OpenAI's Deep Research ​AI indicates a "Hard Takeoff"

[John Clark]

The DeepSeek R1 AI very recently received a 2X speed up, and the software improvement was not written by a human, it was devised by DeepSeek R1 itself. Kevin Roose is the science editor at the New York Times and this is what he had to say: 

"When I wrote about Humanity's Last Exam the leading AI model got 8.3%, five models now surpass that, and the best model gets 26.6%. AND THAT WAS 10 DAYS AGO!" 

OpenAI's Deep Research AI indicates a "Hard Takeoff

John K Clark    See what's on my new list at  Extropolis

fcm