The new Uber Al Labs has been launched. It’s a research group aimed at all-round improvement, from their self-drive car navigation to the route by which your ordered food is delivered. Uber has acquired an entity known as Geometric Intelligence for an undisclosed sum. It will supply the staff of this start-up for the new Lab.
Geometric Intelligence was established in October 2014. The founding members were an NYU graduate, Douglas Bernis, with a PhD in neurolinguistics and three professors. Gary Marcus is a cognitive scientist from NYU and will be head of the new Al Lab as director. Zoubin Ghahramani is a Cambridge professor in machine learning and will serve as a co-director. Kenneth Stanley is a professor of computer science at the University of Central Florida.
Most of the 15 team members in the start-up will relocate to San Francisco. The academics will retain their affiliations with associated universities, although it is understood, certain members will be working at Uber Lab on a full-time basis.
Uber is currently operating a fleet of self-drive cars in Pittsburgh, offering rides to selected customers since September this year. The announcement of the new Lab comes at an opportune time for Uber, with an important regeneration underway for their app. Reportedly, an Al Lab spokesperson said the Lab will be confronting Uber-related issues that include the self-drive cars piloting!
In its relatively short term of operation, the start-up Al Lab has directed its attention towards the “sparse data” issues. It involves determining how an artificial intelligence can be created, capable of recognising objects and situations, quickly. The challenge relates to machine learning, the volume of input data and how recognition can be achieved with less input than is currently available today!
Geometric Intelligence developed software, “Xprop” which is said to require less learning data input than the presently recognised form of machine learning software. It is known as “deep learning” it is applied in the learning of a new visual task. More detailed information is available from MIT Technology Review.
Researchers for the start-up recently became co-authors of a study that was focused on deep generator network. It is a medium that creates the self-drive related images and in turn, shows how individual neurons within the network affect the understanding and interpretation of the complete system. It can be surmised by the acquisition of Geometric Intelligence that Uber is considering the development of commanding speech-image-recognition software, like Microsoft and Google. However, it could be manufactured with fewer data demanding algorithms!
Gary Marcus told a technical related audience in May 2016 that “We live in this era of big data, and there’s this idea that we can just throw more data at the problem, but for some problems, there’s just not enough data.”
Jeff Holden, chief product officer for Uber, made the point that although there has been some significant progress in machine learning, “we are still very much in the early innings of machine intelligence. With all its complexity and uncertainty, negotiating the real world is a high-order intelligence problem,”
He also wrote in a blog post, “With all its complexity and uncertainty, negotiating the real world is a high-order intelligence problem. “It manifests in myriad ways, from determining an optimal route to computing when your car or UberEATS order will arrive in matching riders for UberPOOL. It extends to teaching a self-driven machine to safely and autonomously navigate the world, whether a car on the roads or an aircraft through busy airspace or new types of robotic devices.”
Geometric Intelligence has become the latest acquisition of Uber, in the continued expansion of their business. They are taking advantage of the $ billions capital raised, regardless of a potentially uncertain foundation. This year, also saw Uber buying Otto, the self-drive truck company, which has recently in Colorado, completed the first self-directed shipment!