i have been told that gpt5 is scheduled to complete training this december and that openai expects it to achieve agi. which means we will all hotly debate as to whether it actually achieves agi. which means it will.

Mar 27, 2023 · 5:00 AM UTC

215
424
224
2,714
so how does an LLM, even gpt5, get to agi? don’t you still need to prompt it? doesn’t it need to like, do stuff, have goals? well, imagine you do something like this: give it a goal, put it in a loop, with all of the power of chatgpt plugins, and with gpt5 instead:
Yohei
 @yoheinakajima
27 Mar 2023
So this “AI founder” experiment is kind of blowing my mind. I set an objective, and say “your first task is to create your next task”. It then continues to generate and reprioritize its own task list as it executes them one by one. Only hooked up to search now. Kinda scary.
15
30
3
322
we are SO much closer than people think.
16
11
7
281
uhhhhh
Yohei
 @yoheinakajima
27 Mar 2023
Replying to @blader
Weird fact, I don’t have gpt4 access so this is only gpt3
3
1
116
*also to be clear I don’t mean to say achieving agi with gpt5 is a consensus belief within openai, but non zero people there believe it will get there.
22
4
3
136
Replying to @blader
Not sure if it’s a quip or not. Do you think it’ll be agi?
1
2
definitely (see follow up)
1
8
more replies
Replying to @blader
I think if you showed GPT-4 to the entire world, probably 20% would already call it AGI. I bet by GPT-5, that'll be more like 50%.
7
3
2
162
yeah agi is a continuum not a step function and that would be a pretty good test
3
5
133
more replies
Replying to @blader
in the most recent interview with Sam he doesn’t believe so?
5
23
not a fine line but i think we get there for some reasonable definition of it
Siqi Chen
 @blader
27 Mar 2023
self prompting + toolformers + continuous gradient descent gets you the rest of the way there
1
2
20
more replies
Load more