berkay
431 posts
@BerkayAntmen
pre-training/inference @generalistai
Gen-1 ties zipties Read more about Gen-1 in our blog posts in the comments below ↓
Today marks the end of my first full week @GeneralistAI Last Monday, I was given a challenge: use our GEN-1 model to teach a robot a task of my choosing, using the same no-code platform our customers use. I picked the ball-and-vase magic trick. It was one of my favorites as a kid, and it felt like the right mix of fun and surprisingly hard. A few days later, GEN-1 pulled it off. I left Friday having watched the robot nail it 14 times in a row. What’s wild is that even 4 months ago, if you told me you could go from idea to on-robot skill in a couple of days, I probably wouldn’t have believed you. Really excited to be building with an incredible team. Can’t wait to see what week two brings 🤖
Today marks the end of my first full week @GeneralistAI Last Monday, I was given a challenge: use our GEN-1 model to teach a robot a task of my choosing, using the same no-code platform our customers use. I picked the ball-and-vase magic trick. It was one of my favorites as a kid, and it felt like the right mix of fun and surprisingly hard. A few days later, GEN-1 pulled it off. I left Friday having watched the robot nail it 14 times in a row. What’s wild is that even 4 months ago, if you told me you could go from idea to on-robot skill in a couple of days, I probably wouldn’t have believed you. Really excited to be building with an incredible team. Can’t wait to see what week two brings 🤖
GEN-1 cleans white board Read more about GEN-1 in our blog post in the comments below ↓
GEN-1 cleans white board Read more about GEN-1 in our blog post in the comments below ↓
Can a language model learn, end-to-end, what to keep in its own KV cache and what to throw away? Can it learn to forget while it learns to reason? Deep learning's central lesson: capability emerges from end-to-end optimization, not heuristics/strong inductive biases. But for efficiency, we rely heavily on hand-designed approaches. 🗑️ Introducing Neural Garbage Collection (NGC): we train a language model to jointly reason and manage its own KV cache, using reinforcement learning with outcome-based task reward alone. No SFT, no proxy objectives, no summarization in natural language. New paper with @jubayer_hamid, Emily Fox, and @noahdgoodman!
Nothing shockingly dumb?
GEN-1 plays the 🐚 shell game, trained on just 1 hr of robot data. It also generalizes to unseen objects, like @BerkayAntmen 's car keys. Physical AI models should be capable of benchmark tasks like this one. It's interesting for the all the reasons @RhodaAI calls out -- requires visual memory, and the model must track the cups from the very start, at high frame rates. Interestingly, GEN-1 appears to exhibit a degree of "active perception." It's subtle; the hands can sometimes appear to "follow" the cups, using its own movements to help attend to where it thinks the object should be. Read more about GEN-1 in our blog post in the comments below ↓
GEN-1 plays the 🐚 shell game, trained on just 1 hr of robot data. It also generalizes to unseen objects, like @BerkayAntmen 's car keys. Physical AI models should be capable of benchmark tasks like this one. It's interesting for the all the reasons @RhodaAI calls out -- requires visual memory, and the model must track the cups from the very start, at high frame rates. Interestingly, GEN-1 appears to exhibit a degree of "active perception." It's subtle; the hands can sometimes appear to "follow" the cups, using its own movements to help attend to where it thinks the object should be. Read more about GEN-1 in our blog post in the comments below ↓
GEN-1 plays the 🐚 shell game, trained on just 1 hr of robot data. It also generalizes to unseen objects, like @BerkayAntmen 's car keys. Physical AI models should be capable of benchmark tasks like this one. It's interesting for the all the reasons @RhodaAI calls out -- requires visual memory, and the model must track the cups from the very start, at high frame rates. Interestingly, GEN-1 appears to exhibit a degree of "active perception." It's subtle; the hands can sometimes appear to "follow" the cups, using its own movements to help attend to where it thinks the object should be. Read more about GEN-1 in our blog post in the comments below ↓
Pretty crazy take. If you take our undergrad TA compensation package and measure it as hourly pre-tax income, it's $89 to $102/hr. How is that 'unfair' comp for a 19-year old with no degree and no job experience, when it's 4x+ pretty much every other university?🤔 A good chunk of that is the tuition remission OP mentions, which 80+% goes to students who don't qualify for a penny of financial aid. As I'm quoted in the article, that means we're using other kids' tuition and state funds to subsidize tuition for those wealthier than average, which is exactly ... the opposite of what a government institution like the UC should be doing. I question the negotiation process between public employers and labor unions. The people who run negotiations at the UC Office of the President aren't negotiating with their own money, but rather the money of the taxpayer and the average tuition payer, and the deals they agree to are in my opinion a slap in the face to the taxpayer. P.S. here's how our undergrad TA comp compares to other universities: docs.google.com/spreadsheets/d… (see column G for total comp pre-tax equivalent)
