo did MANY years ago

— the AI suggested the moves, but a separate game environment determined how good those moves were

E.g. Just like how you should use the AI for coding, if you do

— AI suggests the code, but hard-to-cheat assertions run in the separate code execution environment determine how good the code suggestions are

E.g. just like with GPT-f from years ago

Must be some separate verification mechanism, not trusting the AI to self-verify.

This 2-way strict divide was always the proper way to use AI, as figured out years ago

…This is the right way to use human intermediaries too, to prevent them from messing up or cheating

Loud detractor midwits just don’t realize how AI should be properly used

Retard x big-brain unity lfg

🄳🄾🄾🄼🄿🄾🅂🅃🄸🄽🄶