The future of artificial intelligence depends on designing computers that can think and explore as resourcefully as babies do. The Defense Advanced Research Projects Agency (aka. DARPA), inventors of or contributors to small, non-world changing technologies like the internet, GPS and Windows! - are tackling this future. Their approach - build AI that can learn the same way babies do.
Babies seem to learn much more general and powerful kinds of knowledge than AIs do, from much less and much messier data. In fact, human babies are the best learners in the universe.
One of the secrets of children’s learning is that they construct models or theories of the world. Toddlers may not learn how to play chess, but they develop common-sense ideas about physics and psychology...They are surprised if they see a toy car hover in midair or pass through a wall, even if they’ve never seen the car or the wall before. Babies know something about people, too...1-year-olds see someone accidentally drop a pen on the floor and reach for it, they will pick up the pen and give it to them. But they won’t do this if the person intentionally throws the pen to the floor.
AI has traditionally been "children with super-helicopter-tiger moms—programs that hover over the learner dictating whether it is right or wrong at every step." If possible, getting the algorithms and models to instead self train on a smaller sample set of imperfect data, much as children do, would be a major breakthrough in AI development. This is a big if. Imitation, curiosity and millennia of genetic development are hard to program. I also don't particularly love the idea of robots imitating babies if it means more Johny Johny Yes Papa.Read more...