The tension between Google and its AI team DeepMind has always been fascinating. To recap, DeepMind was founded in 2010 and has some of the best ai researchers in the world, publishing a steady stream of in-depth academic papers and Nature front pages. Google, meanwhile, bought DeepMind in 2014 and financed its huge losses, but it’s really trying to Squeeze Some Money out of all this brain trust.

That’s why the Wall Street Journal’s latest story on the two companies is interesting. Pammy Olson reports that Google has ended year-long negotiations between the two companies and rejected DeepMind’s request for more independence.

One idea reportedly on the table is to give DeepMind non-profit status.

Olson said DeepMind told employees “late last month” that negotiations were over. According to people familiar with the plans, one of DeepMind’s founders’ apparent recommendations was for the company to have the same legal structure as a nonprofit, arguing that “the powerful AI they are working on should not be controlled by a corporate entity.” But Google disagreed and told DeepMind that it did not make sense given the amount of money it had invested in DeepMind.

The conflict is not surprising. Google executives have repeatedly said that the company’s future lies in artificial intelligence, and numerous news reports suggest that Google has been pressuring DeepMind to commercialise its work. This has led DeepMind to use its research on projects to improve Android’s battery life and reduce energy costs in its data centres, but the economics of these efforts are unclear. Meanwhile, the UK company’s losses mounted – hitting a high of £477m on the UK Companies Registry. In its most recent public filing for 2019, about $660 million. If Google wants value for money, it cannot afford to give DeepMind anything resembling non-profit status.

In addition to financial pressures, another bone of contention between the two companies appears to be ethical oversight. One much-touted element of Google’s acquisition of DeepMind was its promise to set up an ethics committee to ensure that its technology was always deployed fairly. But the precise nature and scope of the commission, including who serves on it, has never been clear. According to a 2019 report from The Economist, the council even has ownership of any AI created by DeepMind — the term for ai that meets or exceeds human capabilities in a wide range of tasks.

The status of the committee was not mentioned in the WSJ story, but Mr Olson noted that DeepMind’s future work would now be overseen by a separate ethics committee “staffed largely by senior Google executives”. That’s the Advanced Technology Review Board, or ATRC, Olson noted in a tweet. This is said to be Google’s “supreme censorship committee”.

In a statement to The Verge, a DeepMind spokesperson said. Over the years, of course, we have discussed and explored different structures within the Alphabet Group to find the best way to support our long-term research mission to address intelligence. We are proud to be able to accomplish this incredible mission while continuing to have operational autonomy and the full support of Alphabet.”

The original link: www.theverge.com/2021/5/21/2…