It's the first time I've had this much fun with an AI model since O1 launched. The new OpenAI image gen is crazy good for some good memes lol. It's also amazingly good at making cool profile pictures.
Hats off to the image gen team at @OpenAI
@kimmonismus It's more like , in three hours it played millions of games which you as a human could never dream of haha, it's mostly got the advantage of processing data much faster. Would be interesting to see who's better after 1000 games a human or something like alpha zero.
@kimmonismus You can get very impressive numbers depending on what you test it on and the dataset used . On the official university website they do mention that it does generalize well on several other datasets; would be curious to see if it performs that well in real world settings.
So you kinda start thinking o this is nice but then try to empirically justify some odd property of a theorem , all you can do is hold on to those initial axioms lol 🥺
Studying rigorous math is hard, as it often, the further you advance, tries to pull you away from empirical experience into rational deduction that arise from often intuitively defined axioms.
@aidan_mclau O and the EQ update seems to be very real. I asked the same question to both o4 and 4.5 and the 4.5 response felt way more genuine, authentic and thorough. It doesn't overuse emojis or specific phrases it's very balanced. So far enjoying this update 🥹
@aidan_mclau It feels good so far.
Only major downsides are rate limits and speed of response. 4o feels crispier in terms of output speed and the rate limit feels unlimited.
Would be nice to increase it for 4.5 since if your'e just having small chats with it you'd get to 50 prompts fast.