Georgia Tech prof. @animesh_garg says sub-$1,000 robots are about to do for robotics what the Apple I did for personal computing: "Mac Mini analog, the use case is programmable. Giving people standardized hardware at a cost that is roughly less than a phone makes it available to developers." "Robotics is fundamentally going through what I call the renaissance that computers went through in late 1970s when Apple 1 came out. Sub-$1,000 robots which are relatively stable, suddenly make technology go brr." "Just in the four days since @huggingface Microduck has been released, people have done so much crazy stuff with it, from reinforcement learning at scale to building simulators where these things are grabbing pins." "I'm very certain that adding two arms to that interface at the $1,000 price point in six months will probably be very likely." @GeorgiaTech

Georgia Tech prof. @animesh_garg predicts humanoid hardware will get 10X cheaper in just 2-3 years: "A small or medium scale humanoid like G1 is available at BOM of under 10K. If you can build it at scale, at 500 units or above, you can get it there." "Number of US entities have attempted this. I do expect that by next summer we should have US-made or mostly US-made developer platform humanoids available to us at $25,000 give or take price points." "The price for hardware will drop by over 10X within a two to three-year time horizon. So two years from now, I absolutely anticipate a toy level humanoid to be available at 5K, and an actually useful humanoid to be available at somewhere between 10 to 15K." @GeorgiaTech

Georgia Tech prof. @animesh_garg predicts the next 2 years of robotics will be "the return of the RL" as the field shifts from collecting data to learning from it: "Pillar one: how do you represent the environment? Should the input be objects? Should the input be completely latent? Should the input be somehow processed with a VLM or LLM? There is debate in the field." "Pillar two is fundamentally what people now call models of the world. World models or critic models or reward models, some sort of world models that allow a very good prior on what to do." "Robotics is classically a reinforcement learning problem, so what we call post-training or recursive improvement or continued learning, those are exactly the problems that will show up in robotics in maybe a year or a year and a half." "In the same way we saw the jump from early GPTs to GPT-3 and ChatGPT, but ChatGPT by itself did not end up becoming enterprise useful. It took another two or three years from there to becoming the Claude moment." "The next two years of robotics will basically be, now that we have the data, what do we do in RL? As a lot of people call it fondly, the return of the RL." @GeorgiaTech