|
Getting your Trinity Audio player ready...
|
“I started in the MIT AI lab, and it was one of the very early places that did symbolic AI – what’s called good old fashioned AI. I ended up coming to the AI lab at Stanford. Over time a couple of things happened.
My approach shifted, dropping the assumption that we were modeling human thought. (Instead), thinking about, ‘What is it in human interaction with computers that helps people, hurts people, matters to people?’ So I focused on people, not as cognitive engines, but as beings with needs.
Current work on Human Computer Interaction – which I’m not doing now, I’m retired – has to take into account the mix of people trying to get things done with computers. So AI is now at the center of human computer interaction, because that’s how people are interacting.
What machines take over depends on what you let them take over. They don’t have armies to go out and do things.
terry winograd
AI is tremendously useful in an operational sense. There’s all sorts of things that it is good at. The question is, ‘How do you identify those places where it’s good, and those places where it may appear to be good, but is misleading from a human point of view?’
Elon Musk says (that) in 10 to 20 years, AI will replace us, and work will be optional, and money will be irrelevant.
Then you look a little closer, and you see productivity could lead to replacement. It can enhance jobs, it can also degrade jobs. Who decides which of those happens? That’s not a technical decision, that’s an economic and business decision. If you do replace jobs, you can get universal prosperity or you can get destabilizing unemployment. The punchline here is there are lots of directions things can go, and the technology doesn’t determine those directions. The political economic system does.
We live in a capitalist system where efficiency is a very, very high priority in your profit making. But it always has unforeseen consequences. Sometimes being inefficient is good for things.
My basic bottom line is you don’t want to use AI in those cases where human care is really what matters….If it’s something that gives them a feeling of meaning or value or whatever it is, don’t give it to the computer.
Risks are on different scales. You can say, ‘Here’s where it’s going to go in the future, and in the meantime….’ But of course we always live in the meantime. The only time we have is the meantime.
What machines take over depends on what you let them take over. They don’t have armies to go out and do things. The other side of it is the hollowing out of humanness, which is the big threat: that we will all become machines because that’s what we’re used to dealing with. Of course, there’s the other big potential, which is social unrest, and if you have social unrest, you get chaos. I think it’s very likely in the next couple decades we’re going to have some major social unrest, in which computers play a part.”
Terry Winograd is a professor emeritus of Computer Science at Stanford University. Terry created and directed the Human Computer Interaction Group at Stanford and was a founding faculty member of Stanford’s Hasso Plattner Institute of Design, known as the “D school.”
This profile is one of nineteen featured in the “Humans in an age of AI” portrait series. See more profiles here.
