Sponsors

Hi! I'm Eliza!

Location: Berlin
Year: 1966 circa.

Once upon there was...? A wooden puppet –you might guess− who wanted to be real? Nice try!
Once upon a time there was a machine who wanted everybody to believe she was real, in flesh and
bone.
Joseph Weizenbaum is a German-American early computer scientist. Born in Berlin to a Jewish
family, he escaped Nazi Germany in January 1936, emigrating with his family to the United States.
Today, he just published a comparatively simple computer program called “Eliza” named after
Eliza Doolittle, the flower girl with the battered straw hat and cockney accent from George Bernard
Shaw's Pygmalion.


Eliza is an early example of Artificial Intelligence. Following in Turing’s footsteps, Weizenbaum
envisioned a machine-chatter program able to trick people into thinking she was real. To make
things easier to handle a conversation with a human, Eliza was programmed to be a therapist.
Indeed, a therapist is not supposed to give you advice, per se, but he/she would instead ask open-
ended questions. Through these questions, the therapist will lead you into saying more, so that you
will figure out by yourself, why you are laying on a couch and talking to yourself out loud.


I mean no stigma to professional therapists. To be fair enough, every IT guy is also a kind of
therapist who needs to survive in a digital world where none of us has all the answers.
Just to give you an example, if you talked about your father, Eliza would have replied by saying:
“Tell me more about your father.” This wouldn't work to replicate any human conversation pattern,
though. Imagine being in a place where everybody replied with a question after any sentence that
you said. You'd realize that a kind of sick game is going on or that, maybe, you miserably ended up
in an episode of “Twin Peaks”!

However, if you pretend to be in a stuffy and noisy office, it might work pretty well as confirmed
by some experiments done during the 60s when some folks were fooled by Eliza. They were told
that a real therapist was chatting with them from a computer keyboard and most of them believed it.
The lesson learned is that people living in the 60s were arguably a bit crazy, but that's OK. If you
watched the documentary film “Woodstock”, then you knew it already.
I briefly include a sample of a typical interview between Eliza (E) and a human interlocutor (U):

E: "I'm Eliza, how are you doing? Talk about your problem. "
U: "Are you a human?"
E: "Why do you care if I'm human or not?"
U: "I do not like computers."
E: "Do computers worry you?"
U: "Absolutely"
E: "Please continue."
U: "I'm scared of computers."E: "Why do we talk about computers?"
U: "Are you following me?"
E: "Would you prefer not to follow me?"

As you can see, it is unlikely to go longer than one minute with Eliza before realizing that it is a
robot. Mostly because it asks you things anyone is supposed to know already like “Tell me more
about a cup of coffee.” Imagine a dinner with your friends all speaking like Eliza: “Do you want
some more pizza?”,”I am very interested in pizza”, we would reply. “Please say more why you are
liking wine” and “why you just repeating words back to me?”
When I was a teenager, I spent a lot of time talking to Eliza. And not just because I was depressed,
though sometimes I was, but mostly because it was funny to discover “hidden words” that turned
the conversation into a new pattern.

I wasn’t alone in search of hidden treasures. So were the Gods of Economy, which smelled a good
business; a few years later, an invasion of the so-called “adventure games” filled the shelves.
Interestingly enough, those games featured natural language processing to solve riddles, using
complex commands like "take the key from the desk" or “kill the ogre with the sword.” We all
eventually ended up in a dungeon called “Zork” where "it is pitch black, and you are likely to be
eaten by a grue."


Sponsor 2

Post più popolari