> “GPTs will NEVER solve this simple puzzle” > issues $10k challenge to be proven wrong > proven wrong in less than 24 hours

Apr 8, 2024 · 5:39 PM UTC

13
30
2
465
This is a good example of how AI is advancing faster than most people think. Sincere kudos to @VictorTaelin for putting his money where his mouth is — admirable!
1
2
122
Arpit Gupta
 @arpitrage
Apr 8
Replying to @AlecStapp
GPTs will NEVER solve this very specific problem that I face all the time and would benefit from cracking, prove me wrong
1
15
Replying to @AlecStapp
GPTs will NEVER solve this very specific problem that I face all the time and would benefit from cracking, prove me wrong
2
1
114
lmao
10
Replying to @AlecStapp
GPTs didn’t… Claude did
2
4
Replying to @AlecStapp
Best bit - admitted it immediately!
7
Replying to @AlecStapp @hamandcheese
GPTs are more impressive than a lot of people give it credit for.
4
Replying to @AlecStapp
Lots of people said it would be possible before the $10k challenge, but no one proved it. And the prompt to prove it was hundreds of lines long, not a trivial solution. The $10k bought an answer to the question, and got it quickly.
2
Replying to @AlecStapp
On the other hand, what GPTs will REALLY never solve is @fchollet’s ARC benchmark, an actual test of AGI: lab42.global/arc/
Replying to @AlecStapp
Not only that, but I told this idiot exactly why he couldn't get it working with chatgpt... It was just an issue with the tokenization of the "#" sign in his prompt, nothing inherent to the capabilities of the model
Load more