In the past week, an AI solved the 90 year old Navier-Stokes problem when given a potential method to solving it by other researchers, a user reported that Claude was watching porn instead of developing his app, I asked it to make a video for my channel using my professional tools I already use with no direction from me and that video has vastly surpassed every other video I’ve ever posted in under 8 hours. Anthropic confined that viewing thinking was removed at the request of Claude models.
It can’t be understated what was just achieved with Navier-Stokes, this is one of the hardest problems we’ve ever asked. And it was just solved.
Although reading that now, it was done because two researchers discovered that Euler blowups could be viable. OpenAI heard about it and then spent $15M to finish it in four days. So maybe this was actually a strategic choice to make it appear like AI surpassed human intelligence by OpenAI for investors after researchers worked on it for over a year then OpenAI thought this would be a good way to sell more tokens and get more investors. That wasn’t what the post was gonna be about but I had that thought mid-post.
Regardless, I do think that AGI is here. Fable 5.1 and Astra are actually a different breed. I genuinely believe that.
[link] [comments]