Well, what would you do, given a choice between spending a weekend typing up a mundane manuscript for someone, or taking Saturday off in favor of being available on Sunday to take part in the Fourteenth Annual Loebner Prize for the bot that comes closest to seeming human in a Turing Test? I knew you would! So did I, foregoing the more remunerative typing job for the more interesting opportunity to convince a number of judges that I, not some machine, am the human. At stake for the entrants: $2,000 of Loebner's money, a bronze medal, and bragging rights until the next year's contest.

Four bot entrants were set up in Dr. Loebner's apartment on the Upper West Side of Manhattan. The test required 4 human "confederates," 4 judges, 2 rooms, 12 computers, and Hugh Loebner himself overseeing the proceedings, from greeting the participants at the door to ensuring that the programs were started correctly, and everything in between. Also present were a number of bot programmers themselves and a few onlookers.

As it happened, my job was switched from human interlocuter to judge; I would join three others to try to determine which in each of four pairs of human/computer was the more human-like, dividing 100 points between each member of each pair, for a grand total of 400 points among them all. Suddenly the pressure was on: what would be best to ask of each pair to distinguish between machine and human? Should I attempt to have the same conversation with each one, or would it be all right to try to discuss something different each time? I had some time to think about it: there were seven rounds of testing, but each judge was needed for only four rounds; I had the first two rounds free to think things over.

There were too many distractions on the painting- and photo-filled walls in Dr. Loebner's apartment, however, for me to spend much time overanalyzing things. Nor need I have much concerned myself: once I sat in one of the "hot" seats and began conversing first with A (machine? human?) then with B, then A again, and so on, I discovered just how quick and easy it was to determine which was who or who was which. One of the programs announced itself to me early on, proudly telling me the name of its Instructor/programmer as well as just how many words in how many categories it knew; another of the bots gave itself away for certain when it began suddenly spewing a string of totally unrelated words IN ALL CAPS; a third revealed its bothood when it complained that "Oklahoma" was not a helpful association for "OK" then sealed the deal by declaring that Novgorod equals 8; the fourth one was actually doing pretty well until it began repeating that it had no idea, no clue, and was really bored. Okay, I felt kind of insulted by that one.

In the end, no bot received more points from any judge than any human had received. As for me, the bot I gave the most points to was the one that was bored, and the one I gave the least points to was the one that had blurted out the name of its programmer. For a while, it even seemed as if I knew whereof I spoke, as the bot I had favored received the most overall points and was declared the winner ... momentarily ... until one of the human 'confederates' pointed out that its communications program had garbled her input. Indeed, it had garbled all the human confederate input, and so, alas, it was declared disqualified. Instead, the program which I had presented with a score of zero had the next highest overall score and accordingly won the coveted prize. Obviously, A.L.I.C.E., programmed by Dr. Richard Wallace, had not been quite as forthcoming with its true identity with the other judges as it had been with me. Do I bring out the bot in a bot, or what?

Wine was served afterwards, photos taken, congratulations passed around with the snacks. All in all, I considered it a delightful way to have spent a weekend in New York. And I hope I can do it again, next year, too.

Talking to a bot -- trying to convince it to convince you it might be human -- is rather fun, I found. Have a look at the transcripts from the 2004 competition ... and then go ask 2004 Loebner Prize Winner A.L.I.C.E., or additional contenders Toni, Alex and Jabberwacky.