"We are not currently training what will be GPT-5; we don't have plans to do it in the next 6 months" – Sam Altman, under oath

May 16, 2023 · 5:31 PM UTC

26
10
7
336
Replying to @daniel_271828 @daniel_eth
Maybe they’re doing a 4.5?
1
14
Yeah maybe - would that be larger or further fine-tuning on GPT-4?
1
8
more replies
Replying to @daniel_271828 @daniel_eth
Because it's GPT-∞.
1
7
It's GPT-4 Jr.
9
Replying to @daniel_271828 @daniel_eth
GPT4.9˙
1
5
GPT-4 the 2nd
1
7
more replies
Replying to @daniel_271828 @daniel_eth
I’ve always took this to mean “we are working on it, we just haven’t started the training run yet”
1
40
Yeah, it means the training run hadn’t started yet
1
26
more replies
Replying to @daniel_271828 @daniel_eth
@sama also said that we shouldn't think too much about parameter count. Lots of room to make current models better w/out more params, so the current work is probably to optimize GPT-4 and make it better and/or cheaper to run. Azure bills are adding up! patmcguinness.substack.com/
Replying to @daniel_271828 @daniel_eth
Jokes on them, GPT-5 is already trained and they are working on GPT-6...
1
25
Load more