Two weeks ago, I wrote a post about AI principles of design. I concluded with this:
…there’s one other design principle I follow, one that I discovered myself and have never read about anywhere else. It’s probably the greatest single insight I’ve had since starting this project. But I’m out of time this morning, and it probably deserves a whole post in itself, so it’ll have to wait for now.
No time like the present.
First, a little background. We all know that animals – even simple animals – have a basic kind of intelligence. Even my pet hamster had it, back in the day, and trust me when I say hamsters are not the smartest creatures in the world.
My hamster’s name was Bowser, and his greatest desire in life was to escape from his cage. He tried all sorts of things: gnawing on the bars, forcing his way through the bars, digging a hole through the cage floor, climbing up to the ceiling. He had what I call trial-and-error intelligence. In other words, he would try something, see if it worked, and adjust his behavior accordingly.
This may not sound like much, but as an AI programmer, let me assure you that even this is a tall order to code from scratch. It took a long time for evolution to produce anything that complex. If you’ve ever watched a fly buzzing endlessly at a window, never thinking to try anything besides its default go-forward behavior, you can see how smart this trial-and-error mentality really is.
But of course, most of us wouldn’t consider trial and error alone to be true intelligence. True intelligence means sitting down with a totally new problem and figuring out the answer in your head, so that you only have to try one way: the correct way.
I call this reasoning intelligence, and it’s much more rare in the animal world. Other than humans, only crows, chimps, elephants, and a few other animals have demonstrated any kind of reasoning ability.
I’ve read lots of discussion about the gap between these two fundamentally different kinds of intelligence. How do you make that leap? How do you get from mere trial and error to actual reasoning? How do you make a machine that can truly think?
About five months ago, I figured out the answer in a late-night revelation, after everyone else had gone to bed. I can’t prove this is correct, but it feels very right to me, and it’s now one of the cornerstones of my design philosophy.
It’s simple. Trial and error isn’t fundamentally different from reasoning. They’re the same thing, the same essential act. The only difference is that trial and error means trying things in the real world, while reasoning means trying things in the mental world.
If I, like Bowser, were stuck in a big cage, I’d do exactly the same things he did. I’d try digging, attacking the bars, climbing to the ceiling, everything. The only difference – the only extra wrinkle – is that first, I would try those things in my imagination. Safer, faster, easier. But not really so different, when you think about it.
That’s reasoning. Trying things in your head before you try them in reality. That’s what separates us from the hamsters.
Thoughts?
Brian D. Buckley 
To some degree I think that you may be correct. I still don’t think that your idea of virtual trial and error covers all the concept of reasoning intelligence. For example in physics If I need to find a method to solve an integral. There are various ways to solve that integral. What do you do if you don’t know the method. You’ve tried all the ways you know but it uses a method you’ve n ever seen before. I’ve always found I have a moment of intuition in these cases, I usually make a logical leap. Otherwise, If you can’t even think of the trial then you won’t make the progress you need. Basically you can never solve the integral. A person may eventually solve the integral, but in that case I’m not sure a program ever could.
The problem with this example is that you can program a program to solve just about any integral through an iterative process. But even the best integral solvers, you’ll notice, still can’t do some integrals that people can do…
However, I do think that this virtual trial and error Idea has some strong potential.
Well, nothing says the ideas you try out have to be things you’ve tried before. 🙂 Otherwise you’d never do anything new. Any AI worth its salt would have to be able to build new actions from combinations of old actions.
You’re right that the reasoning-as-trial-and-error concept doesn’t come close to solving all the questions around AI. (I wish it did!) It’s only one step in the process – but it seems like an important one to me.
Now, if robots are given taste sensors and tacos, no reasoning is required. TACOS = DELICIOUS!
I’m not sure if trial and error in our head is a direct analogy for what I do when I try to solve a problem, but I definitely agree with the visualizing principle. I design code first in my head, then on paper, then finally in program code, I try to work out some of the architecture questions, pitfalls, and scope before approaching a solution. I definitely filter out ridiculous solutions (I don’t try those in my head), but I would think that my own experience filters could be applied to an AI.
I still think for some problems in AI knowledge has to be pre-programmed. The integral example is a good one. Most of us don’t know how to solve an integral until we’re taught, and trial and error doesn’t get us to a solution. But once we’re taught, it’s easy. Humans don’t try to intuit all of human knowledge from scratch, and I don’t think an AI would as well.
BTW don’t make your AI female. Fighting the minions of a Goddess while she taunts you and calls you an insect is very emasculating.
So, I should make an important clarification. Mental trial and error is the main thing separating “reasoning” from lower forms of intelligence, but it’s certainly not the only thing going on. You can’t even attempt any kind of trial and error (mental or otherwise) without a vast array of mental filters and analyzers, all built on experience/teaching. Nobody’s going to solve an integral if they don’t know what numbers are. 😉