Robots aren’t going to take over the world, but humans could use them for some scary stuff. That’s Amazon CEO Jeff Bezos’s take on artificial intelligence (AI), according to CNBC. And in the wake of Google’s latest tech revelation, many experts are agreeing with him.
Google revealed its human-sounding AI assistant Google Duplex during this week’s I/O conference. Google Duplex can make phone calls for users to perform tasks such as scheduling appointments and reservations — even completing those conversations with natural, human-sounding “ums” and “uhhs.” Though the technology demoed at the event is still in its early stages, it has drummed up a fair share of controversy as people debate the ethics of letting robots do the talking. Should robots have to identify themselves as such when conducting business on behalf of their owners? Is it deceptive if they don’t?
For Bezos, though, it didn’t take Google Duplex to get him worried about the potential abuses of AI. Just because we aren’t headed toward a future of digital overlords or a full-on, Matrix-style robot apocalypse, it doesn’t mean this tech can’t do some serious damage. In particular, Bezos says he finds the concept of autonomous weapons “extremely scary” — and the AI tech that already exists today is more than enough to take many frightening sci-fi concepts out of the fiction genre and into the real world.
AI could potentially be abused for synthesizing speech for impersonation, analyzing human behaviors, moods and beliefs for manipulation, automated hacking, and physical weapons — such as swarms of micro-drones. Bezos says the world needs to come up with some sort of international treaty, similar to the Geneva Convention, before this happens.
But he’s far more worried about how humans will use AI than about how AI will treat humans. He said that today’s technology is only capable of building a “narrow” AI, designed to assist in specific tasks, rather than a general AI that could set its own objectives. Even if (or when) technology reaches the point of creating a general AI, Bezos doesn’t find it likely that the machine’s first thought would be to kill or subjugate all humans.
The late Stephen Hawking, however, was more cautiously optimistic. Yes, an effective AI could help humans — but it could go either way, and Hawking felt it would all depend on how the machine’s creators prepared for and avoided the potential risks. Otherwise, he said, this technology could be worse than disruptive — it could be downright dangerous, driving powerful autonomous weapons or creating new means by which the few can oppress the many.
Tesla and SpaceX CEO Elon Musk falls on the more cautionary side, calling it an “existential crisis” to ensure that future digital superintelligences are symbiotic with humanity.
On a less existential note, Bezos said that the impact of AI on the job market will not be as dire as everybody thinks. While some reports suggest that tens of millions of jobs could be lost or cut back due to automation, Bezos sees this as a lack of imagination around what future jobs will look like.
Today, Bezos noted, there are massage therapists and dog psychiatrists. If someone had told 1918 farmers that those jobs would exist in 100 years, it would have blown their minds. The future of AI is similarly inscrutable, he said. He is hopeful that automation could reduce the amount of routine jobs that people have to do and elevate the quality of work — enabling more people to have more than just a job, but a career or even a calling.