Google's Eric Schmidt: There's no question AI will put jobs at risk, but it's natural
Published 14/03/2016 | 07:39
Few company chairmen could justify taking a 10- hour flight to travel 5,638 miles to watch a board game being played. But Eric Schmidt could.
The Alphabet chairman last week took the trip from Google’s holding company’s headquarters in California to Seoul, South Korea, to watch world Go champion Lee Se-dol go head to head with AlphaGo, an algorithm created by Google-owned British artificial intelligence company DeepMind, over five rounds of the ancient east Asian board game.
“When I was a young computer scientist in the Seventies, there were many claims that we would beat human intelligence. None of it happened,” Schmidt said over a gourmet Chinese meal a few hours before the first Go game. “Now there is a sense that AI [artificial intelligence] has finally arrived.”
Last summer, Google announced its plan to restructure itself into several companies under the Alphabet umbrella, to give investors greater visibility into how the group spends its money. Some observers see a division between the core business, made up of its eponymous search engine, YouTube and its Android phone operating system, and so-called “bets”, including its work on driverless cars and its health business. But at the suggestion that machine learning sits in the latter division, Schmidt recoils, curtly: “The cast you gave, which is roughly mature versus experimental stuff, is not what Alphabet is about...It’s about building real businesses in new areas. Google is about performance.”
Schmidt’s high-profile trip is part of a public relations exercise to show just how some of its new technologies can perform. Se-dol yesterday won his first game against his machine opponent, but given it won the first three, he has already lost the challenge.
Google is investing heavily in AI, in particular in machine learning, or the ability of computers to learn to spot patterns from large datasets, and achieve an ambitious goal.
“I can’t quantify our investment into AI, but it will eventually be embedded in everything,” Schmidt says. “It started small, with hundreds of people and now we have thousands. It’s a very significant investment.” The growing team in this area will form the core of the company, Schmidt said. “They’re not in cul de sacs, they’re building things that will touch millions and millions of people.”
Machine learning technology is now essential to Search, which remains the crown jewel of Google services.
Google’s most high-profile engineer, Jeff Dean, has said machine learning is the third most important signal, out of hundreds, that determine how Google search results are ranked. Numbers 1 and 2 are a closely-guarded secret.
Considering the software being used only developed in 2015, its fast-rising importance has surprised even company insiders. “One of the concerning things is that [machine learning] is so new that even the best engineers don’t know the techniques,” Schmidt said.
In response to this need for education, Schmidt said the company runs “a huge internal training programme” where engineers from across Google’s business are encouraged to start using AI in everything from Android to Gmail.
This, Schmidt says, has resulted in an enthusiastic uptake, with machine-learning algorithms powering the 100 languages on Google Translate, for instance.
“Now you can actually talk into a phone, it translates and comes out in another language on the other side, which is sort of amazing,” Schmidt says. Machine learning is also behind Google Photos, which automatically labels pictures based on what it thinks they are of.
And that’s just the start. “Think about all the things Google does that are big – we have lots of searches, lots of ads, lots of customers, lots of data centres, lots of people using Google Compute, lots of people using our security software,” Schmidt says. “Whenever you have a large number of people using something, we can probably use machine intelligence to make it more efficient.”
But it’s not just Google’s normal line of business where Schmidt thinks benefits can be reaped. The company is particularly interested in two areas – healthcare and smartphone assistants, which it believes can be boosted by the new technology.
“Let’s take healthcare,” explains the 60-year-old. Recently we [tested] diabetic retinopathy, where your blood vessels change shape in your eye, causing blindness,” he says. “We can diagnose it more accurately than an ophthalmologist.” Why? “Because we see more eyes, we see a million eyes and an ophthalmologist may see 10,000.”
Now that a machine has beaten a Go grand master at a game he’s been playing professionally for 20 years, surely there is a concern that AI-fuelled robots will be able to replace humans in other areas, hurting jobs?
“There’s no question that as [AI] becomes more pervasive, people doing routine, repetitive tasks will be at risk,” Schmidt says.
“I understand the economic arguments, but this technology benefits everyone on the planet, from the rich to the poor, the educated to uneducated, high IQ to low IQ, every conceivable human being. It genuinely makes us all smarter, so this is a natural next step.”
A natural next step for Alphabet, perhaps, but for those whose jobs may be displaced by robots and the like, Schmidt may yet have to do some convincing.