Jacob Van Meter

wow @jacobvm04 and i just took first place on one of robotics’ hardest problems. machines knowing where they are in the world. slam dunk.

when LaMAria was introduced a year ago, Meta's closed-source Aria SLAM was far ahead of academic systems. one year later, we've surpassed it in 3/5 categories: medium, long, and low light. here's what closing that gap looks like:

interpretability might be the most important scientific problem in the world right now, as I don't think we can solve alignment if we don't know what we're building. I think I have good news about it. I co-authored early work in mechanistic interpretability at OpenAI and came away pessimistic, because neural networks appeared to fold into arbitrary tangled knots. years later, I started openai's first large-scale automated interpretability project, finding about 2% of neurons legible with the primitive tools we had with modern agentic tools and a much closer look, the picture looks totally different. the vast majority in models up to trillions of parameters are legible, and their circuits beautiful. many famous examples of tangled neurons turned out to be misunderstood. The roadblocks I expected mostly aren't there. I now think ambitious interpretability will be solved by the time we need it most, and possibly quite soon, as automated scientists keep improving. it may be a bitter lesson for interpretability: the simple thing, at absurd scale, works already-clean.goodfire.com/

This is the real gpt 3 moment for robotics by the way. Astra has essentially perfect vision. What’s left is latency (gpt 3.5 moment). Once that is solved, we can just rl on a bunch of robotics tasks.

it's crazy how robust rl policies can be if you design the training to scale and use proper dr/sensor noise augmentation. trained this 2d lidar point nav policy locally at 1m steps/s and it is very robust