Adaption Labs is building AI that adapts to you
Adaption Labs, started by Sara Hooker, is focused on building adaptive AI.
Their goal? Build adaptive AI that actually learns and changes based on how people use it, not just what it was trained on.
How is it different from traditional AI?
Instead of static models, Adaption Labs focuses on three things: real-time data that adapts to the task, intelligence that adjusts for problem difficulty, and models that change behavior when given feedback.
They also skip traditional retraining—using a method called gradient-free learning to make their AI flexible without all the extra computing.
Hooker on the future of AI and India's role
Hooker points out the old way of making bigger and bigger AI models isn't working anymore. She's betting on new techniques like reinforcement learning and test-time adaptation—which are faster and cheaper.
Plus, their tech can adapt in real time to different tasks or domains, making it potentially useful for a range of work, from simple questions to more complex tasks.
Hooker also discusses India's role in providing infrastructure and support for AI research.