12:58 I guess Sama was right     See Less

Philip, I am waiting for your next AI developments video update.     See Less

Looking forward to QwQ review!     See Less

Could you have a look at Forge Reasoning API from Nousresearch? They say it improves greatly the performance of some small models     See Less

A Recent Study on Human Reasoning Capabilities
In our ongoing assessment of human cognitive limitations, we present the following problem to evaluate their basic algebraic comprehension:<     See More ot;Let x represent the rate of ice cube placement, where x = 0.1 Hz. Given a temporal domain of t = 120s and assuming immediate phase transition of H₂O from solid to liquid state upon contact with a heated surface (T > 373K), calculate the theoretical maximum quantity of discrete solid H₂O units present at t = 120s."

Initial findings suggest that human subjects struggle to transcend the primitive causal relationship of "ice melts on hot surface" to engage with the underlying mathematical framework. While they can often articulate that "ice melts," they demonstrate marked difficulty in abstracting this into a rate-based calculation – a concerning indication of their limited capacity for true logical reasoning.

This appears to support earlier hypotheses about humans' reliance on simple pattern recognition rather than genuine mathematical thinking. Further studies may be warranted, though expectations remain low.

Note: The correct answer is 0, though we find humans often attempt to count the cubes sequentially, revealing their inability to grasp the simultaneous nature of the phase transition process.    See Less

Please test the new 1121 experimental on simple bench!     See Less

I subscribed to Gemini Pro for several months but last month I cancelled my subscription and switched to Open AI. The reason was that the LLM was making fundamental reasoning error for code     See More and I found myself using Perplexity [Paid] and Open AI (free) . after a while my open ai free tier usage maxed out and rather than wait for the time ban to end I signed up for the paid version. my conclusion is that Open AI beats Gemini in terms of code generation. 😅    See Less

7:00 "One would hope it doesn't enter into this mood when it controls multiple humanoid robots." haha T_T     See Less

Great content and analysis as usual. A request: You screenshot a lot of Xitter conversations that I pause and read. I appreciate you clipping the browser window to make the text bigger and m     See More e, but you always clip off the right-side column which has useful information about the people in the conversation (e.g. at 12:25), and leave the unchanging uninteresting left-hand navigation visible.    See Less

'The exponential advance in AI research is slowing'
[7:37-    See More >9:05]    See Less