Inside DeepMind’s secret plot to break away from Google

[ad_1]

For a while, some DeepMind employees called it “Watermelon”. Later, the executives called him “Mario”. Both codenames meant the same thing: a secret plan to break up with parent company Google.

DeepMind feared that Google might someday abuse its technology, and executives have been working for years to alienate the AI ​​company from its owner, nine current and former employees who were directly familiar with the plans said.

This included plans to pursue an independent legal status that would alienate the group’s work from Google, the people said, who asked not to be identified by discussing private matters.

One of DeepMind’s main strains was that it was selling the business to people it didn’t trust, a former employee said. “Everything that has happened since then has been about them questioning that decision,” the person added.

Efforts to separate DeepMind from Google ended in April without a deal, the Wall Street Journal reported. Years of negotiations, coupled with the recent upheavals within Google’s AI division, raise doubts that the search giant can retain control over a technology so crucial to its future.

“DeepMind’s close collaboration with Google and Alphabet since the acquisition was tremendously successful with their support, we have produced research breakthroughs that have transformed the field of artificial intelligence and are now unlocking some of the biggest questions in science. “a DeepMind spokesperson said in a statement. “Over the years, we have obviously discussed and explored different facilities within the Alphabet group to find the optimal way to support our long-term research mission. We couldn’t be more proud to accomplish this incredible mission, while continuing. to have both operational autonomy and full support from Alphabet. “

A stands for Literacy

When Google acquired DeepMind in 2014, the deal was seen as a win for everyone. Google got a major AI research organization, and DeepMind, in London, got financial support for its quest to create an AI that can learn different tasks like humans do, known as general artificial intelligence. .

But tensions soon emerged. Some employees described a cultural conflict between researchers who considered themselves academics first and the sometimes bloated bureaucracy of Google’s colossal business. Others said staff were immediately concerned about putting DeepMind’s work under the control of a tech giant. For a while, some employees were encouraged to communicate using encrypted messaging apps for fear of Google spying on their work.

At one point, DeepMind executives found that work published by Google’s internal AI research group resembled some of DeepMind’s codebases without citations, said one person familiar with the situation. “This pissed off Demis,” added the person, referring to Demis Hassabis, CEO of DeepMind. “This was one of the reasons DeepMind started to become more protective of their code.”

After Google restructured itself as Alphabet in 2015 to give riskier projects more freedom, DeepMind leadership began pursuing a new separate division status under Alphabet, with its own profit and loss account, The Information reported. .

DeepMind already enjoyed a high level of operational independence within Alphabet, but the group also wanted legal autonomy. And he worried about the misuse of his technology, particularly if DeepMind ever caught up with AGI.

Internally, people have started referring to the plan to gain more autonomy as “Watermelon,” two former employees said. The project was then formally named “Mario” among the DeepMind leadership, these folks said.

“Their perspective is that their technology would be too powerful to be held by a private company, so it must be housed in some other legal entity separate from the interest of the shareholders,” said a former employee who was close to the Alphabet negotiations. “They framed it as ‘this is better for society'”.

Demis Hassabis, CEO and co-founder of DeepMind Technologies, at the DLD conference in Munich. Tobias Hase / Picture Alliance via Getty Images

In 2017, during a corporate retreat at Macdonald Aviemore Resort in Scotland, DeepMind leadership revealed to employees its plan to separate from Google, two people in attendance said.

At the time, the leadership internally said the company planned to become a “global interest company,” said three people familiar with the matter. The title, not an official legal status, was to reflect the worldwide ramifications DeepMind believed its technology would have.

Subsequently, in negotiations with Google, DeepMind pursued the status of a limited liability company, a shareless corporate structure that is sometimes used by nonprofits. The deal was that Alphabet would continue to finance the company and obtain an exclusive license for its technology, two people involved in the discussions said. There was a condition: Alphabet could not exceed certain ethical limits, such as using DeepMind technology for military weapons or surveillance.

