www.mitchellhowe.com


The Singularity Q&A

Q: What are some noteworthy examples of "imposter AI"?

A: These are just a few...

AARON

This program is at once fascinating and disappointing. Billed as the first “cybernetic artist”, AARON creates “original” artwork in real time. (You can watch it draw right now by clicking here.) The remarkable thing about AARON is that it is unlikely to ever create the same work twice, and yet each work adheres to basic rules of “balance” in composition, and hence appears at least vaguely artistic – as opposed to randomly generated brush strokes. But if you spend even five minutes watching AARON draw, you will probably grow tired of observing the same loose patterns governing the creation of people and potted plants – the subjects that seem to comprise the whole of AARON’s repertoire. It is as though a human short-order artist has a vision of a particular painting, and is spending every waking moment trying unsuccessfully to paint it. The result: thousands of very mediocre paintings that fundamentally resemble each other too closely to be independently interesting. AARON’s creator, Harold Cohen, has spent over 30 years tweaking this program, which should give you some idea of how hard it is to fake the visual creativity of a human-like intelligence.

ELIZA

The grandmother of all “chatterbots”, ELIZA was created in 1966 by Joseph Weizenbaum, partly to mock the very Turing Test it was designed to pass. Consisting of just 200 lines of code, ELIZA responds to textual statements by looking for key words and turning the content back on the user, in the style of a psychologist. Because ELIZA does not learn or remember anything from one statement to the next, it is quickly unmasked by anyone who already has reason to be suspicious. But many have been fooled, at least for a short time, into thinking that they were communicating with an interested human. Weizenbaum was alarmed at the initially enthusiastic response given to ELIZA by many psychologists – he was, after all, trying merely to demonstrate that the Turing Test may reveal more about humans than about AI. The uncomfortable truth, rediscovered by many of us, is that human conversation is often as devoid of any real intelligence as ELIZA itself. (You can chat with ELIZA here, among other places) .

Ramona

Ramona is, quite frankly, the most annoying feature of futurist Ray Kurzweil’s website [No longer pops up when you go to the site... thank goodness], which is otherwise quite good. If you take the time to install the required “Life FX Player”, the popup window that appears anytime you visit this site will contain a talking head spitting out the textual responses of this modern chatterbot. Ramona is essentially a friendlier, better rehearsed version of the ELIZA program that also combines features of an internet search engine. While users can expect to become impatient with Ramona very quickly, the experience is illustrative as a glimpse of a primitive “personal agent”, a type of narrow AI some have predicted will become a popular means of intuitively using the internet. Go ahead. Ask Ramona what movies are showing in your area, or what the weather is supposed to be like tomorrow.

Other Chatbots

It should be noted that many people customize chat bots in their spare time, often as entertaining additions to online chat rooms. In the 40 years since ELIZA, hundreds of bots have been created or modified in many shades of personality. This link provides an index to some of the more polished ones.




[Back to Futurism] [Back to Main]  ©2002 by Mitchell Howe