#DeepSeekR1 drives the world crazy but no, it's not the best Chinese #ChatGPT! @AlibabaGroup just released its AI, Qwen 2.5 Max, and it's just crazy. I test it in my latest video right here → https://youtube.com/watch?v=zE_dM7DwPQw … Everything is free… Even generate videos.
RESEARCH
-

What Could Go Right: Reid’s Optimistic AI Perspective
By
–
Reid's favorite question is "What could go right?" It turns out the answer is "A lot!" I really enjoyed this book
-
AI Making Humanity Smarter for Future Challenges
By
–
This is what we urgently need. AI that makes humanity smarter will be what helps us navigate the craziness of what's coming in the next century.
-
Transformational AI: The Next Era of Human Empowerment
By
–
This week yet again underlines that this moment is very nearly here. …it’s the next step toward truly transformational AI. If we can harness this safely and contain its downsides, this really will become the most empowering and creative phase of human history.
-
AI Paradigm Shift: From Imitation Learning to Reward Learning
By
–
This is a monumental shift in AI. We’re moving from a world of imitation learning (with SFT), to reward learning (with RL). AI that not just creatively imitates images and language but truly invents new strategies and insights.
-
AGI Systems for Knowledge Discovery: A 2010 Vision
By
–
What drew me to work on AGI back in 2010 was the dream of creating systems that could discover new knowledge – the most inspiring vision I could imagine then and now.
-
Bloomberg segment on AI and DeepSeek developments
By
–
I got on Bloomberg News yesterday to talk about AI and the DeepSeek hullabaloo. You can find my segment starting on minute 58.
-

Teaching LLMs with Textbook Information Structure
By
–
We have to take the LLMs to school. When you open any textbook, you'll see three major types of information: 1. Background information / exposition. The meat of the textbook that explains concepts. As you attend over it, your brain is training on that data. This is equivalent
-
Speed and Quality Correlation in Expert Performance
By
–
Some key takeaways: 1. Speed and quality aren’t actually at odds—they’re often positively correlated. The best practitioners in any field, from chefs to programmers, work quickly because they’re experts. The speed doesn’t come from rushing or cutting corners but from competence.
-

DeepSeek R1 70B Outperforms GPT-4o and o1-mini
By
–
DeepSeek’s R1 70B combines the powerful reasoning ability of the full R1 model with the size and speed of Llama 70B. R1 70B outperforms GPT-4o and o1-mini across a range of general and reasoning benchmarks, making it the most capable Llama 70B variant by far.
