The AI war is not over. But at this point in time, PRC related entities seem to be pulling ahead. U.S. based companies may lead in cutting-edge closed models like Gemini, Grok, ChatGPT. But Chinese firms are dominating in open source and in applied AI diffusion (for example, into
RESEARCH
-
LLMs World Models Robustness Debate LeCun Bengio Sutton
By
–
If you want to actually understand the arguments about LLMs and world models—and why LeCun, Bengio, @RichardSSutton and I have landed in similar places—read my essay “Generative AI’s crippling and widespread failure to induce robust models of the world”
-

CoRL 2025 Data Curation Workshop Recording Now Available
By
–
Thx Chris! #CoRL2025 Workshop on Data Curation recording is now online https://
sites.google.com/stanford.edu/c
orldata25/home
… -

ByteDance Lynx: Photo-to-Video AI Model with Identity Preservation
By
–
🚨 ByteDance unveils Lynx, a next-gen video model that turns a single photo into smooth, lifelike clips while keeping facial identity intact.
— Futurepedia – Learn to Leverage AI (@futurepedia_io) 28 septembre 2025
Key highlights:
▪️1 photo → full video with strong identity, motion & quality scores
▪️Built on a Diffusion Transformer with two clever… pic.twitter.com/LZfwiJ65i1ByteDance unveils Lynx, a next-gen video model that turns a single photo into smooth, lifelike clips while keeping facial identity intact. Key highlights:
1 photo → full video with strong identity, motion & quality scores
Built on a Diffusion Transformer with two clever -

GitHub E-book Treasure Chest with Free Technology Downloads
By
–
Found an e-book navigation site on GitHub with many categories, free downloads https://
github.com/jbiaojerry/ebo
ok-treasure-chest
… -
OpenAI Researchers Allowed Outside Contrary Popular Belief
By
–
Contrary to popular belief, OpenAI researchers are in fact allowed to go outside
-
CaMeL Pattern: Building More Sophisticated AI Systems
By
–
The most convincing pattern I've see for building more sophisticated things than that is CaMeL, but it's extremely hard to implement
-
Reachy 2 robots deployed to research institutions worldwide
By
–
https://t.co/2ieEakgPF9 https://t.co/HAdQ3vVs1J
— clem 🤗 (@ClementDelangue) 27 septembre 2025While we're working hard on shipping the first Reachy Minis (planned before the end of Sept), we shipped with @pollenrobotics @LeRobotHF a bunch of Reachy 2 to Research labs all over the world this summer. This is the unboxing from the Fraunhofer Institute for Industrial
-
Authoritarian Bias in Western AI Happiness Research Methodology
By
–
Western “non-partisan” studies should know better than to compare happiness across authoritarian and free societies. In places where dissent has consequences, answers reflect survival as much as sentiment.
-

Can Large Language Models Understand the Real World?
By
–
Can large language models figure out the real world? How deeply predictive AI systems understand their subject matter? https://
news.mit.edu/2025/can-large
-language-models-figure-out-real-world-0825
… #AI #MachineLearning #MIT #Harvard @mitsmr@AlexMachicado@Analytics_699
