Had a great time talking to @lachygroom & @hausman_k of @physical_int at @Lux_Capital
's event in NYC this week. Fascinating to hear about surprising progress on robot models. But hey still don't know where we are on the road to GPT-4 level robot models. (Photo: Ben Hider)
@willknight
-
Robot Models Progress Compared to GPT-4 Level Capabilities
By
–
-
AI Journalism: Mapping CEO Alignments Through D&D Charts
By
–
I think 75% of AI journalism is figuring out where various CEOs sit on the D&D alignment chart.
-
Data Centers and the Religious Fervor Around Super-Intelligence
By
–
Given the quasi-religious way people think about super-intelligence, I keep wondering if today's incredible data-center build out might be comparable to the construction of cargo cults.
-
Humanoid Robot Sprinting: Can It Reliably Recover From Falls?
By
–
The lab mentioned developed a humanoid for sprinting, with very explosive leg movements. https://
newshub.sustech.edu.cn/en/html/202508
/46874.html
…
What's unclear from the video is if this is a fluke or can reliably help the robot get up after falling. -
Boston Dynamics’ Atlas Robot Demonstrates Generalized Learning Capabilities
By
–
This week's newsletter: Boston Dynamics’ dancing humanoid, Atlas, can now use a single model for walking and grasping, and exhibits some 'emergent' skills—a significant step toward more general robot learning. https://t.co/G7Qn2GAMBV
— Will Knight (@willknight) 5 septembre 2025
Subscribe here: https://t.co/NUffPrZt92 pic.twitter.com/kzDK35UuoTThis week's newsletter: Boston Dynamics’ dancing humanoid, Atlas, can now use a single model for walking and grasping, and exhibits some 'emergent' skills—a significant step toward more general robot learning. https://
wired.com/story/this-hum
anoid-robot-is-showing-signs-of-generalized-learning/
…
Subscribe here: https://
wired.com/newsletter -
Human Learning vs AI Rule-Writing: An Inverse Problem
By
–
This is such an interesting prompt-question. It's kind of the inverse of the way humans can learn to do all sorts of physical things but wouldn't be able to write down the "rules" for how to do them.
-
Why LLMs Struggle to Follow Chess Rules Consistently
By
–
@hyhieu226 Why can't an LLM write down the rules of chess and then follow them? A limited kind of world model.
-
Robots Taking Jobs: Separating Hype from Reality
By
–
None of these are examples of robots taking people's jobs en masse. Videos of bots doing cool stuff is fine, but be wary this kind of scaremongering. Robots are having an impact on labor and wages but these videos are misleading. https://t.co/SxukZmhkk5
— Will Knight (@willknight) 21 août 2025None of these are examples of robots taking people's jobs en masse. Videos of bots doing cool stuff is fine, but be wary this kind of scaremongering. Robots are having an impact on labor and wages but these videos are misleading.
-
AI Model Training Energy Requirements Forecasted by Epoch
By
–
A whitepaper from Epoch AI, a company that tracks AI trends, and the Electric Power Research Institute attempts to forecast the energy that will be required to build future AI models.
-
AI Model Evolution Safety and Training Infrastructure Challenges
By
–
Other AI research in last week's AI Lab newsletter: The evolutionary tree of AI models. How to stop chatbots learning to be unsafe. And the insane power needed for future AI training runs. …