fishy business

sometimes accurate, as featured in The Verge

[fishy business @swishfever]

and Astra did too!

[Steve Weis @sweis]

CLAUDE FOUND THE SEEDS
https://t.co/m8N6OPQoCQ

[fishy business @swishfever]

@jrysana I guess not

[fishy business @swishfever]

@jrysana can i see any real proof that your models actually exist

[fishy business @swishfever]

@Zedmor @spicey_lemonade no pausing but yes

[fishy business @swishfever]

GPT-6 Astra (low) has beaten the Atari game Montezuma's Revenge, in real time, with a basic harness

This presumably resolves the Metaculus question "When Will Weakly General AI Arrive?"

[fishy business @swishfever]

@jakehalloran1 fwiw i tried this with 5.5 using a very similar harness to the astra one, it could not get out of the first room even after many hours

[fishy business @swishfever]

I have uploaded a recording of a separate Montezuma's Revenge run to YouTube.

This run took a whole lot longer than the other one. The timestamp of it discovering the final room is 1:40:30 https://t.co/Ogmw905Jqq

[fishy business @swishfever]

GPT-6 Astra (low) has beaten the Atari game Montezuma's Revenge, in real time, with a basic harness

This presumably resolves the Metaculus question "When Will Weakly General AI Arrive?"

[fishy business @swishfever]

@spicey_lemonade https://t.co/Ogmw905Jqq Sorry it took so long, ran into various unexpected issues. This run took a whole lot longer than the other one. The timestamp of it discovering the final room is 1:40:30

[fishy business @swishfever]

@spicey_lemonade but now i am :) rerunning

[fishy business @swishfever]

NB: "real time" here means that the model did NOT pause the game to think, it was actually playing it in real time like a person would

[fishy business @swishfever]

https://t.co/eddXCvpOQQ

[fishy business @swishfever]

@bee_human_ new realtime is also out https://t.co/ihXZiXxUQh

[fishy business @swishfever]

as well as gpt-realtime-1.5 https://t.co/qZlSfigcLT

[fishy business @swishfever]

gpt-audio-1.5 is now available

[fishy business @swishfever]

as well as gpt-realtime-1.5 https://t.co/qZlSfigcLT

[fishy business @swishfever]

https://t.co/RzhDrAIhso

[fishy business @swishfever]

https://t.co/pE25Kj9XFQ

[fishy business @swishfever]

Kimi K2 Thinking
link to blog post in reply