DeepMind would have 'probably failed' without Google, says early investor

Google Deepmind head Demis Hassabis speaks during a press conference ahead of the Google DeepMind Challenge Match in Seoul on March 8, 2016.
Jung Yeon-Je | AFP |Getty Images | Getty Images

LONDON — One of DeepMind’s first investors believes the artificial intelligence company wouldn’t be around today if it hadn’t been acquired by Google.

Founding investor Humayun Sheikh, who held 43,750 of DeepMind’s 3,386,754 shares in 2011, told CNBC that the London AI lab would have “probably failed” if Google hadn’t bought it for a reported $600 million in 2014. Facebook was also interested in buying DeepMind but ultimately the company accepted Google’s offer.

“Commercialization for any AI based company is very difficult, unless you get absorbed by a big corporate,” said Sheikh, who now runs his own AI business called Fetch.ai, on a call last Friday. 

“If you think about DeepMind on its own today, DeepMind would have probably failed as a company because you wouldn’t be able to commercialize anything.”

A DeepMind spokesperson said: “We chose Google as a partner because it was clear they were as passionate about AI as we were and since then our partnership has gone from strength-to-strength in pursuit of our shared long-term ambition for responsible and impactful AI.”

Sheikh, who claims to be a friend of the company’s CEO and co-founder Demis Hassabis, said there were concerns about how DeepMind was ever going to make money from the beginning. “We spent five years in close discussions about this whole DeepMind thing,” he said.

DeepMind set out to crack artificial general intelligence (AGI), which is often referred to as the holy grail in AI. But, Sheikh said: “The concern, the question marks, have always been in commercialization … How do you do it?”

While Google has found uses for DeepMind’s AI, its technology has not been widely applied elsewhere. It has partnered with the U.K. National Health Service but the size and scope of the project is relatively limited. DeepMind was also in talks with electricity grid operator National Grid, but ultimately it came to nothing.

The Google acquisition gave DeepMind access to vast amounts of computing power that Google has across its data center network, which has allowed it to train AI models to do things like play abstract strategy board game Go. It also gave the start-up access to engineering talent, and a steady stream of finance. 

DeepMind costs Google’s parent company Alphabet hundreds of millions of dollars every year. In 2018, it made a loss of £470 million ($622 million), up from £281 million in 2017 and £127 million in 2016. Its losses are growing because it continues to hire hundreds of expensive researchers and data scientists but isn’t yet generating any significant revenue.

Jon Crowcroft, a computer science professor at the University of Cambridge, told CNBC that it’s hard to see a business case for DeepMind’s scale and burn rate.

“A normal investor would be looking at big customers and an income stream well before now,” Crowcroft said. “They have developed techniques and prototype tools that could lead to that, but [there is] a long way between lab and reality.”

Another industry source, who did not wish to be named due to the sensitive nature of the matter, agreed that DeepMind would have eventually needed a big tech parent to survive indefinitely. “The company is engaging in research, not products, and wouldn’t be able to obtain indefinite VC funding without eventually having a business model,” they said, adding that Google is an excellent fit.

“Google is large enough and AI is fundamentally important enough for its (highly profitable) business model that having DeepMind as a subsidiary makes solid commercial sense,” the source added.

This post was originally published on this site