Slashdot Mirror
The Profoundly Weird, Gender-Specific Roots of the Turing Test
malachiorion writes: Alan Turing never wrote about the Turing Test, that legendary measure of machine intelligence that researchers claimed to have passed last weekend. He proposed something much stranger — a contest between men and machines, to see who was better at pretending to be a woman. The details of the Imitation Game aren't secret, or even hard to find, and yet no one seems to reference it. This article explains why they should — in part because it's so odd, but also because it might be a better test for 'machines that think' than the chatbot-infested, seemingly useless Turing Test.
That would be a hard test. Most guys can not even pull it off. Lord knows I have seen a lot of guys online trying to pretend to be women who just can not even get the basics down.
You're asking a machine to mimic something profoundly alien to it's nature, to put things on an equal footing the man should have to do so as well.
And please, let's not get in to the similarity/difference argument. Yes, the similarities between the minds of men and women far outweigh the differences, but the differences are still profound.
--- Most topics have many sides worth arguing, allow me to take one opposite you.
It's more like a case of people pushing the goal posts back to where they originally were, after some over zealous AI zealot moved them forward to make his failure look like a success...
Which has more power: the hammer, or the anvil?
to understand how a machine could be made to pass the Turing test (or the woman test) honestly and thoroughly. To do so, it would have to understand arbitrary human statements and questions: not just "why is the sky blue?" (relatively easy) but "why doesn't my wife understand me?" and "is the real rational, as Hegel posited, and if so (or not) why (not)?" Note that the machine could reasonably pretend to know nothing about Hegel, but it would have to react like a normal human being. No obfuscation such as pretending to be foreign, a child, thoroughly ignorant, or befuddled by drugs should be accepted.
Going a little further, it would have to cope with (very) simple jokes such as "I asked my dog which team would win the World Cup"/"What did he say?"/"Nothing. HE'S A DOG".
I am sure that there are many other solipsists out there.
Arguably, the 'AI curse' is only partly goalpost-moving by humans obsessed with their special uniqueness.
The other aspect of it (as well a genuinely interesting result of whatever attempt at AI was made) is the discovery that we actually handle much less of what we do in some sort of naive, idealized, "high-level, general-purpose, cognition" than we naively suspect, and rather more in specialized and unconscious mechanisms(this lesson can be learned from the other direction as well: something like face blindness, in otherwise cognitively and visually normal people, simply wouldn't be possible if object recognition were a general-purpose function handled by an 'intelligence' with access to a video stream...) And, even when we aren't crunching stuff unconsciously on quite special-purpose mechanisms, we are sometimes just dumb. Good old Youtube comments would probably allow a babbling Markov chain to pass a Turing test.
There's also the fact that humans are frequently easier to fool than they would like: It's not exactly news that people will see faces on the moon, a known sterile rock, impute emotions (and sometimes entire spirits or deities) to the weather, emotionally bond with tamagotchis and similar nonsense.
Yes, people are likely to dismiss every AI that doesn't end up murdering them all and rendering them for computronium as 'just an expert system'; but it's also arguably the case that using a Turing test to check for AI substantially rests on the assumption, more or less Cartesian and more or less nonsense, that man is first and foremost an abstract 'thinking thing', with some other stuff tacked on that philosophers and mathematicians needn't really worry about.
"Passing" the Turing Test is about as valid as Asimov's Three Laws of Robotics.
Proverbs 21:19
Turing's imitation game was a thought experiment. To explain it simply to an audience unfamiliar with the idea, he started with man against woman, then proceded to man against machine. The specific genders were not important or significant in the thought experiment, just the existence of some difference between the contestants that could potentially be spoofed over a teletype. There was nothing gender-specific or weird about it.