Read on Twitter
The big multimodal currently in the works at Openai is called Gobi.
Should I leak more?
They have an internal 64k model, can’t release with the huge compute issues atm.
Dear @sama I wrote you, but you still ain’t callin’
I left my cell, my pager and my home phone at the bottom
I sent two letters back in Autumn, you must not’ve got ‘em
Dear @sama I wrote you, but you still ain’t callin’
I left my cell, my pager and my home phone at the bottom
I sent two letters back in Autumn, you must not’ve got ‘em
"the plan is to not talk about any of the training specifics of GPT-4, including trying to avoid indirectly exposing specifics by deduction"
But it’s very similar to 3.5 ;)
They’ve had a jukebox v2 with diffusion as a slack bot for well over a year now.
And for the “ big goal “ ;) that’s for me to talk about later. Maybe.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
