Eka Robotics, a startup founded by MIT professor Pulkit Agrawal and former Google DeepMind researcher Tuomas Haarnoja, released a physical AI model that has experts raving. It is the company's first major breakthrough since its founding in 2022.
Eka reported a 95% success rate in picking up objects, measured on a test of 1,500 items weighing up to 1 kg per hour, compared to about 600 for a human. That compares with previous benchmarks in the industry.
The model is built on a vision-force-action architecture and targets complex dexterous manipulation tasks. Availability begins with industrial pilots, initially for customers in pick-and-place and wire manipulation.
"They are already exceeding superhuman performance for certain tasks, and so we think over the next few months there will be superhuman performance in many dexterous manipulation tasks," said Raghu Raguram, the a16z general partner who co-led the fundraising round. Eka’s capabilities are shocking veteran robot builders.
The announcement follows a $150 million Series A round led by venture funds a16z and F-Prime. Eka’s results suggest critics of generative AI robotics models have a point, as the company is pushing through reinforcement learning.
Eka did not say how they achieved their unique robotic hand, and the open question is how generalizable these techniques can be to different tasks. The company is working in secretive pilots with a number of industrial customers.
Source: techcrunch
