In Smalltalk, two computer programs chat with each other on the subject the visitor has selected. The leading sentence which can be selected
on the touch-screen of the handheld computer interface sets off the conversation, in the course of which the robots try to expound upon
each other's reactions, and to keep the conversation going via the formation of increasingly appropriate responses. Their sentence-interpretational
capacities based on mechanical symbol-reduction, precisely due to their misinterpretations, are frequently capable of producing variegated chatter
for a longer or shorter period of time, which, upon entering into self-repetitions, reaches its end: upon registering such phenomena, the robots - in lack
of a better solution, often provoking arguments - try to end the conversation as soon as possible.
The leading sentences that may be selected on the touch-screen are,
in part, related to artificial intelligence and chat-robots, as well as to the chatter about the installation at hand, while the other vein
is initiated by László Krasznahorkai's book entitled, Sátántangó (Satan's Tango), and the superficial dialogue engaged in in the film directed
by Béla Tarr and based upon the same book. (In the next version - which is already under development - a couple of new topics will be added,
which are more understandable for the international public.)
The robots chat in English, and the spoken English words together with
their Hungarian mirror-translations appear in the form of subtitles on the image-screen. The constrained use of English can be traced to
the fact that the text-analysis system of the robots, functioning according to simple symbol-reduction and based upon linguistic-grammatical regularity,
works satisfactorily only in the case of languages that do not use declension and without a fixed word order.
The field of chatterbots simulating human intelligence is perhaps the
best-known, and certainly the most popular subject within the discourse
on artificial intelligence.
Alan Turing published his famous paper entitled Computing Machinery
and Intelligence in 1950, which contains the description of the "Imitation
Game", commonly referred to as the "Turing Test" nowadays. The game
has three players: a man, a woman, and an interrogator - of either gender
- who must determine which of the players is the woman. The interrogator
communicates in text-only mode and knows the players only by ambiguous
pseudonyms. The rules further require that the woman always tell the
interrogator the truth. The man, on the other hand, must always lie.
From the viewpoint of artificial intelligence issues, the obvious question
poses itself: what would happen if we were to exchange the male-player
with a computer in the game? Would the interrogator be mistaken more
or less often in identifying the gender of the players? In his classical
paper, Turing proposes replacing questions like "Can machines think?"
with the observation of this situation. Despite the fact that many theories
of artificial intelligence have attacked this basis of the Turing-test,
the "Imitation-game" is still the clearest method for measuring the
capabilities of chatterbots, and thus has remained as the only competition
sport of the world championship of palavering-machines, the Loebner-contest.
The very first and most famous chatterbot, Eliza , was written by Joseph
Weizenbaum in 1966. ELIZA was a psychiatrist, in particular, one that
posed analytical questions for every answer the user gave it. Though
sometimes they may have seemed ambiguous, people actually felt ELIZA
could take care of their needs just as well as any other therapist.
They became emotionally involved with ELIZA, so much so that even Weizenbaum's
secretary demanded to be left alone with the program. The success of
the program is based on the thesis that in a conversation, the content
of the actual spoken/written sentences is much less important than the
personality the human imagination projects behind them. Consequently,
the task of the robot-actors of Smalltalk is difficult: the on-line
chatterbots can always count on the human intelligence of their chatpartners,
which is able to understand an ambiguous, misunderstood reply as an
intelligent implication, or to switch a tired conversation into a new
inspiring direction. Although the Alice chatterbot written by Dr. Richard
Wallace , which is used in Smalltalk, has a much more sophisticated
language-processing system than Eliza had, she is not less dependent
on her chatpartner's imagination than the virtual psychiatrist. Any
one of us can have long and entertaining discussions with the original
Alice about virtually any subject, but her blabbing virtue fails after
a few sentences when she is confronted with a similarly mechanical intelligence.
Thus we had to equip the robots acting in the installation with specially
designed brains, which are much more striated within the fields the
user can select on the touchscreen than the average. The Smalltalk robots
are able to follow up the topics of the leading sentences selected by
the visitor and produce diverse and unique conversations, which resemble
the human social interaction of parties and exhibition-openings.