This past week, I had the rare opportunity to interview Millicent Robinson, Professor of Aritificial Curiosity at the County Durham College of Agriculture, Animal Husbandry and Fine Art and one of the designers of GEBx project. What follows is an excerpt from that interview:
CTR: Good morning, Prof. Robinson, and thank you for taking the time to answer a few questions.
Prof. Robinson: My pleasure. How would you like to begin?
CTR: Well, everyone is familiar with artificial intelligence but I suspect most people have never heard of artificial curiosity. Could you give us a brief description of the differences between the two?
Prof. Robinson: Certainly. I’ll keep it very brief. Artificial intelligence is a lie and artificial curiosity is the only real way to achieve a truly thinking machine.
CTR: Whoa, now! You’re calling AI a lie? You’re talking about the technology that’s given us robots that vacuum our homes, algorithms that can translate Scottish into English (sort of) and that shooting-a-car-into-space-guy, what’s his name, Elon Musk.
Prof. Robinson: Typical. You non-scientists are all alike. “What has your tech done for me lately?” you say, or, “Quick, do something useful or I’ll take away your funding.” You’re only interested in results, in achieving a specific goal, and usually a stupid one, at that.
CTR: Hey, I’m all for blue sky research. I just don’t see how you can call AI a lie when it has proven itself in so many ways.
Prof. Robinson: I see we’ll have to do this the hard way. I assume you know what the Turing Test is.
CTR: Of course. Alan Turing’s idea that a computer would be considered intelligent if it could convince a human being that it was a person.
Prof. Robinson: Precisely. A complete lie. The whole premise of the Turing test is deception. A computer is intelligent if it can trick a human. A chicken can do that, for Pete’s sake. Complete BS.
CTR: Okay, I see what you mean by a lie. But that is only one way of defining AI.
Prof. Robinson: Wow, you’re even dumber than my fitbit. The lie isn’t just in the deception. The Turing Test sets a specific task for a computer, i.e. to trick a human, and success is measured solely by achieving that goal. Same goes for the Winograd test. In fact, a large number of researchers define intelligence as “the ability to achieve goals”. By that measure, I can build a robot to pick my nose and rate it’s intelligence in the weight of boogers it collects.
CTR: Please tell me you haven’t done that.
Prof. Robinson: Don’t be stupid.
CTR: That’s my goal for today. But let’s move on to artificial curiosity. What is it, exactly, and how does it differ from AI?
Prof. Robinson: Ah, an Artificial Curiosity is a computer that can set its own goals.
CTR: Wait, you mean there’s no one telling the computer to perform a specific task, to drive a car or whatever. You mean the computer decides for itself what it is going to do.
Prof. Robinson: Precisely. That’s what the GEBx project is all about. We designed GEBen–the first iteration of what we hope will be many such computers–with a kind of feedback loop that makes GEBen pay particular attention to surprising outputs and to focus on analyzing data that doesn’t fit existing models.
CTR: So GEBen is identifying anything that is unusual or unfamiliar and paying particular attention to that.
Prof. Robinson: Yes, GEBen is driven to learn more about anything different or new. In other words, GEBen is curious.
CTR: I thought GEBen was supposed to analyze aesthetonomic data, looking for links between human creativity and radioactive decay.
Prof. Robinson: Yeah, that’s what we wrote in the grant applications. But when we got to the point of actually building GEBen, there were too many other supercool things we wanted to do. The aesthetonomics people still get their data crunched, they just have to wait a little longer when GEBen goes off on a tangent.
What’s important, however, it that this is precisely the kind of thing that traditional AI will never be able to do. GEBen isn’t trying to trick us into thinking it is human. GEBen is just being GEBen. And it’s doing it in the most delightful and imaginative ways, producing research and, dare I say it, art, that no human could ever come up with.
And that is what artificial curiosity is all about.
End of interview
That’s all I’m going to publish today. Our talk lasted a good hour and Prof. Robinson went into fascinating detail about GEBen but the more more complicated things got, the more I had to stop her and ask for clarifications, definitions and non-mathematical explanations. And that resulted in a lot of swearing. I’m trying to edit that out so I can print a family-friendly version of the interview, but I’m not sure that will be possible.