Bullshit detector going off     See Less

It’s refreshing to see someone dive into the nuance behind the hype and skepticism around AI progress. The contrast between the slowed improvement in models like Orion and the breathtaking     See More from Sam Altman really highlights how narratives can shift depending on what you choose to focus on. It’s also fascinating that even OpenAI itself seems uncertain about how much further scaling will take us, which feels like a rare admission in such a competitive space.

The idea that progress in reasoning benchmarks like Frontier Math is limited by data scarcity is eye-opening. It raises an important question about whether breakthroughs might come from entirely different approaches, like test-time compute or new methods of training. The analogy of solving math problems to tackling other domains really brings home the challenges AI still faces, especially in reasoning-heavy tasks.

But I love how this video also pointed out the explosive progress in areas like video generation and speech-to-text, where the abundance of training data seems to be a huge advantage. It’s a reminder that while some parts of AI may slow down, others are likely to leap forward—and the balance between those will shape what AI can do in the near future. Definitely a thoughtful take that avoids leaning too far into either extreme.    See Less

I came back to this video for the FrontierMath paper... Man o3 getting 25% there... I have no future in science anymore, and my career was just getting started.     See Less

I'm slightly confused: ChatGPT hasn't even mastered counting, as in the 'r's in 'strawberry' and all that. How do you expect it to solve Olympiad-level problems that     See More world's smartest math geniuses?    See Less

Sam Altman is heavily incentivised to overhype the product though, because he's constantly trying to attract more investors. He's the least reliable messenger in AI     See Less

Yea GPT-5 might not be here anytime soon, but you can check out Flock's customized GPT for try, no hurt right     See Less

That Sora rumor was correct!     See Less

I was never betting on these models, I've been betting on Kurzweil     See Less

I still remember. When GPT4 was relesead and "AGI soon" reached pitch fever i used to comment:
"guys, like everything else, this will not go that far. It will hit a wall, l     See More soon". You guys have no idea how many times the "believers" called me d.u.m.b. .
Here we are.     See Less

I believe we must slow down the progress of AI at this point to allow for rigorous testing, ensuring accuracy and safeguarding humanity.     See Less