This was originally posted 16 years ago. Opinions, facts and reality may have changed.
Now, this might not sound like a big deal but it could be: the Turing Test has been passed.
Woah, not so quick, I promise this is interesting.
In 1950, British mathematician Alan Turing – whose previous hits included cracking the Enigma code during World War II – developed a test to ascertain whether a machine was intelligent.
In its simplest terms, a person would “chat” to a machine and a human, via a keyboard and a screen. If, after a pre-determined period of time – say 30 minutes – the person couldn’t tell which was which, or indeed believed the machine was human, then the machine would pass the Turing test.
And something just has. Or should that be someone? When do “inanimate” objects gain the right to be considered sentient? Anyway, her name – the chat bot in question – is Suzette. And after 25 minutes chatting to a judge, he guessed wrong – believing Suzette to be human.
If you have any experience engaging with these chat bots, you’ll know the tricks for catching them out. You ask a question to test their vocabulary and understanding, and they’ll fire back with a “What do you mean by do I like dancing in the street with pandas?” (or whatever other inane question you’ve asked to elicit a response) or they divert the subject to something else (“Pandas are nice but I’d rather discuss the weather…”).
Well, apparently that’s what Suzette did: parry, tangent, ignore, repeat to fade. And the judge believed she was human.
Now, it’s easy to mock. And, frankly, it seems like an opportunity wasted if I don’t. For reasons too lengthy to explain (which could also alienate the few friends I have that don’t consider me a complete geek), I understand the ins and outs of the Turing test and, yes, it’s easy to say “oh well, it’s obvious it’s a chat bot” when you start chatting to Suzette – because we already know before the conversation starts.
It’s like knowing the “whodidit” before watching the Sherlock Holmes film. It’s all so obvious when you know. So, let’s allow the judge a little leeway before we begin mocking him (I’m assuming it’s a he for some reason).
Right leeway over…
The thing is, human conversation doesn’t work in the way that Suzette and the dozens of other chat bots attempt to emulate it.
If I was to ask a friend if they liked dancing in the street with pandas (how I wish my examples weren’t so random sometimes), then they’d undoubtedly ask me what they hell I was talking about.
My friend would understand the words, the concepts I was communicating and would be intelligent enough to know that, frankly, I was talking complete bobbins.
A chat bot doesn’t have this intelligence.
It may be able to recognise the words I’m writing but the chances of it being able to decipher what precisely it is I’m talking about… well, it would be a strange programmer who put this particular sentence into its database.
And that’s why, in my humble, amateur scientist/psychologist, opinion, we’re pretty safe from hordes of self-aware Terminators. For now anyway.
Because if a chat bot can’t deal with the randomness of the human brain, it won’t be rocket science (literally or otherwise) for us humans to maintain the upper hand.
So you can sleep safely in your beds tonight. The Turing test may have been passed but I’m starting to think it was a technical knock-out only.
(All this said, knowing that the Ministry of Defence’s military satellite network is called Skynet does trouble me somewhat…)