Worried about AI? So’s the Nobel Prize-winning physicist who helped invent it.
Artificial intelligence has the power to end life as we know it—both in positive ways and in terrifying ones, says Geoffrey Hinton, known as the “godfather of AI” for his pioneering work in the field.
“The other thing to remember is: You ain’t seen nothing yet,” Hinton told attendees at NCSL’s Legislative Summit in Chicago. AI has developed much faster than even its biggest enthusiasts expected, he says. “And the other thing that’s happened is society’s become more aware of it. And a lot of the public are now worried about it. And I think they’re correct to be worried about it.”
Hinton says risks of AI include everything from massive job losses to the end of humanity itself, as super-intelligent machines decide that humans are no longer necessary. “It’s very, very early days,” he adds. “I should emphasize that nobody knows what the future’s going to be like.”
Among the risks Hinton cites:
- Mass unemployment: AI could replace routine intellectual and administrative work, potentially eliminating millions of jobs without creating a proportional number of new ones.
- Erosion of information integrity: The flood of highly convincing fake videos, audio clones and targeted disinformation could make it increasingly difficult to determine what’s real.
- Political manipulation: Bad actors could easily exploit intelligent chatbots and personal data to manipulate electorates and fuel divisive online echo chambers.
Hinton says advanced AI agents must generate subgoals to complete complex tasks, and they will determine that self-preservation and acquiring power are key to achieving those tasks.
“They want to survive,” he says. “There’s an example of an AI that was shown the internal mail of a fake company. And it was obvious that one of the engineers was having an affair with somebody else in the company. And then it was shown a message that said that it was going to be replaced by another AI, and this engineer was going to be doing the replacing. So, this AI, by itself, dreamed up a simple blackmail: It sent a message to the engineer saying, ‘If you try and replace me, I’ll let everybody in the company know you’re having an affair.’ It made that plan.”
As scary as the risks might be, AI brings with it enormous benefits in productivity, Hinton says, adding that healthcare and education can expect to reap major rewards. AI chatbots can become the private tutors that many people can’t afford for their kids, he says. “We know that a private tutor can teach children stuff about twice as fast as the classroom. If you follow their curiosity and answer what they’re wondering about, they’ll learn very fast.”
AI already is better at diagnosis than human doctors, Hinton says, citing a Microsoft experiment that showed chatbots correctly diagnosed difficult cases with 80% accuracy, compared with 20% for humans. “In the United States and Canada, about 200,000 people a year die from bad diagnosis. AI should be able to at least halve that.”
Asked about possible AI guardrails, Hinton says big tech companies bristle at the idea of regulation, which they insist will cede progress to the Chinese.
“They want you to think of innovation as like the accelerator of the car, and they want to use the idea of regulations like the brakes of the car,” he says. “There’s an alternative model, which I think is far more accurate, which is (that) innovation is like the accelerator of the car, and regulation is the steering wheel. Now, you wouldn’t want to make very fast cars without steering wheels. The whole point of regulation is not to stop progress—it’s to cause progress to happen in the right direction. To make these technology companies produce things that are good for people, not things that are bad for them.”
Hinton likens creating AI to creating a new kind of being—and we still have power over what kind of being we’re going to create.
“Just creating beings to be as smart as possible and make short-term profits seems crazy to me. We should be thinking very hard about how we create them so that they’re good, and so that they care about us,” Hinton says. “We need to design these beings to care more about us than we do. And we still have that possibility.”
Lisa Ryckman is NCSL’s associate director of communications.