Meet TOPIO 3.0 (TOSY Ping Pong Playing Robot). It’s a big step for AI, but can a machine ever be conscious?

What makes humans conscious? As mere collections of organic matter, it’s pretty impressive that we’re even able to ask ourselves this question. What is it that makes our power of self awareness possible?

Setting aside the task of defining what consciousness really means, where does it come from? Here are the options:

1. Humans are not conscious – it’s just an illusion.
2. Consciousness comes from something physical (the brain).
3. Consciousness comes from something non-physical (a soul).

These are the only possibilities, right? Let’s look at each one.

1. Humans are not conscious – it’s just an illusion.

If this is the case, then we immediately run into an apparent contradiction. How is it possible to think about whether you’re conscious without actually being conscious? This is very similar to the argument behind “I think, therefore I am.”

Still, I don’t think we can rule it out entirely. Sometimes I look into the mirror and wonder, “Are you really me, or do I just think you are? And if the latter, do I really think you are, or do I just think I think you are?”

2. Consciousness comes from something physical (the brain).

This seems like a simple and obvious answer, but it’s really not. Because if consciousness comes from the brain, then there’s no reason we can’t build a conscious machine simply by replicating the brain with mechanical parts.

Sure, we’re currently far from having the technology to interconnect 100 billion artificial neurons with 100 trillion artificial synaptic connections. But technology has improved a lot over the last hundred years. What will happen over the next million?

I wouldn’t be too quick to predict limits on technological progress. When we build semi-intelligent nanobots, won’t they be capable of helping us build more intelligent nanobots? And then won’t it be easy to build a brain far better than what we have now?

But no matter how advanced future technology may be, I still have a hard time seeing the leap from artificial intelligence to artificial consciousness.

Artificial intelligence is easy. Even if a program isn’t truly intelligent, it can easily give the appearance of being so. As an example, consider Harold the tic-tac-toe AI.

Harold is a tic-tac-toe program I wrote the other day to test out an idea. He’s really not intelligent in any sense. For example, he doesn’t understand (nor will he ever learn) that if you put two X’s in a row, he needs to block you. All he does is make arbitrary decisions, then he sees what happens.

If he ends up losing, he knows he made a mistake, so he won’t do that again. And if he ends up tying, he knows to try something different next time, in case a win was possible. It’s a bit agonizing to wait for him to learn by playing out all the different variations, but after he’s made every mistake once, he’ll play perfectly.

Now, if Harold just played out all these games in his head before playing against a human, he’d have the appearance of being intelligent. And given an arbitrarily high processing power, there’s no reason we couldn’t generalize this concept to have him play perfect checkers, chess, or indeed solve any problem that had well-defined rules and goals.

And that’s without having any actual intelligence at all – just brute force and a good memory. The possibilities will become really interesting when we start making significant progress on true AI: deduction, reasoning, problem solving, knowledge representation, planning, perception, creativity, etc.

We still have a long way to go before you can have a conversation with a robot without figuring out he’s a robot. But I don’t think intelligent robots (or at least, robots that appear intelligent for all practical purposes) are much of a stretch at all in the very long term. Last April, a program extrapolated the laws of motion from a pendulum’s swings, including conservation of momentum and Newton’s second law (F = ma), without having been programmed with any knowledge of physics.

Let’s say we get to the point where we have true AI. You can talk to a robot (or really, just a program – an AI doesn’t need a body) and fully enjoy its witty banter, beautiful poetry, insightful Zen koan interpretations, etc. It’s still just a program, right? It has no sense of awareness or subjective experience.

Can you imagine a program being truly conscious? Wondering what its life purpose is, whether this external hard drive makes it look too fat, and when it will finally get the right to vote? Moving it to the recycle bin would be kidnapping, and deleting it would be murder. Ridiculous, right?

3. Consciousness comes from something non-physical (a soul).

This would provide a nice answer to the previous question – a program can’t be conscious, because it doesn’t have a soul. Of course, this option comes with its own problems, not the least of which is that it’s a severe violation of Occam’s razor.

The absence of supernatural phenomena is the simplest possible explanation, and therefore most likely to be the correct one. Unless, of course, it’s too simple to be possible.

When a car shuts down from a dead battery, you just put in a new one and it comes roaring back to life. Why doesn’t the same thing happen with people? If someone dies from a heart attack, why can’t you just repair their heart (and anything else that may need it) and watch them come back to life?

What part of them has really died, if all their organs are completely intact? Why doesn’t Frankenstein work in real life?

Then again, maybe it does. The real problem with a fatal heart attack is probably that it causes brain death, and nerve cells aren’t easy to repair. But is this just a matter of technology?

Can we someday inject nanobots into someone’s bloodstream, having programmed them to repair any and all cell damage, and expect the person to live indefinitely in perfect health, not even aging?

Which of these is the right answer? I have no idea.