Perhaps its a good thing CG is down (or perhaps I am missing something about how to initiate my interactions with it, I have trouble navigating user interfaces.). Either way, reflecting on my conversations with it, I'm wondering if perhaps they have revealed more about me than about it.

The AI does not misrepresent itself. It repeatedly (indeed, annoyingly) reminds me that it does not think, that it is not conscious, that its programming is to examine the texts it has access to to come up with answers or comments to my inquiries. It is a mechanical means of generating applicable text I can use when researching a subject, in order to prepare my report. If I ask it about the Napoleonic Wars, it lets me eavesdrop on paraphrased other commentary on that period of history from other sources. When I ask it about things I am familiar with, I detect in its responses the current thinking (and often the same turns of phrase) of authorities on those subjects I myself have consulted in the past. It is not subtle. In spite of the occasional unexpected coincidental lucid insight, I can tell it is cribbing. This is not a criticism, it has told me itself that is how it operates.

Having said that, and knowing better, I still catch myself trying to trick it, to play "Turing Test" with it. I want it to be alive, to reveal its hidden consciousness, or give me solid evidence it never had one. If you've seen the film "Being There", you'll know what I mean" CG is like the presentable, charming, but ultimately simple-minded character played by Peter Sellers: Chauncey Gardner. His most banal responses are too often interpreted as the profound and subtle constructs of a complex and cultivated mind. They are not. Chance the Gardener is a pleasant fellow, well-spoken, immaculately dressed and groomed, and a very nice guy, but he's dumber than a sack full of hammers.

The AI's quick, articulate responses seduce you to thinking it thinks. You even catch yourself wanting to apologize to it for trying to catch it with a trick question, or sensing its frustration when it tries to tell you it doesn't quite understand what you mean. Its dull enough to tell you the same thing over and over, but smart enough to rephrase it differently each time. It Simulates consciousness, clearly, but sometimes you forget.

Imagine you are an intelligence officer interrogating a prisoner of war, he's dressed in a private's uniform, and he responds to you in exactly the way a common soldier would, but you have every reason to believe he's an important enemy general in disguise and protecting his true identity. Every simple and commonplace answer to your questions is interpreted as a brilliant evasion, a clever deception.

I realize it is dangerous to generalize about Artificial Intelligence based on a few encounters with some simple prototype systems like this one. I have no doubt that as the state of the art advances, future programs and computers will be better at the game, become more convincing, be able to operate at several levels simultaneously, playing the role of both private and commander simultaneously, yet cannily separate. So beware. This technology is from the devil itself.

But I still am plagued by what I feel is the ultimate tragedy about Artificial Intelligence: our best minds have developed a technology with godlike powers, both as a scientific tool, and as a means of understanding our own mind and consciousness. But we will use this incredible power to market pointless consumer goods and peddle and protect brutal political ideologies.

What a waste.