Tech titans create company to develop AI
New York - Elon Musk, Peter Thiel and other technology entrepreneurs are betting that talented researchers, provided enough freedom and money, can develop artificial intelligence systems as advanced as those being built by the sprawling teams at Google, Facebook and Microsoft.
Along the way, they’d like to save humanity from oblivion.
The pair are among the backers of OpenAI, a nonprofit company introduced Friday that will research novel artificial intelligence systems and share its findings. Musk, chief executive officer of Tesla Motors Inc. and Space Exploration Technologies Corporation and Sam Altman, president of the Y Combinator, will serve as co-chairman. The nonprofit has received financial backing from Musk, Thiel, co-founder of PayPal Holdings and Palantir Technologies, Reid Hoffman and others as well as companies including Amazon Web Services and Infosys.
The group’s backers have committed “significant” amounts of money to funding the project, Musk said in an interview. “Think of it as at least a billion.”
In recent years the field of artificial intelligence has shifted from being an obscure, dead-end backwater of computer science to one of the defining technologies of the time. Faster computers, the availability of large data sets, and corporate sponsorship have developed the technology to a point where it powers Google’s web search systems, helps Facebook understand pictures, lets Tesla’s cars drive themselves autonomously on highways, and allowed IBM to beat expert humans at the game show Jeopardy!
That development has caused as much trepidation as it has optimism. Musk, in autumn 2014, described the development of AI as being like “summoning the demon.” With OpenAI, Musk said the idea is: “if you’re going to summon anything, make sure it’s good.”
“The goal of OpenAI is really somewhat straightforward, it’s what set of actions can we take that increase the probability of the future being better,” Musk said. “We certainly don’t want to have any negative surprises on this front.”
OpenAI’s chief technology officer is Greg Brockman, formerly the CTO of Stripe, a startup valued in excess of $1 billion. Its research director is lauded AI researcher Ilya Sutskever, formerly with Google. At Google, his work included research into the technology that became Smart Reply, the auto email-writing feature, as well as systems that can learn to write their own algorithms. This year, MIT Technology Review named him one of their 35 innovators under 35.
The organisation has attracted other talented researchers, whose past work ranges from developing robots that can learn to perform tasks based on human demonstrations, to software that can improve its own code to solve new problems.
The idea, is that “real breakthroughs in research are serendipitous,” Brockman said, so the best way to develop powerful artificial intelligence systems is to take a group of accomplished people and give them the latitude to focus on solving novel problems. A potential criticism is that the lack of a defined commercial motivation can make it difficult to home in on a precise area for research.
“This collection of people is stunning,” said Pieter Abbeel, a professor at the University of California at Berkeley and an adviser to the company.
The best measure of the organisation will be to see the reaction to its research and the ideas it fosters, Abbeel said. “I expect it’s going to be impressive and surprising,” he said.
Other advisers to OpenAI include Yoshua Bengio, one of the founding figures of a powerful form of artificial intelligence called Deep Learning, and Alan Kay, a lauded American computer scientist.
Wojciech Zaremba, one of OpenAI’s researchers, said the group’s openness can help serve as a balance to the narrower interests of researchers at commercial ventures.
“Here Elon is saying, in a philanthropic way, ‘I have money, I want to solve AI for the good of humanity.”’Bloomberg