I don't know if any of you heard of this, but there is a contest every year dealing with AI. The rules are simple: Create a program to converse with judges, run the program, and at some point during the "conversation" switch it with a human (or maybe it's the other way around? Could someone help me with that, if it isn't going out of their way?). The scoring: There are three medals. I forget the criteria for the bronze, but the silver is awarded when at least half the panel of judges think the computer is a human when conversing with text, and a gold is awarded when a computer fools the judges talking with speech, not text. Anyway, just a tidbit I thought might be relevant to this forum.
End of line.
-Snoopy
hmm, i think that the competition is to trick the judges into believing that your computer is in fact a human...
Just wondering, how would you differentiate between human and computer, while talking to them? What signs would you look for that the subject is human?
You would probably look for text and/or speech articulation. With text: Speed, grammar, expressions, legible sentences that are on topic and make sense, and maybe even the occasional misspelling (Get it? It's supposed to be mispelling. Ha, ha, ha). With speech, you would need one of those new, next-gen speech synthesizers that will change tones, pitch, volume, and timing depending on position in word, homophones, accents, and stuff like that. You'd look for accent, sound (does it sound like a human, or Stephen Hawking's computer?), speed, articulation, no "hiccups", like in Microsoft Sam's voice, grammar, sentences that make sense and are on topic and maybe the wrong pronounciation every once in a while. This is just a guess, but that's what I would look for.
This strikes me as a very poor and highly flawed contest. Even the judging is left open to massive interpretation. I cant imagine this is taken too seriously in the communitiy. A much better platform for this is the desert run event that DoD and NASA personnel host every year.