@BorisMPower@thecaptain_nemo I think in the long run it's the hardware manufacturers who will benefit with more need for compute and the technology will eventually be deployable on local hardware. It's then we will see major economic impact. Btw, how much does it cost for one deep research query on average?
@GaryMarcus Actually there will be a lot of demand for better hardware and chips, although for a different segment. Everyone would love to run these models locally. There will also be a lot of potential for funding small niche AI apps for whom operating costs will keep going down.
@_xjdr R1 is damn good. Claude has better function calling and is better for api usage. 4o has been giving out gibberish lately. I think they have been shrinking and distilling model sizes to cut costs.
@fchollet How is it different from alphazero for games? Your approach would require RL if you want it to continuously learn. And how does your approach use LLM's or do you want to opt for a completely different framework?
@huybery Having used earlier versions and the tiny ones as well, I think the Qwen models are the best in knowledge retention, better than the llama's and also more steerable.
@imjustnewatai You could also have a slider or something that controls how much freedom you are giving o1 to take it's own decisions or asking it to strictly adhere to the prompt.
@imjustnewatai I still think o1 preview is not able to follow instructions properly. In my workflow and testing I found it going off and editing some already tested functions inspite of clear instructions to keep those functions in my code unaltered. Maybe we need a better o1-instruct first.
254K Followers 7K FollowingOG GenAI Skeptic; spoke at US Senate. Warned about hallucinations in 2001. Advocating world models & neurosymbolic AI ever since. Author, Marcus on AI & 6 books
225K Followers 0 FollowingOG of: P and T in ChatGPT, recursive self-improvement, neural distillation, GAN/World Model, deepest learning. Co-authored most-cited AI paper of 20th century
32K Followers 225 Followingscaling @ Nemotron
getting us to singularity with friends | angel
computers can be understood: https://t.co/doHE1Quv2L
x founder, @GoogleDeepMind @Microsoft
44K Followers 224 FollowingI built a C library that lets you compile 12kb static binaries that run natively on Linux, Mac, Windows, FreeBSD, OpenBSD, NetBSD and BIOS using just GCC/Clang.
26K Followers 0 FollowingPure Mathematician+pianist. Grieving involuntary childlessness. Author: Unequal: The mathematics of when things do and don't add up, out 2025
50K Followers 126 FollowingMechanistic Interpretability lead DeepMind. Formerly @AnthropicAI, independent. In this to reduce AI X-risk. Neural networks can be understood, let's go do it!
2.0M Followers 179 FollowingNobel Laureate. Co-Founder & Chair @GoogleDeepMind; Chief Scientist of Alphabet. Founder & CEO @IsomorphicLabs. Building the future...