Ray Kurzweil, Nick Bostrum and others have been warning of “The Singularity” for some time now.
In their scenario, an AI wakes up one day and is suddenly smarter than any human, and alone, or by coercing humans to help it, this single AI either takes over the world, or helps we poor dumb humans out of all of our problems in a flash of intelligent light.
On the other hand, Daniel Dennett talks at length on youtube and in his books about the power of culture, the power of memes and “Intuition Pumps” and other tools for thinking. It seems to me that the Singularity Folks have it wrong and Dennett is a lot closer to the truth.
We all seem to realize, and agree, that the AI, when / if it wakes up, is going to have to learn about the world. Just like we do from infant-hood. Of course this learning, a lot of it at least, can be very fast since not only will the AI brain be faster, but it can suck down the internet at blazing speed. But then the Singularity Folks make the magical leap to the place where this Super Brain is all of a sudden smarter than a Whole Culture of people, Literally Billions of Humans. I think that’s wrong. This AI may be faster, but it does not have a culture separate from our own. It’s embedded in our culture.
We are unlikely to spring for 100s or 1000s of Super Brains before one or a few have proven themselves to be useful. And, for reasons that I outlined in my previous papers, one or a few brains cannot take us over and build an army. So I submit that we can relax about the sudden AI Apocalypse.
I submit, along with Dennett, that it is our Culture that makes us smart, not just our brains. One mind could no more develop an understanding of Particle Physics than it could build an iPhone. We are smart, as a Culture, and because we have Culture. And just as we are not smart alone, not stupid, but not that smart either, this Super Brain AI will not be smart alone either. Let’s look at the reasons for this:
First: Invention is the combination of ideas, many of them created randomly, not a straight line path to progress. Progress is a random walk through lots of possibilities and lots of failures. No matter how smart a brain is, the random walk requires many of these brains with slightly or wildly different perspectives coming up with ideas and then the physical testing of those ideas.
No brain, no matter how smart or how quick can simply think itself to a solution to a complex problem, if the solution is not a mathematical proof that is. Sure math proofs can be a matter of thinking, if all the techniques for solving it are known. But if the techniques for solving the proof come from other disciplines of mathematics then we use multiple brains to solve it. So even here, a single smart brain trained in one way on one set of examples is not enough to get the job done.
Second: A blazing fast AI Brain does not greatly speed up the work in the laboratory, of any kind, to do the experiments to find out whether an idea is worth pursuing. Sure, a Super Brain in each laboratory may speed up the work in each laboratory to some extent. I’m not saying that we should not or will not employ them. But they aren’t going to solve the problems in a flash of insight. Consider the Large Hadron Collider of recent fame. The Higgs Boson Discovery would not have been accelerated that much, I contend, by one or even 100s of Super AI brains. The development of the 1000s of solutions to problems to allow this masterwork machine to be built required 1000s of small experiments to be carried out all around the world by 1000s of people in 100s of labs.
I’m not saying that Super or General AI is not worth pursuing, but only that from these principles, we have little to fear from the AI Apocalypse and I believe that we will find that we need to use these Super Brains in combination of many of them and with us to solve the really hard problems, just as we find we must work together to solve those problems.
Daniel Dennett – Wikipedia
And look for his many talks on YouTube and many books.
Artwork by Adam Burn at Deviant Art used without permission.