Technology

OpenAI wants to create an all powerful AI. But in a good way

One Saturday last month, five menage 19 to 26 years confidently emerged from a cloud of purple smoke at a converted auto show in San Francisco. They sat down at a row of computer keyboards, loudly greeted by a crowd of several hundred people. After ninety minutes of intense mouse clicks, the smiles of the five turned dull and the applause was comforting. The OG team, the world champions of the most profitable competitive Dota 2 game in the world, lost two games in a row to the team of bots with artificial intelligence.

This result was remarkable becauseComplicated video games are mathematically more complicated than board games like chess or go. Still, leaning against the wall behind the scenes, Sam Altman, the general director of the OpenAI Research Institute, who developed these bots, felt a festive relief.

Artificial Intelligence vs. Doters

"We were all very nervous that morning - I thoughtthat the odds are 60 over 40, says Altman. He became the CEO of OpenAI in March, after leaving the post of president of the influential incubator startups YCombinator. To succeed in the new position, Altman needed bots that not only beat people in video games - he needed bots that are better than people in everything.

The stated mission of OpenAI is toensure that all humanity will benefit from any future AI capable of surpassing “people in the most economically viable work”. Such technologies - general artificial intelligence, “strong” artificial intelligence, or AII (AGI) - are not even looming on the horizon, but OpenAI claims that both it and others are rapidly progressing. The organization has shown that it can conduct research at the level of the best in the world. And she got the accusations of fanning the hype on the topic of AI, as well as of fixing on the AIS and potential threats of artificial intelligence.

According to Altman's plan, an OpenAI study - andprovocations - should only pick up the pace. Being previously the chairman of the board of directors of this organization, he took the post of CEO after helping to transfer most of the nonprofit staff to a new commercial company, hoping to attract investors with the billions needed to shape the fate of AI and humanity. Altman says that large technology laboratories at Alphabet and other companies should be under pressure from colleagues who are not seeking to increase their shareholder value. “I don’t want to be in a world in which OII will create one company, capture all its value and make all the decisions,” he says.

At the MIT event at the end of 2014, the generalTesla director Ilon Mask described the AI ​​research as an attempt to “summon a demon”. In the summer of 2015, he spoke with Altman and several others at dinner and proposed to create a research laboratory, independent of the technology segment, which would send the AI ​​in a positive direction. In the same year, OpenAI was created, in which Altman and Musk found themselves. In early 2018, Musk left the board of directors, citing potential conflicts with his other roles.

In its short life, OpenAI has positionedyourself as a serious place for AI research. Ilya Sutskever, co-founder of the organization, who resigned from an AI research management position at Google, is overseen by staff that includes former people from Google, Facebook, Microsoft and Intel. Their work on the topic of robotics and machine learning appeared at the most prestigious conferences. The victory over the OG team in Dota 2 earned them respect in the field of AI and gaming.

Turning OpenAI into a commercial corporation wascaused by the feeling that in order to keep up with such giants as Alphabet, you will need access to ever-growing computing resources. In 2015, OpenAI announced that it had $ 1 billion from Altman, Mask, LinkedIn co-founder Reid Hoffman, Facebook early investor Peter Thiel and Amazon. Now Altman says that one billion will not be enough. “The amount of money we need for the success of a mission is much more than I thought,” he says.

IRS applications show that in 2017, whenOpenAI showed its first bot dotera, she spent $ 8 million on cloud computing. Since then, her expenses have probably increased significantly. In 2018, OpenAI announced that the predecessor of the system that won the OG connected more than 120,000 processors leased from the Google cloud division for weeks. The champion-winning version has been training for 10 months, equivalent to 45,000 years of Dota 2 playing against itself. How much did it cost? Greg Brockman, CTO OpenAI says that “millions of dollars” went to the project, but did not disclose details.

Altman is not sure that OpenAI will continuerely on competitors' cloud services - it is still open to purchase or even to develop equipment for artificial intelligence. The organization is closely following the new chips being developed by Google and a lot of startups to squeeze out more machine learning algorithms.

To raise funds required for accessto the future equipment, Altman tried to sell the idea to investors according to a scheme that was wild even for Silicon Valley. Invest in OpenAI, wait, and the company will return a hundredfold investment to you - as soon as it invents robots that surpass people in most economically valuable tasks.

Altman says meeting these expectationsIt was "the most interesting experience of fundraising in my life - it is not suitable for any other model." According to him, the greatest interest is shown by rich people interested in AI. Hoffman and venture capital firm Khosla Ventures invested in a new commercial OpenAI, but did not respond to requests for comment. No one says when to expect a refund, but OpenAI bets are not for the impatient. Companies are informed that the return time may exceed the industry standard (ten years). “We immediately tell them that you will not receive funds in the next 10 years,” says Altman.

Despite all efforts, OpenAI is often criticized.industry leaders. In February, the company revealed details about language processing software that was capable of generating surprisingly lively text. Some immediately called the invention “dangerous” and expressed concern that it could be used to harm, to litter social networks.

Some prominent people in the field of AI researchwere annoyed, including the main AI on Facebook, Jan Lekun. In his posts on Facebook, he defended an open publication of research on the topic of AI and joked that people should stop having children, because one day they would publish fake news.

For some, this episode emphasized that the missionOpenAI is not much different from other corporate or academic laboratories. “They do the same kind of research, like the others, but they want to attract billions of dollars,” said Zachary Lipton, a professor of machine learning at Carnegie Mellon University. "The only way to do this is to wry a little soul."

Altman admits that OpenAI may have scoredanxiety is too early, but says it is better than being late. “The technology industry has not done a very good job trying to be proactive against abuses,” he says. Self-censorship is also required by OpenAI itself.

After the defeat of the OG last month in order toto express sympathy, OpenAI decided to conduct a second experiment, designed to demonstrate the auspicious side of superhuman AI. Dota 2 experts and several newbies played in teams with bots.

Unlucky softwareto become partners of the newbies on the team, for the most part showed superhuman indifference to help the beginning player. Instead, it focused on winning, following instincts, honed over months of expensive training.

No one knows how to make a focused AI.more flexible, or what components may be needed to make OII more than technological fantasy. Even Altman is scared of the scale of the problem. “I have days when I believe that all this will happen, and days when it all seems like a impossible dream,” he says.

What do you think will work out for him? Let's discuss in our chat in Telegram.