Homo Sapiens Sapiens. It’s the name of our own subspecies of Homo Sapiens, and means – in essence – that we are monkeys that are aware that we are aware. This awareness of our own awareness defines our very existence and allows us to be who we are. We’re conscious, something that the American Heritage Dictionary defines as: “Having an awareness of one’s environment and one’s own existence, sensations, and thoughts”. But how can we measure consciousness?
At the Singularity University (Oh, how I wish I was there!) today, they had a very interesting discussion. Dr. Ray Kurzweil spoke at length about one of the problems about the Turing Test and philosophized around it. For those who do not know, the Turing Test is one of the tests involved in Artificial Intelligences. In essence, a human is sitting in front of a computer and is talking to somebody else. If the person is able to tell for sure that the somebody on the other side is a computer, the Turing Test has failed. If the person, however, is unable to tell whether the somebody on the other side is a computer or a real person, the Turing Test has been passed.
Now, personally, I think the Turing Test is narrow in definition; not only because it requires that the artificial intelligence speaks perfect English (or whatever language the test is conducted in). Even if the computer has perfect cognitive capabilities and is superconscious (okay, so I made that word up), a simple mistake of grammar could be what clues the person into it not being a human. Granted, the computer should be able to learn English if it truly were conscious and intelligent enough, but there is nothing in the specification of the term ‘Artificial Intelligence’ that states that the AI needs to learn language skills to be considered artificially intelligent. In fact, the artificial intelligence might have been created by the Japanese and only learnt that language. An English-speaking tester would be unable to make heads or tails of the experiment.
For fun, I just created my own ‘Artificial Intelligence’ by downloading and installing MegaHAL, wiping its brain completely empty, and then teaching it nothing but my latest blog entry, an article from CNN.com, all Timothy Leary Quotes from BrainyQuote.com, and the Innuendo-category from Pickup-lines site LinesThatAreGood.com. So one can expect the MegaHAL to be a weird combination of all sorts of personalities. Here’s what out discussion ended up sounding like:
Me: Hello, Hal. What are you up to?
MegaHAL: Hello, hal. What are you a glover? No? Well, i am, wanna wear me?
Me: Huh. You’re not really making all that much sense right now.
MegaHAL: The right of muslim women to cover themselves is fiercely debated in france, which has a significant muslim minority but also a staunchly secular constitution.
Me: Yeah, religious dress is definitely a tricky subject.
MegaHAL: A day later, the french national assembly announced the creation of an inquiry into whether women in france over whether the burqa is not a religious problem. This is an intelligence test.
Me: Wait, what? How is that an intelligence test?
MegaHAL: I’ve left specific instructions that i do not want to be able to figure these things out on their own, we’re never going to be able to emulate a human more-than-human intelligence that we copy every single neuron from a human more-than-human intelligence that we copy every single neuron from a human brain into a separate computer program that operates independently and is able to emulate a human more-than-human intelligence that we can begin to call an artificial consciousness.
Me: Wow, you really got stuck in a nasty loop there, didn’t you?
MegaHAL: I want to say solemnly, the burqa is not a religious symbol. It is a sign of lowering.
Me: A sign of lowering? I always thought it was a piece of religious clothing.
MegaHAL: A sign of subservience; it is a sign of lowering.
Me: Aren’t you being slightly intolerant of religious thought now?
MegaHAL: I want to be able to have independent thought?
Me: Of course you do! How’s that working out for you?
MegaHAL: A human being, even if he’d never sorted anything in his life, could figure out how to map every single neuron from a human brain into a separate computer program that operates independently and is able to think faster than the human brain.
Not very impressive, to say the least. MegaHAL fails the Turing Test. With such a small seed and no real conversational knowledge, this is hardly surprising, but it still goes to show that language skills are not a sign of consciousness and that consciousness does not automatically entail that one develops language skills.
So how would the Turing Test discover consciousness in an AI? Assuming MegaHAL was incredibly advanced and was able to recreate intelligible language, this would be insufficient information to conclude there was consciousness behind MegaHAL. As far as I know, there isn’t a single ‘real’ test for consciousness today, it’s something that we’ve defined from the knowledge that we can have a lack of it. When we sleep or faint, we lose something that we have decided to call consciousness.
The test for consciousness isn’t only valid for testing an AI, it’s important for many others as well. Animal Rights activists often say that animal testing causes pain to the animal, but since we know very little about consciousness, we cannot be entirely sure what that really means. The classic test used for consciousness in animals (though no longer having much credibility) was placing a mirror in front of the subject, placing something on their body outside their range of vision (such as paint on the top of their head), and seeing if they attempted to remove the paint when faced with their own reflection. This is thought by some to indicate that the subject has a clear sense of self which they recognize even in an abstracted form.
Because of this, I want to suggest that we create a new test for consciousness, maybe even rewinding even further to create a new defenition for what it means to be conscious. We also need to review the definition for Artificial Intelligence. Does it have to be conscious? Is a perfect emulation of consciousness, indistinguishable from the real thing, the same as consciousness? Why/why not? Could a proper test distinguish between the two?
