Artificial Consciousness: What Does It Take?

In this series (intro here), I have suggested that consciousness consists of thinking about thinking, and that thinking consists of symbol-manipulation. In the last post, I presented a case that our biological brains do not have to manipulate the symbols. Artificial brains would do.

But how advanced must those artificial brains be in order to qualify as conscious?

Because consciousness is a matter of degree, artificial brains could be specialized and still be conscious. The key ingredient, I have argued, is that a conscious being is able to think about its own thinking. Technically, this means that the input to its symbol-manipulation includes its own symbols.

Computer program can do exactly that.

For example, some chess-playing programs learn from their mistakes, in effect reprogramming themselves based on experience. To be sure, this is extremely limited consciousness, but our consciousness is also limited, is it not? We are unconscious of the earth’s magnetic field, but some migrating birds do seem to have this awareness. Would such a bird label us “unconscious” because we are unaware of something so fundamental in its world?

If a computer program that is only conscious of itself as a chess player is not good enough, consider the article in Time Magazine last February. Titled 2045: The Year Man Becomes Immortal, it projected that computers’ intelligence will exceed ours within 35 years. We already  have a computer program that beat two Jeopardy! champions. What will the world be like when computers beat humans at every creative pursuit?

When computers are able to process every sensory input better than we can, as well as some we can’t; when they are able to reflect on what they’ve learned faster and more accurately than we can; when they can think about their own behavior and adjust it for optimal results without even seeing a psychiatrist; when in fact they are better psychiatrists than we are — when all these things take place (and they will), will we be ready to concede that software can be conscious?

Maybe a better question is, will computers think that we are conscious?