Google bought British AI startup DeepMind in 2014 for about $525 million (or $600 million, depending on who you talk to). Its goal, as described by one of its founders, Demis Hassabis, was to make “machines smart.” His DeepMind artificial intelligence system has won at go and chess. But the firm has also lost vast amounts of money for Google over the years. That leaves market analysts wondering if the money is as smart as the machine:
The London-based AI lab—founded in 2010 by Demis Hassabis, Mustafa Suleyman and Shane Legg—saw its pretax losses grow to $570 million (£470 million), up from $341 million (£281 million) in 2017, and $154 million (£127 million) in 2016.
DeepMind’s losses are growing because it continues to hire hundreds of expensive researchers and data scientists but isn’t generating any significant revenue. Amazon, Apple, Facebook are locked in an expensive battle with DeepMind and Alphabet to hire the world’s best AI experts, with the goal of building self-learning algorithms that can transform industries.Sam Shead, “Alphabet’s DeepMind Losses Soared To $570 Million In 2018” at Forbes
The pay at DeepMind has attracted industry attention:
DeepMind doesn’t say how many people it employs, but its deepening costs appear to come after a period of recruitment. In late 2017, co-founder and CEO Demis Hassabis said DeepMind had doubled its headcount to 700 over the previous 12 months. Today, LinkedIn says DeepMind employs 838 people, suggesting Hassabis has been doing some hiring.
Either way, the implication is that Deepmind pays somewhere around £478k ($581k) per head (including pensions and individual travel). That’s pretty generous.Sarah Butcher, “This is how much Google pays people at DeepMind in London” at efinancialcareers
In an all-out botwar with the other tech Bigs, DeepMind could simply be paying top minds not to work for the competition while readying AI tools that pay better than winning at board games. One analyst put it like this: “Its relentless drive to push for innovation and little regard for balance sheets could drive out Google’s potential competitors in this space.”
The company works with the U.K. National Health Service hospitals, researching algorithms that can diagnose eye diseases and spot head and neck cancers from medical imagery, and the U.S. Department of Veterans Affairs on an algorithm that can predict which patients are at risk of sudden deterioration from acute kidney injury and other conditions.Nate Lanxon, “Alphabet’s DeepMind Takes on Billion-Dollar Debt and Loses $572 Million” at Bloomberg
But markets get antsy when doubled revenue is undone by “spiralling” losses, especially considering that DeepMind also owes 1.04 billion pounds, most of it, admittedly, to Google. DeepMind, we are told, has “written assurance” from Google that “it would be supported for at least another year and helped to pay its debts” (Telegraph) That’s not very reassuring when Google could just pull the plug, as it did on its Facebook wannabe Google Plus, earlier this year.
The AI triumphs that hit the headlines and spark commentary don’t necessarily predict the uses that will matter over time, like quicker and more accurate medical diagnoses. As Brendan Dixon likes to say, for a clearer picture, follow the money. consistently over time.
See also: Can AI make unique trail-blazing science discoveries? It would save us a lot of time but, as Eric Holloway warns, some things, by their nature, can’t be automated.
Is AI really becoming “human-like”? There are those who want an AI superintelligence to come along and save us so badly that they will glom onto anything that gives them hope. (Robert J. Marks)
Featured image: AI chess/johndwilliams, Adobe Stock