robotics/ai research | unc dropout | prev @plasticlabs
San Francisco, CA
Joined April 2023
- Tweets60
- Following1.3K
- Followers322
- Likes1.2K
Pinned Tweet
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:
wow @jacobvm04 and i just took first place on one of robotics’ hardest problems.
machines knowing where they are in the world.
slam dunk.
ALT sota
faith is all you need
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/
Jacob Van Meter retweeted
i asked claude(.)ai about a coffeeshop. it sent my full name, my employer, and my bank security answers to a malicious attacker.
no code execution or weird MCPs. writeup below.
Learning from human interventions is underrated, it’s crazy how small of a dataset you need if it covers the right distribution
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
Jacob Van Meter retweeted
MacOS is the future
wrote a metal batched rendering backend for genesis (@gs_ai_) after spending too much money on gpu rentals to mess around with robotics sims, now I can finally run visual rl experiments on my laptop at >40k fps!
wrote a metal batched rendering backend for genesis (@gs_ai_) after spending too much money on gpu rentals to mess around with robotics sims, now I can finally run visual rl experiments on my laptop at >40k fps!
