This can be the Apple II of AI-driven robotic arms



Robots in factories at the moment are {powerful} and exact, however dumb as toast. 

A brand new robotic arm, developed by a staff of researchers from UC Berkeley, is supposed to vary that by offering a cheap-yet-powerful platform for AI experimentation. The staff likens their creation to the Apple II, the non-public pc that attracted hobbyists and hackers within the 1970s and ’80s, ushering in a technological revolution.

Robots and AI have developed in parallel as areas of analysis for many years. In recent times, nonetheless, AI has superior quickly when utilized to summary issues like labeling pictures or taking part in video video games. However whereas industrial robots can do issues very exactly, they require painstaking programming and can’t adapt to even the slightest adjustments. Cheaper, safer robots have emerged, however most are not designed particularly to be managed utilizing AI software program.

Join The Algorithm

Synthetic intelligence, demystified

“Robots are more and more capable of study new duties, whether or not by trial and error or by way of professional demonstration,” says Stephen McKinley, a postdoc at UC Berkeley who was concerned with creating the robotic. “And not using a low-cost platform—an Apple II-type gadget—experimentation, trial and error, and productive analysis will proceed to maneuver slowly. There may be potential for analysis to be drastically accelerated by making extra robots extra accessible.”

The brand new arm, often called Blue, prices round $5,000, and it may be managed by way of a virtual-reality headset—a method that’s proving helpful for coaching robot-controlling AI algorithms.

Blue is able to carrying comparatively heavy hundreds however can also be extraordinarily “backdrivable,” which means it should comply when pushed or pulled. This makes it secure for folks to work alongside, and permits it to be bodily proven how one can do one thing. The system supplies low-level software program for controlling the robotic and for the VR system, and it’s designed to be appropriate with any pc operating AI software program.

The undertaking comes from the lab of Pieter Abbeel, a professor at UC Berkeley who has pioneered the applying of AI to robotics (see “Innovators Beneath 35: Pieter Abbeel”). The IP for the undertaking has been licensed from UC Berkeley by a brand new firm known as Berkeley Open Arms, which is able to develop and promote the {hardware}.

It stays extraordinarily troublesome to translate machine studying from a digital surroundings to the true world. Regardless of this, tutorial researchers have made progress in making use of machine studying to robotic {hardware}, resulting in some spectacular demonstrations and a couple of business ventures.

Some canny corporations have taken discover of the development. Nvidia, a chipmaker that has ridden the AI increase by making microprocessors and software program for deep studying, not too long ago launched a lab devoted to exploring functions of AI to robots (see “This Ikea kitchen would possibly educate industrial robots to be much less dumb and extra useful”). 

Nvidia’s CEO, Jensen Huang, describes the Berkeley robotic as “very thrilling.”

Huang notes that whereas an industrial robotic could price round $50,000 to purchase, it may price many instances that to reprogram one for brand spanking new a sequence of various duties. “We have now it the improper method round,” he says. He expects large advances in robotics in years to come back due to advances in machine studying and virtual-reality simulation: “Robots and AI are actually the identical factor.”


Like it? Share with your friends!

0 Comments

Your email address will not be published. Required fields are marked *

Send this to a friend