trisweb.com

I was forced against my will (child labor) to mow my parents’ lawns, and while I was pushing the mower around in circles (some people go back and forth, some go diagonally—I’m a spiral mower myself) I got to thinking about my Artificial Intelligence class at Berkeley and how I was basically ridiculed for saying that Intelligence is complex!

I guess it sounds like an ignorant viewpoint for a computer scientist to have. Intelligence is complex—or should it be just a very powerful computer with the right program? It seems to me like most people in the strong AI group think this way; that it’s just a lot of complicated parts needing independent solutions and some putting-together. This view is almost required to believe it can be done in the first place! I do believe it can be done—in fact, I believe the only way it can be done is the way it already has. We need to simulate the evolution of a strong AI and all its sources—essentially, us. I believe the evolutionary processes of an intelliegence—artificial or not—are as important as the end result.

This begins with recognizing the complexity of a human being’s intelligence. And I’m not even saying we’re that intelligent, but all of us, stupid or smart, are at least sentient. What is sentience exactly? It is by its very nature a combination of millions of complex things. Let’s use bullets:

• Our environment. Itself extremely complex, moreso perhaps than us. It includes:
• The natural world. Mountains, oceans, water, food, predators, prey, animals, plants, air, cold, hot, shelter.
• The man-made world. Cities, roads, buildings, etc.
• People. Society. Social interaction.
• Physics and science, “how things work.”
• Our physical bodies—also inherently complex.
• Hormones, chemical triggers, cravings.
• Physically manifested emotions, especially in social contact.
• The mind-body connection.
• Our perception of our own physical presence.
• Sensory input and response.
• Instinct
• Our minds.
• Memories, imagination
• Ideas and thoughts
• Connecting ideas with memories with thoughts and imagination.
• Connecting everything at once, almost automatically.
• Sensory perception
• Learning, especially from other people.
• Our evolution and genetic construction.
• DNA and its affect on all of the above.
• The environment, and its affect on DNA (evolution)
• Human effects on the environment through all of the above.

I have to say a little more about that last point—it’s not enough to say simply “evolution made us complex”—it made us extremely unimaginably incomprehensibly complex. We’ve probably got material in our genes that’s the way it is because some ancient squirrelly animal got eaten by a velociraptor. Our evolution was absolutely affected by the extinction of the dinosaurs and great climate changes. We’re entirely adapted harmoniously to our environment. We probably even have survival strategies dating back to some of the earliest mammals—if you ever doubt that, just think of fear and the kind of fright-or-flight response you immediately get.

See, this is how evolution works. It is the balance point in the middle of everything that simply allows life (I’m not trying to prove evolution here either; if you’re not with me this far, you might as well leave now).

We are a product of this incredible amazing process. This is life, this is intelligence, this is sentience. It’s not a computer program, it is the whole thing—not only us, but everything around us too, and every other us. It’s all one intelligence, and it would be different if any little thing went differently at any point in the last 4 billion years or thereabouts. We are the whole thing.

So talking about Artificial Intelligence in all these parts—of natural language processing, image recognition, even cognition and brain theory. It seems to me that no one gets it about the whole picture! Everything is important and you’ll never get a strong AI without understanding that!

I think it’s possible. What we need is a simulation of everything. But it can start simple, as simple as we did. With the beginning of the universe. Make some physical laws, design a universe program, get a computer that’s fast enough to simulate everything (for now at least, we can upgrade) and run the damn thing. It’s the only way to get everything.

No one wants to do that though, so we’ll keep focusing on the computer science problems behind it, and it might get us somewhere. It will not, though, get close to human intelligence without everything, quite literally. It will never think of crazy ideas while mowing the lawn without a lawn to mow (metaphorically speaking), and there’d be no lawn without, well, the rest of the world. Every little thing is part of it.

I believe it’s better to start with an understanding of these complex origins, and how important it is to our “program.” Perhaps a computer scientist with that realization could approach it from a different direction, and make it easier to bypass steps of evolution, or parts of the environment. Maybe cognitive scientists with a respect for complexity will figure out how to quickly evolve a brain in an appropriate system. Maybe multi-disciplinary biologists will think of a way to mimic the complexity of DNA. Historians will bring past human environments and societies to bear, Archaeologists will help with ancient physiologies and cultural clues, Sociologists will help us understand human interaction better, Writers past and present will provide content for an AI to learn, and Poets must help us understand our own humanity in this new context.

Every scientist must understand complexity in the 21st century. There is no other way. It is a shame, then, that we teach them so specifically when they should be trained as the worlds’ best generalists, with a special respect for complex systems. Science needs that now. It’s not just Intelligence that’s complex, but also our natural environment, our societies, our technologies, our weapons, our genetics; never was there a time in more dire need of scientists who understand this, nor in more danger from those who do not.