Some startups are trying to create another form of AI than deep learning, to minimize the amount of training, data and server power needed. Samsung Next, the South Korean company’s venture capital unit, just launched the Q Fund to jumpstart this idea by funding companies focusing on new ways of developing artificial intelligence. One of Q Fund’s first investments is Vicarious, a startup that wants to give machines “imagination” and is inspired by biology to make machines learn more quickly.
Quartz reports that Samsung Next director Ajay Singh noted that, “deep learning has its own baggage,” including “its propensity for bias and the amount of data needed.” Last July, Q Fund “participated in a funding round of Vicarious led by Khosla Ventures,” and “also invested in automation-robotics startup Covariant.ai, previously called Embodied Intelligence,” which was founded by noted AI researcher Pieter Abbeel.
The latter startup aims to use virtual reality and simulations “to reduce the amount of time robots need adjusting to their jobs in the real world.” To vet future investments, Q Fund relies on “a deep stable of industry and academic experts from MIT, Carnegie Mellon, Princeton, and Toronto’s Vector Institute.”
“We’re okay if it’s unproven,” said Samsung Next’s principal Vin Tang. “As long as the upside is big enough.”
ZDNet reports that the Q Fund “will seek to work exclusively with startups developing software and services that tackle the biggest challenges of artificial intelligence,” focusing on “the areas of learning in simulation, scene understanding, intuitive physics, program learning programs, auto ml, robot control, human computer interaction and meta learning.”
Singh noted that the Q Fund wants “to invest in the people and teams who will try new approaches to lay the groundwork for what AI will be,” which is why it “will prioritize technical diligence over revenue models.” The investment in Covariant.ai, he said, “highlights Q Fund’s focus on startups replacing conventional algorithms, data structures, and general purpose computer programs with learned systems.”