Questions and Comments.

Dateline: November 23, 1997

I THOUGHT you might like a break from the drudgery of reading yet another of my interminable essays.

OK, so I lie. I would like a break from the drudgery of writing etc.

To tell the truth, I nearly relented and wrote an essay after receiving the following encouraging email yesterday from Dr. Alan Peters at Vanderbilt University:

Subject: I thoroughly enjoy your ...
Date: Sat, 22 Nov 1997 01:29:35 -0600
From: Alan Peters <rap2@vuse.vanderbilt.edu>
Organization: Vanderbilt University School of Engineering
To: ai.guide@miningco.com

...pages. Lots of good things there. I just an hour or so ago ran into the mining company pages. What a neat place. I'll definitely be coming back and sending the URL around.

I enjoyed your critique of the "Mind" book. I've read Dennett (and Minsky). As far as I can tell from your review, there is nothing new in it. (Did he ever _explain_ how to pronounce !Xo~?)

Your robotics links are especially cool, but if I may be so bold, incomplete. Please check out

http://shogun.vuse.vanderbilt.edu/CIS/IRL/

My students and I (and three other faculty) are constructing a dual-arm humanoid (designed expressly to interact closely with people) to be an intelligent assistant. The robot also includes a mobile platform capable of autonomous navigation that has a single arm. The mobile and stationary platforms are designed to interact. Our humanoid, Isac, plays the theremin, and can draw. The arms on our robots are actuated by McKibben articial muscles, they act like human arms. Also, you might check-out the following article: Freedman David, H., "Chasing the Jetsons," Discover, vol. 18., No. 9, pp. 48-55, September 1997, in which Isac is featured.

Keep it going. It's good.

Alan

I'm posting this not so much to bask in glory, but because Dr. Peters provides information that should be shared. He's right: the Vanderbilt IRL website is thoroughly deserving of attention (i.e., it's cool!) And that brings me to my first request:

Please write to me if you (1) see any deficiencies in the resource listings or (2) have interesting information to share about AI-related projects. I'll gladly and promptly add stuff to the resource list, and pass on information to other readers through this forum. AI is such a huge domain, and there are so many links in the resource sections, that I can't hope to maintain them unaided.

A second question concerns "this forum." Right now, I'm maintaining the front page of this site, adding four or five interesting news snippets each week, besides writing my weekly essay. But thanks to the sterling (or should I say greenbacked?) efforts of The Mining Company staff, we can now expand the forum to include weekly chat sessions, a newsletter, and (soon) a bulletin board. The thing is, are you interested in participating in chats, receiving a newsletter, reading and writing to a bulletin board? To implement these things will take considerable planning and effort on my part, and I don't want to waste time if there's not enough demand out there.

Not being in an academic setting (would that I were, but I've spent the past 10 months on a personal sabbatical to write a book manuscript, now completed in first draft) I do not get to rub shoulders and debate ideas and issues on a daily basis with other minds. So I'm anxious for your feedback on the stuff I write in these articles. I'm only too well aware that much of what I write is open to challenge, so go ahead -- challenge me!

Finally, here's something I'd really appreciate your help with. You're undoubtedly familiar with the Turing Test, designed to test whether a given machine can be considered intelligent. I'm trying to devise a kind of obtusely reverse Turing Test, to determine whether an apparently intelligent machine is truly so or is, in fact, under human control. The problem I'm trying to solve is exemplified by an experience I had in a London department store one Christmas about ten years ago. On one of the floors was a mobile wheeled "robot" that would approach customers and chat with them.

Of course it was not a robot. It was just a glorified remote-controlled vehicle made to look like a robot, and of course it could not chat. It housed a radio transceiver, a speaker, and a microphone. A human operator hidden somewhere in the store was using a joystick to move the robot around, and a microphone/transmitter to talk to customers through the robot. It was a great hit, and fooled many people.

As my regular readers will know, I believe that we will one day have autonomous, intelligent, sentient, conscious robots, but the question is, how could you be sure, if you met one, that it was not just a department store robot? Put it another way: suppose you've just taken part as the interrogator in a Turing Test, and have just been told that the contestant you identified as the human was in fact the computer under test. How can you be sure it was not under some kind of remote human control, carefully concealed from the test organizers?

If you think about it, the answer is not easy. Not to me, anyway. In fact, I don't know the answer. Which is why I am asking you!

Until next week,

NEXT WEEK: Figments of Reality: Notes from my reading of Ian Stewart and Jack Cohen's recent book.

Previous Features