No, the robots are not going to rise up and kill you

We have seen astonishing progress in artificial intelligence, and technology companies are pouring money into AI research. In 2011, the IBM system Watson competed on “Jeopardy!,” beating the best human players. Siri and Cortana have taken charge of our smartphones. As I write this, a vacuum called Roomba is cleaning my house on its own, using what the box calls “robot intelligence.” It is easy to feel like the world is on the verge of being taken over by computers, and the news media have indulged such fears with frequent coverage of the supposed dangers of AI.

But as a researcher who works on modern, industrial AI, let me offer a personal perspective to explain why I’m not afraid.

Science fiction is partly responsible for these fears. A common trope works as follows: Step 1: Humans create AI to perform some unpleasant or difficult task. Step 2: The AI becomes conscious. Step 3: The AI decides to kill us all. As science fiction, such stories can be great fun. As science fact, the narrative is suspect, especially around Step 2, which assumes that by synthesizing intelligence, we will somehow automatically, or accidentally, create consciousness. I call this the consciousness fallacy. It seems plausible at first, but the evidence doesn’t support it. And if it is false, it means we should look at AI very differently.

A T-600 Terminator in a scene from Warner Bros. Pictures action/sci-fi feature Terminator Salvation (Richard Foreman/Richard Foreman)
A T-600 Terminator in a scene from Warner Bros. Pictures action/sci-fi feature Terminator Salvation (Richard Foreman/Richard Foreman)

Intelligence is the ability to analyze the world and reason about it in a way that enables more effective action. Our scientific understanding of intelligence is relatively advanced. There is still an enormous amount of work to do before we can create comprehensive, human-caliber intelligence. But our understanding is viable in the sense that there are real businesses that make money by creating AI.

Consciousness is a much different story, perhaps because there is less money in it. Consciousness is also a harder problem: While most of us would agree that we know consciousness when we see it, scientists can’t really agree on a rigorous definition, let alone a research program that would uncover its basic mechanisms. The best definitions capture the idea that consciousness grounds our experiences and our awareness. Certainly consciousness is necessary to be “someone,” rather than just “something.” There is some good science on consciousness, and some progress has been made, but there is still a very long way to go.

It is tempting to conflate something that we understand better with something we hardly understand at all, and scientists are not immune to this temptation. There is also the chauvinism of people who are smart for a living. To the professors, research scientists and engineers working on AI, intelligence is obviously the most impressive thing about being human. If we build the most impressive thing, of course the less impressive things will follow. But this ignores other qualities of humanity — empathy, hope, love — that may be at least as important. If we think that intelligence is sufficient, then we will be constantly thinking that consciousness is just around the next bend.

The history of AI shows just such a pattern. For a long time, experts scoffed at the idea that a computer would ever play a good game of chess. Then IBM made Deep Blue, which in 1997 beat Garry Kasparov, the world chess champion. We were forced to confront the idea that Deep Blue was intelligent enough to play chess better than any human but was still not at all conscious. The same pattern repeated itself with Watson and its ability to process natural language.

To someone who succumbs to the consciousness fallacy, this history must be deeply disappointing. Because consciousness has not magically appeared, we must be missing something deep about intelligence. But if we let go of the fallacy, we immediately benefit in two major ways. The first is in ethics: An intelligence that is not conscious is not a person, so any ethical problems related to the treatment of an AI system evaporate. The second is that AI may be much less dangerous than many people believe. Without a consciousness to drive it, there is no reason to expect such an intelligence to rise against us.

If the consciousness fallacy is indeed a fallacy, then the AI of the next decades will look very different from what we have been conditioned to expect from science fiction. We have been expecting software and robots to one day have secrets and dreams like we do. Instead, even as they advance almost magically, our AI tools are likely to continue to derive all of their root volition from us. Your self-driving car will take you places you want to go and even make suggestions. But it won’t argue, and when you’re done with it, it will just sit in your garage, recharging. There is simply no business value in pursuing a self-driving car that has all of these properties but also wants to go to college. If we can make the first without making the second, then things will be much less complicated for us. And much less frightening.

David W. Buchanan is a researcher at IBM, where he is a member of the team that made the Watson “Jeopardy!” system.

1 comentario


Deja un comentario

Tu dirección de correo electrónico no será publicada. Los campos obligatorios están marcados con *