Hacker News new | ask | show | jobs
by Tepix 743 days ago
Arguably we're seeing bad engineering by OpenAI leading to their superalignment team quitting and raise the alarm.
2 comments

No.

Re-wind the clock:

- A google engineer says "google has AGI"... this leaks and is dismissed.

- MS writes a paper that says GPT has hints of AGI.

- OpenAI signs deal with MS saying they can have everything OpenAI does till they hit AGI.

That last one. Thats not a bet you make if you think AGI is a decade away. Its a bet you make if you think your close.

OpenAI thought that they were close. They fell for the classic AI trap of "if only we had a bigger model". They got high on their own supply and thought that some sort of tipping point would come with size or scale. That a bigger model with MORE data and MORE information would have AGI, sapience or sentience emerge.

It was a bad bet. Q* isnt going to get there, the LLM line is dead.

Why do you need to worry about safety or super alignment if you arent going to get to AGI. They left because there is no THERE, there..

> They left because there is no THERE, there..

They left because they lost the power struggle.

That makes zero sense.

If your safety oriented, and your company is close to AGI you stay regardless. Your boss isnt going to push you out because you leave and scream "it will kill us all"... All the safety people have been doing that for years, now they are quite.

If you aren't close, or realize "we're never getting there" leaving, vesting makes sense. There's NO existential risk.

Talk of AGI is hubris, vanity and snake oil.

Philosophy majors judging bad engineering and making grand stands.