Robotics startup Generalist AI has unveiled GEN-1.5, an AI model that teaches robots new tasks from a single demonstration. A 3- to 12-second demo gets loaded into the model's context window as a "physical prompt," essentially the model's short-term memory. After that, the robot performs the task with no training at all. Across ten tests like opening a jar or pulling money from a wallet, the company reports an average success rate of 59 percent. With ten training steps on five minutes of data, that number jumped to 83 percent.
The model can chain two prompts together into longer sequences, use demos from simulation, and partly imitate human hand movements. According to Generalist, these abilities emerged on their own during more than eight months of pretraining on interaction data. They were never explicitly trained.
Other research teams have shown similar in-context learning before, but only for a handful of task types. Generalist claims to be the first to make it work across a wide range of tasks. The tasks shown are simple and short, though, and all results come from the company itself. None have been independently verified.
AI News Without the Hype – Curated by Humans
Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.
Subscribe now