Artificial intelligence now it is possible Advanced math solvingperform complex reasoningand even use personal computersbut today’s algorithms can still learn a thing or two from microscopic worms.
Liquid AIA startup spun out of MIT, today is unveiling several new AI models based on a new type of “liquid” neural network that has the potential to be more efficient, less power-hungry, and more transparent than with models underpinning everything from chatbots to image generators to facial recognition systems.
Liquid AI’s new models include one for detecting fraud in financial transactions, another for controlling self-driving cars, and a third for analyzing genetic data. The company introduced new models it is licensing to outside companies at an event held at MIT today. The company has received funding from investors including Samsung and Shopify, both of which are testing its technology.
“We are expanding,” said Ramin Hasanico-founder and CEO of Liquid AI, who co-invented liquid networking while a graduate student at MIT. Hasani’s research inspired C. eleganta millimeter-long worm commonly found in soil or decaying vegetation. The worm is one of the few organisms to have its entire nervous system mapped, and it is capable of remarkably complex behavior despite having only a few hundred neurons. “It used to be just a science project, but this technology is fully commercialized and ready to bring value to businesses,” Hasani said.
Inside a typical neural network, the properties of each simulated neuron are determined by a static value or “weight” that influences its activation. In liquid neural networkThe behavior of each neuron is governed by an equation that predicts its behavior over time, and the network solves a series of coupled equations as the network operates. This design makes the network more efficient and flexible, allowing it to learn even after training, unlike conventional neural networks. Fluid neural networks can also be tested in ways that current models cannot, as their behavior can essentially be rewinded to see what output it produces.
In 2020, researchers showed that such a network with just 19 neurons and 253 synapses, tiny by modern standards, could control a simulated self-driving car . While conventional neural networks can only analyze image data at static intervals, fluid networks capture how image information changes over time very effectively. In 2022, the founders of Liquid AI Found a shortcut that makes the mathematical labor required for liquid neural networks feasible for practical use.