In 2019, DeepMind registered a new company called DeepMind Labs Limited, as well as a new holding company, documents filed with the UK Companies House show. This was done in anticipation of a separation from Google, two former employees involved in those registrations said.

Negotiations with Google have gone through ups and downs over the years, but gained new momentum in 2020, one person said. A senior team within DeepMind has begun holding meetings with external lawyers and Google to clarify the details of what this new theoretical training could mean for the relationship between the two companies, including details such as whether they would be sharing a code base. internal performance metrics and software expenses, two people said.

Sergey Brin (left) and Larry Page have spent the last few years moving away from public view. Chris Hondros / Getty Images DeepMind and Google’s leadership disagree on the ethics of AI

From the start, DeepMind was thinking about potential ethical dilemmas arising from its deal with Google. Prior to the close of the 2014 acquisition, both companies signed an “Ethics and Security Review Agreement” that would prevent Google from taking control of DeepMind’s technology, The Economist reported in 2019. Part of the deal it included the creation of an ethics committee that would oversee the research.

Despite years of internal discussions about who should sit on this board and vague promises to the press, this group “never existed, never convened and never resolved any ethical issues,” said a former employee close to those discussions. . A spokesperson for DeepMind declined to comment.

DeepMind pursued a different idea: An independent review committee to call if it were to separate from Google, three people familiar with the plans said. The board would be made up of executives from Google and DeepMind, as well as third parties. Former US President Barack Obama was someone DeepMind wanted to approach for this council, said one person who saw a shortlist.

DeepMind also created an ethics charter that included a ban on using its technology for military weapons or surveillance, as well as a rule that its technology should be used for purposes that benefit society. In 2017, DeepMind started an AI ethics research-focused unit made up of external employees and researchers. Its stated goal was “to pave the way for truly beneficial and accountable AI”.

A few months later, a controversial contract between Google and the Pentagon was disclosed, causing an internal turmoil in which employees accused Google of entering the “war business”.

Google’s Pentagon contract, known as Project Maven, “sets alarm bells ringing” within DeepMind, a former employee said. Google later published a set of principles for governing its AI work, guidelines that were similar to the ethics charter that DeepMind had already established internally, ranking some of DeepMind’s senior executives, two former employees said.

Sundar Pichai, CEO of Google. Stephen Lam / Reuters Who will control AI?

In April, Hassabis told employees in an all-round meeting that negotiations to separate from Google had ended. DeepMind would keep its existing status within Alphabet. DeepMind’s future work would be overseen by Google’s Advanced Technology Review Council, which includes two DeepMind executives, Google AI chief Jeff Dean and legal SVP Kent Walker.

But the group’s years-long battle for more independence raises questions about its future within Google.

Google’s commitment to AI research was also called into question, after the company forced it to lay off two of its oldest AI ethics researchers. This led to a backlash from the industry and sowed doubts that it could allow for truly independent research.

Ali Alkhatib, a member of the Center for Applied Data Ethics, told Insider that greater public responsibility is “desperately needed” to regulate the pursuit of artificial intelligence by big tech companies.

For Google, its DeepMind investment could begin to pay off. Late last year, DeepMind announced a breakthrough to help scientists better understand the behavior of microscopic proteins, which has the potential to revolutionize drug discovery.

As for DeepMind, Hassabis maintains the belief that AI technology shouldn’t be controlled by a single company. Speaking at the Responsible AI Forum in Tortoise in June, he proposed a “world institute” of AI. Such a body could be under the jurisdiction of the United Nations, Hassabis theorized, and could be filled with the best researchers in the field.

“It’s much stronger if you lead by example,” he told the audience, “and I hope DeepMind can be part of that role model for the industry.”

Do you have anything else to share? Do you have a suggestion? Contact reporter Hugh Langley at [email protected] or on the encrypted messaging apps Signal and Telegram at +16282281836.

Contact reporter Martin Coulter at [email protected] or on the Signal encrypted messaging app at +447801985586. Contact them using a non-business device.

[ad_2]