Dara Akbar Ladjevardian discusses Liquid AI, an MIT startup building neural networks for robots and safety-critical systems. The company focuses on physical intelligence and edge computing, processing data directly on devices like satellites instead of sending it to the cloud.
What is liquid AI and why is it so special? We started liquid AI as a foundation model company that we invented liquid neural networks. This was a class of neural networks that allowed us to build AI systems for robotic systems that work in safety critical applications.
We realized that there is the nugget of a new idea looking towards what is known about the neurons of small species and using that as the basis of a new approach for machine learning. The for AI.
Physical intelligence and AI in reality is much farther behind the AI that we have in papers, in research and in digital reality. The intelligence needs to live on the body of the physical agent. For example satellites. So on satellite whenever you want to process information you have to send like these big pictures of earth that you're taking. You have to send them back on earth, get the computation and then bring them back up. If you can do it on the edge that would be a massive.
The AI Agent Getting More Supplier Data in 5 Weeks Than a Team Did in 2 Years
