Google has a responsibility to protect DeepMind data
Google’s £400m acquisition of the UK artificial intelligence research company DeepMind in 2014 was testimony to the quality of British scientific research.
Furthermore, the insistence of the three UK co-founders that their company would not move to California was seen as evidence of London’s potential to become a successful centre for technology innovation.
Four years later, the future of the UK capital’s tech aspirations and of DeepMind’s centre of gravity look a lot less certain.
DeepMind’s announcement last week that it would transfer control of its health unit to a new Google Health division in California has raised questions about data privacy.
The health unit has access to the records of 1.6m patients of Britain’s National Health Service.
After four years of relative operating freedom, the company is confronting the hard reality of being owned by Google.
For Google, however, which has been patient so far about its return on investment, the time for DeepMind’s work to be commercialised -specifically a patient management app called Streams – appears to have arrived.
The UK company founded by Demis Hassabis, Shane Legg and Mustafa Suleyman has repeatedly vindicated Google’s assessment of its world class artificial intelligence research.
In 2016, its AlphaGo programme beat the world’s best player of the fiendishly complex board game “Go” after thousands of practice games.
In 2017 its progeny, AlphaGo Zero, did it again – without any expert human input.
When algorithms beat humans at their own games it is impressive; when they start beating them at their work it becomes unsettling.
This year, another DeepMind algorithm proved better than retinal specialists at London’s Moorfields Eye Hospital at making referrals when tested on patient scans. This was clear progress.
DeepMind’s health work is what is most immediately relevant to Britons since, through a partnership with the Royal Free Hospital, it has access to the data of so many patients.
The move to California has understandably raised privacy concerns at a time when big tech companies, including Facebook, are coming under growing scrutiny for the careless way they have exploited private data for commercial gain.
Moreover, the transfer appears to contravene promises by DeepMind that “at no stage will patient data ever be linked or associated with Google accounts, products or services”.
It is worrying that at the same time DeepMind’s independent review panel – set up to scrutinise its sensitive relationship with the NHS – is also being wound up.
DeepMind, which sees the move as a away of ensuring millions benefit from its work, claims that its contracts with the NHS are sufficient to protect patients’ data, which will remain under the strict control of Britain’s health service.
Google has said nothing. There is a clear need for both companies to offer much greater assurances.
Last year, DeepMind set up an ethics and society department, whose independent advisers were selected for their integrity.
They had a reputation for asking tough questions which set the company apart in the tech sector.
If indeed the founders believed this culture would be unaffected by the gravitational pull of a buyer as powerful as Google, they were naive. WhatsApp and Instagram made the same mistake.
But for the sake of the NHS patients whose data are at issue, it is to be hoped that the same culture and integrity survives in California.
The Silicon Valley mantra of “move fast and break things” might work for companies developing software. It has no place governing healthcare and technology.