1. Home
  2. tech
  3. Google's cutting edge artificial intelligence unit is costing millions

Google's cutting edge artificial intelligence unit is costing millions

Google's cutting edge artificial intelligence unit is costing millions

lee sedol & demis hassabis

Google Deepmind

DeepMind CEO Demis Hassabis and Lee Sedol, the Go champion defeated by DeepMind's AI in 2016.

  • Google's artificial intelligence business, DeepMind, has reported rising expenses for 2017.
  • It's also generating more revenue thanks to commercialising its cutting-edge research to help its parent company.
  • In financial filings made in the UK, the company said expenses had more than doubled thanks to staffing and other administrative costs.
  • Hiring top-tier academics, doctors, and engineers isn't cheap.
  • DeepMind is best known for its AlphaGo algorithm, which beat the human champion of strategy game Go in 2016.

Revolutionising the world with cutting-edge artificial intelligence is, as it turns out, an expensive endeavour.

DeepMind, the AI research unit belonging to Google, has seen its costs almost triple in 2017, thanks to rising staff and infrastructure expenses.

It is also making more revenue by commercialising its technical breakthroughs to help its parent firm with things like cooling server warehouses, according to earnings filed at the UK's Companies House.

Here are the key numbers for the year to December 2017:

  • Revenue: £54.4 million ($71 million), up 35% from £40.3 million ($52 million) in 2016.
  • Losses before tax: £281.9 million ($366 million), up 123% from £126.6 million ($164 million) in 2016.
  • Expenses: £333.8 million ($433 million), up 104% from £164 million ($213 million).
  • Staff costs: £200 million ($260 million), up 91% from £104.7 million ($136 million) in 2016.

DeepMind is one of Google's most fascinating projects.

Google acquired the UK-based business for £400 million (then $600 million) in 2014, in a major bet on the future of artificial intelligence. The company has made history in its mission to create general artificial intelligence. In 2016, its AlphaGo algorithm beat the world's human champion at the strategy game Go, one of the most difficult games ever invented.

Last year, the firm attracted controversy for its work in healthcare, with the UK's data watchdog finding that its data-sharing deal with the Royal Free hospital was illegal. It has tried to move past that scandal and is cementing more clinical partnerships and healthcare research projects. In August, the firm unveiled research showing its AI could detect eye disease.

Hiring all those top-tier doctors, academics, and engineers is costly though and, as yet, it isn't clear how DeepMind will become profitable. Expenses more than doubled, thanks to technical infrastructure, staff costs, "professional service fees," and charitable gifts.

Staff costs almost doubled to £200 million. That figure covers salaries as well as benefits and costs like pensions and travel. Staff numbers were not reported in its earnings, which is relatively unusual, but CEO Demis Hassabis said in 2017 the firm had around 700 employees.

DeepMind's revenue doesn't, as yet, reflect any money it makes from its healthcare work. A spokeswoman said its reported turnover mostly reflects the work DeepMind does for its parent Google, such as using AI to cool Google's server warehouses and work on Google Assistant. According to a June report by a panel that oversees the firm's healthcare work, the company doesn't make money from its health business yet.

The filings show DeepMind spent £8.1 million on academic donations. Asked for clarification, a spokeswoman said the company had given grants to New York University, University College London, Imperial, and the University of Alberta "to support AI research initiatives." The donations don't come with conditions, she added, and all institutions disclose their grants publicly.

Demis Hassabis, CEO and cofounder of DeepMind, said in a statement: "We're on a long-term scientific mission to solve intelligence and use it to benefit the world, and we're really proud of the impact our work is already having in areas from healthcare to energy.

"Since our founding in 2010, we've assembled a world-class interdisciplinary team of machine learning experts, neuroscientists, engineers, ethicists and more, and created a unique environment where ambitious, long-term research can flourish. We intend to keep investing in our mission, and look forward to the scientific breakthroughs that lie ahead."

Get the latest Google stock price here.


Popular Right Now