Chris Paxton ✓

In the back of my mind, part of me has always wanted to make a video game some day, and it really feels like i'm running out of time to make that happen before gpt can just one-shot it, and no one bothers buying video games any more. 5.6K views · 50 likes · 1 reposts · 17 replies Open on X →

You either die the disruptor or live long enough to make a clumsy attempt at regulatory capture 1.8K views · 17 likes · 0 reposts · 1 replies Open on X →

This: "You don't have to post-train ChatGPT on every user" is exactly what I think robotics has been missing and why every "chatGPT moment" for robotics has rung a bit hollow. we're at a really exciting time! 2.4K views · 24 likes · 1 reposts · 0 replies Open on X →

Learn contact-rich humanoid manipulation policies, purely within a world model. This is usually not tractable, because you need a very large model to get the contact dynamics right -- instead, @Jsphamigo @Rk4342R couple a large global model with a local model. Very cool work on h 5K views · 59 likes · 6 reposts · 3 replies Open on X →

Reinforcement learning has unlocked tremendous gains in broader AI and humanoid robot whole body control, but still has yet to deliver strong results for robot manipulation due to the difficulty of capturing tasks in an accurate simulation; one way to fix this is to create https: 6.2K views · 29 likes · 7 reposts · 1 replies Open on X →

This is fantastic 4.5K views · 23 likes · 0 reposts · 2 replies Open on X →

The irony of it all. The disruptor becomes the disrupted. 2.3K views · 10 likes · 0 reposts · 0 replies Open on X →

I was really not expecting ARC-AGI 3 to be solved so quickly. But these benchmarks have been dropping so, so fast. In some ways not surprising: humans are comparatively good right now at long horizon tasks and out of distribution generalization, and try as you might it's hard to 3.9K views · 33 likes · 1 reposts · 10 replies Open on X →

I still fundamentally believe there will be some measure of comparative advantage to human intelligence -- there will be things that we are better at than machines. But the trend lines are incredible, and you (as an individual) may not be happy with what ends up on the "human" 2.8K views · 21 likes · 1 reposts · 1 replies Open on X →

You don't have to post-train ChatGPT on every user. If you did, it would never have taken off. Yet this is exactly how robotics works today For robots to take off, they need to learn in-context. Great article from @Chris_J_Paxton on this new paradigm and S1's place in it: https 13.3K views · 122 likes · 11 reposts · 3 replies Open on X →

AI researcher @andymasley explains why society is radically underreacting to the risks of machines replicating human intelligence: "The general idea that there could be all these really unique risks from AI being able to actually replicate a lot of what happens in the human http 11.9K views · 42 likes · 1 reposts · 5 replies Open on X →

The funny thing is that i run my writing through pangram all the time and have never had a false positive. Almost like people are lying about whether they wrote stuff or not 3.1K views · 21 likes · 1 reposts · 3 replies Open on X →

This particular one is almost certainly human in the loop or just teleop -- that is currently the only way to e2e autonomy 1.3K views · 0 likes · 0 reposts · 1 replies Open on X →

This does actually worry me, too many people trying the same ideas 5.3K views · 62 likes · 1 reposts · 3 replies Open on X →

This specific video is teleop 5.4K views · 14 likes · 0 reposts · 0 replies Open on X →

We are in one of the most consensus driven market. 6.2K views · 11 likes · 1 reposts · 1 replies Open on X →

This is why we will see more people training robots not to damage themselves. The barrier is going to be maintenance and support (humans) not actual task completion soon 2.4K views · 21 likes · 1 reposts · 2 replies Open on X →

They got a tariff exemption for the base vehicle and there's no way its 125k. More like 60-70k. Double, but its roomier, and we will have to see if its safer, but the gap isnt as big as it sounds 12.8K views · 85 likes · 0 reposts · 27 replies Open on X →

@chris_j_paxton I’m a big fan of Ultra but they also leverage teleop pretty heavily. Do you happen to know if the robots in this video are teleoped or running policies (or perhaps some hybrid?) Either way, super impressive to see robots paying for themselves instead of creating 909 views · 4 likes · 0 reposts · 1 replies Open on X →

Imagine telling people 4 years ago that robots running end to end neural nets would be deployed and paying for themselves Even today some dont believe it 47.7K views · 290 likes · 24 reposts · 7 replies Open on X →

The reason I'm a fan of @Ultraroboticsco is they do real world AI robotics NOW. This is live footage from a 3PL customer site with three Ultra bots coordinating to build boxes, pack, tape, label and place into the correct USPS, Fedex, etc. bins. Ultra writes all the back end ht 34.1K views · 150 likes · 15 reposts · 11 replies Open on X →

robot BOMs are quickly getting to the point where this threshold is met and policies are improving rapidly too the cost factor most don't account for here is support overhead -- installation, training, repairs, babysitting, etc. it's surprisingly hard to drive this one down 5.6K views · 31 likes · 3 reposts · 2 replies Open on X →

Join us to talk about what we have learned in 100 episodes and to welcome a new co-host @ruijie_sg -- so we can record even more episodes New @RoboPapers episodes to follow soon 6K views · 30 likes · 2 reposts · 1 replies Open on X →