One of the most ambitious private artificial intelligence research organisations has turned itself into a profitmaking enterprise in an effort to draw in more capital and talent — though with a cap on how much money its backers can make.
The revamp at OpenAI signals a rethink at the San Francisco-based organisation on the way to what it hopes will be full human-like computer intelligence, known as artificial general intelligence.
The brainchild of some of the tech world’s most radical thinkers, including Elon Musk and Peter Thiel, OpenAI was set up in 2015 to lead the way in developing “safe” AI. Mr Musk began warning loudly at the time that new controls were needed to prevent intelligent computers from one day destroying human civilisation.
The founder of Tesla and SpaceX has since distanced himself from the organisation, leaving its board a year ago partly to prevent any conflict of interest with his electric car company’s own work on AI.
On Monday OpenAI said it would in future conduct much of its work through a new for-profit arm. It put the new commercial remit down to a need to raise “billions of dollars” to pursue its goals, as well as being able to offer greater incentives to top researchers at a time when tech giants such as Google and Facebook are paying huge amounts.
Despite the new focus on making money, OpenAI said it would still be governed by its original remit to produce safe AI and that it would limit the amount workers or investors could make, with the excess going into funding its work.
Investors in the first round can earn up to 100 times their initial investment, with future rounds expected to offer lower returns. The organisation said that, if successful, “we expect to generate orders of magnitude more value than we’d owe to people who invest in or work at” OpenAI.
To keep the focus on its original goals, OpenAI’s work will be governed by the board overseeing its non-profit arm, only a minority of whom will be allowed to have a financial interest in the group.
It also said that, for now, it would continue to look “exclusively” at research on new forms of AI technology, and that it would shift its focus towards commercial ventures “only once we’ve created safe AGI”. When that date will arrive is a topic of heated debate in AI circles, with many experts putting it decades in the future.
The group put its need for capital down to the massive computing resources needed to run its data-crunching algorithms, as well as a desire to build its own AI supercomputers. In a demonstration of how sheer computing brawn can bring big advances, OpenAI last month demonstrated a language-producing system it had built that can construct coherent-sounding text from any prompt. The system works by analysing mountains of text and then guessing which word is most likely to come next in any situation, turning writing into a statistical guessing game.
The group on Monday named Sam Altman, who stepped down at the end of last week as president of the tech incubator YCombinator, as its first chief executive. Thanks to early successes such as Airbnb, Dropbox and Stripe, YCombinator has become the most prominent of the organisations that try to identify the next big tech companies at a time when their founders are first formulating their ideas.
OpenAI’s board includes Reid Hoffman, the founder of LinkedIn, who it named last year as one of its biggest backers when Mr Musk stepped back. Forbes magazine put Mr Hoffman’s wealth last year at $1.8bn, and said he had already given away $1.5bn.
Copyright The Financial Times Limited . All rights reserved. Please don't copy articles from FT.com and redistribute by email or post to the web.