Facebook's test release today of a digital assistant inside its Messenger app is a shot across the bow of the Internet's biggest companies: Apple, Google, Microsoft, and Amazon.com. It's also the latest salvo in a high-stakes battle over the ways artificial intelligence should transform the way we live and work.
Facebook M is intended to allow users of Facebook Messenger to pose any query or service request in natural language and get a personalized answer immediately. The key wrinkle that sets it apart from Apple's Siri, Google Now, and Microsoft Cortana is that there's a team of human "trainers" who will step in when the machines aren't quite up to the challenge.
So far, it's only available to a few hundred people in the San Francisco Bay Area, and its timing and scope are unclear. But judging from a brief post by VP of Messaging Products David Marcus, Facebook has big ambitions for it to become a prime intelligent assistant for a wide range of tasks:
M is a personal digital assistant inside of Messenger that completes tasks and finds information on your behalf. It's powered by artificial intelligence that's trained and supervised by people.
Unlike other AI-based services in the market, M can actually complete tasks on your behalf. It can purchase items, get gifts delivered to your loved ones, book restaurants, travel arrangements, appointments and way more.
This is early in the journey to build M into an at-scale service. But it's an exciting step towards enabling people on Messenger to get things done across a variety of things, so they can get more time to focus on what's important in their lives.
So Facebook M is clearly a major bid in a quickening battle to be the virtual assistant of choice, taking on not only Siri, Google Now, and Cortana, but also a raft of upstarts such as Luka, Magic, and Operator. And in the mobile age, virtual assistants could prove to be the key product that will define which companies dominate the next decade of online services, just as search was for the past decade. “Whoever creates the intelligent assistant will be the first place people go to find things, buy things, and everything else,” former AI researcher Tim Tuttle, CEO of the voice interface firm Expect Labs, said last week.
But what's even more interesting in the bigger picture is how Facebook M plays into a longstanding, fundamental battle over how artificial intelligence should be employed--one that has recently come into sharper focus. Bolstered by new approaches such as deep learning and powered by vast networks of increasingly powerful computers, AI has produced big breakthroughs in areas such as speech and image recognition as well as self-driving cars and robotics.
The progress has been startling enough to suggest that machines will take over a wide range of even sophisticated tasks that provide employment for many millions of people. Even more than that, big brains as diverse as theoretical physicist Stephen Hawking and Tesla Motors and SpaceX CEO Elon Musk, not to mention entrepreneur Jerry Kaplan, author of the thoughtful new book Humans Need Not Apply, worry that AI-powered machines could be a danger to humanity.
Others argue that AI's capabilities are still quite limited, and will remain so for decades--leaving not only room for people but requiring their active participation. Take "driverless" cars, for instance. As New York Times writer John Markoff, author of the new book Machines of Loving Grace: The Quest for Common Ground Between Humans and Robots, pointed out in a recent interview, it's not clear that we can ever solve the handoff problem--when and how the car will hand back control to humans in situations where it can't decide the best course of action. A better use of the technology, he suggests, might be accident prevention rather than taking over driving--essentially aiding rather than replacing human skills.
The issue of how much control to cede to machines, or indeed to what extent machines can ever handle many tasks requiring true intelligence and experience, is an old one. As Markoff himself wrote in a previous book, the issue is whether we should have Artificial Intelligence or Augmented Intelligence. John McCarthy, the Stanford computer scientist who coined the term artificial intelligence, aimed to create an actual superbrain that could outdo humans on tasks. Doug Engelbart, inventor of the computer mouse and other breakthroughs at the Stanford Research Institute, believed to the day he died in 2013 that computers should be tools to augment human intelligence.
While the two poles may not be that far apart in the real world, McCarthy's approach has continued in services such as Siri and Google Now. Facebook M sounds like Doug Engelbart all the way, at least for the foreseeable future. And as it turns out, Facebook is in tune with a number of other people in AI these days, suggesting that we humans still have a long future of productive work ahead of us. XPRIZE, the nonprofit organization that sponsors technology "grand challenge" contests, for instance, plans to design an AI-human collaboration contest.
"In some cases the machines have already replaced jobs, like building cars on an assembly line or defusing mines," says Matt Sanchez, founder and CTO at CognitiveScale, a firm that offers AI-powered computing services in the cloud, and one of the team that started IBM's commercial Watson computer division. "But at the end of the day, man will still have to guide the machines."
To put it more starkly, so to speak, he says he's trying less to create HAL than Iron Man's Jarvis suit. The MD Anderson Cancer Center at the University of Texas in Houston, for instance, uses CognitiveScale technology for a "concierge" to help medical teams provide personalized service to cancer patients. "Machines can augment the humans," he says, "transforming any knowledge worker into a super human ... and using cognitive computing to bring insights and advice directly to them when they need it."
Likewise, even the robots that spark Terminator fears are still used as much to aid human workers as they are to replace them (though replace them they often have, and will). Fetch Robotics, for instance, offers a robot that follows human pickers around a warehouse, collecting products the workers pluck off the shelves. Amazon's Kiva robots bring stuff to workers to process.
In any case, some experts believe that AI simply isn't good enough for a wide range of tasks, and won't be good enough for a long time to come. Gary Marcus, professor of psychology and neural science at NYU and CEO and founder of a new AI startup Geometric Intelligence, reiterated his deep skepticism about AI progress last week at a conference in San Jose.
"We wanted Rosie the Robot and instead we got Roomba," says Marcus (no relation to David Marcus). "AI is still after 50 years basically a collection of idiot savants," unable to truly learn or exhibit judgment. That's because leaders in AI such as Google, he contends, have fallen in love with statistics and forgotten AI's roots in trying to crack the secrets of human cognition.
So, Facebook M will be a key test of what direction AI will take, at least in the short term. The service itself clearly faces challenges, not only from the likes of other intelligent-assistant leaders such as
Not least, Facebook M could face challenges from its own users. Although many who commented on Marcus' post seemed eager to try it out, some expressed worries. Wrote one: "Great idea. Ask for products and services, sell our data to the highest bidding advertisers to put forward their product suggestions. Big data. Ruthless."
The notion of
Not least, it's hard to predict how well Facebook's approach can scale up to serve all its nearly 1 billion users worldwide. Marcus told Wired that eventually Facebook anticipates employing thousands of these trainers, though by calling them trainers, Facebook is also signaling that more and more of their responses ultimately should be automated.
But that clearly won't happen overnight. Call it the revenge of Doug Engelbart: Until and unless AI gets so good that machines can anticipate what we want, people will remain a key component of truly intelligent online services.
