Gigaom brings you our unique analysis and commentary on the present and future of AI.
I've been spending some episodes of the AI Minute talking about common sense, how it is that we know not to walk off a cliff's edge, but a machine has a hard time learning that. And I've been thinking about how is it that you could teach a machine to have common sense. Now I'd like to close the series out by talking about how it is that we have common sense.
Now you might think, well we're born blank slates, these empty minds, we have a set of experiences and we somehow build a web of what we call common sense. This turns out not to be the case because children are born with certain instincts. They generally know not to crawl off a cliff's edge. At a very young age, a frowny face and a smily face on a piece of paper and they can distinguish between the two. They have certain behaviors they know how to do right from the instant they are born. They're probably afraid of snakes right away. We don't know how it is that we're born with this knowledge. If you think about it, you know, we all began as this cell that divided and divided and divided enough times to make us. And yet, somehow, the knowledge that's passed down persists through all those divisions.
It stands to reason it's encoded in our DNA on a level we don't even understand. So, this idea that you can take a robot and basically make it an infant and have it have, quote, "life experiences" and then it will eventually acquire common sense probably isn't the case. Aside from the fact that we don't really know how the mind organizes and connects information. We don't even know what set of information we began with as humans. So we don't really know why it is that we have common sense and therefore we don't really know how to teach a computer to.