Good writeup. I am surprised the anonymous researchers mentioned seem to update more on the DS results, rather than o3 benchmarks? I guess we will just go even faster but I found that event 'spookier'.
Ah, maybe I didn't make explicit enough in this post but the reason they were spooked by the DeepSeek results was because of the way DeepSeek got there. The idea that they were able to generate really strong results with much less human data and without that human guidance was what surprised researchers and made them realise that DeepSeek had brought us closer to a jumping off point.
Good writeup. I am surprised the anonymous researchers mentioned seem to update more on the DS results, rather than o3 benchmarks? I guess we will just go even faster but I found that event 'spookier'.
Ah, maybe I didn't make explicit enough in this post but the reason they were spooked by the DeepSeek results was because of the way DeepSeek got there. The idea that they were able to generate really strong results with much less human data and without that human guidance was what surprised researchers and made them realise that DeepSeek had brought us closer to a jumping off point.