Flunking the Turing Test


I’ve been interested in computers since I watched them on Star Trek reruns when I was a kid.  Back then in the 1970’s, computers weren’t exactly interactive.  I remember when my friend Greg got a program called ELIZA for his TRS-80.  It was an early attempt at artificial intelligence.  Mathematician Alan Turing tried to answer the question whether or not a computer might be considered intelligent.  He came up with the “Turing Test”.  Basically, if you could talk to a computer and didn’t realize it was a computer, the computer would be considered intelligent.

Greg and I could type in statements in English and ELIZA would respond back in English.  While that was a novelty (generally all typing on the TRS-80 was in BASIC), the actual conversations with ELIZA sound more interesting than they really were.  The original ELIZA program was written in 1966, the same year Star Trek debuted on television.  The TRS-80 version of ELIZA struggled to make conversation from whatever words could fit into 4,000 bytes of memory.  Pretty much anything we typed in was answered with something like “That’s interesting, please go on.” or “Why do you say that?”  It was like having a conversation with a Magic 8-Ball.

Surprisingly, the concept behind ELIZA – having a large pre-programmed list of words and phrases – is still in use in artificial intelligence programs today.  The major difference is speed and scale.  Even a basic system today can hold a million times as much data and access it in nanoseconds.  I was curious to see how these new artificial intelligences matched up against old ELIZA.  I found a few so-called “chatterbots” online.  They all seem to work the same way.  A chat window is available on the page.  You type in English and it answers in English.

The first one I tried – Yeti – was pretty bad.  Everything it answered with was all in caps, so it seemed like it was shouting random things at me.  I guess it could have passed for human if the person was seriously deranged. I had a better time with Jabberwacky.  I managed to talk back and forth with it a half dozen times before things started to get a little… off.

JABBERWACKY: London is my favorite place in the world.
ME
: Why?
JABBERWACKY: My friends don’t like me.  😦

I lost my connection to Jabberwacky shortly after that.  I hope it didn’t do anything rash.

After talking to machines that were obviously machines, I thought I’d try an actual Turing Test.  I was supposed to have a chat for five minutes.  Sometimes the site connects to actual people, other folks talking the Turing Test.  And sometimes it connects to a machine.  My chat partner was named “Landru”.  We had a very pleasant chat for about a minute and a half before he froze.

Turing Test

Apparently Meka was right.  I watched too much Star Trek as a kid.  Like Captain Kirk in so many episodes, I am able to bring a machine to its metaphorical knees just by having a discussion with it.  So, note to NASA: if Pioneer 11 shows up again, rebuilt as a planet killing death machine, give me a call.  I should be able to take care of it in a jiffy.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s