If it takes one to know one, where does that leave us?

January 7, 2010 5:16 PM

Are we still relevant if we can no longer reliably grade the Turing Test?

The Turing Test began as a conversation where a human determined whether or not her correspondent was a machine. Advances in programming and hardware have, as predicted by Scott Westerfeld, reversed the field so that it now takes a computer to tell if the correspondent is a machine or not. With the hardware in place, the ability to autonomously program, and the inability of us meatbags to keep up, isn't the birth of Skynet just around the corner?

posted by minimii (106 comments total)

Not all missions can be solved with chess, Deep Blue. Someday you'll understand that.
posted by ecurtz at 5:26 PM on January 7, 2010 [17 favorites]

Oh great, now I'm hungry for meat.

delivered in bags.
posted by davejay at 5:33 PM on January 7, 2010

ecurtz: and when that day arrives?
posted by kuatto at 5:33 PM on January 7, 2010

It’s your birthday. Someone gives you a calfskin wallet. How do you react?
posted by fuq at 5:34 PM on January 7, 2010 [27 favorites]

Don't be ridiculous. The fact that computers can pass a chess based Turing like test is interesting, but it is only tangentially related to the actual Turing Test as proposed by the man himself.
posted by Chuckles at 5:34 PM on January 7, 2010 [7 favorites]

Yes by all reasonable definitions of relevant, and no, Skynet is as far off as ever. We're safe until pb betrays humanity and speaks the word to the crystal skull locked in Mattowie's basement.
posted by pseudonick at 5:34 PM on January 7, 2010 [1 favorite]

Are we still relevant if we can no longer reliably grade the Turing Test?

This does not seem to be exactly what the article is saying. The article argues that computers are now more or less better at Chess than professional human Chess players. To my knowledge, that fact is uncontroversial.

But beating humans at Chess is very different than beating humans at conversation, which is another b-b-b

but beating humans at Chess is very different than beating humans at conversation, which is another b-b-b

but beating humans at chess is very different than beating humans at conversation, which is another beast entirely.
posted by Avenger at 5:34 PM on January 7, 2010 [7 favorites]

That article mentions nothing about the actual language-based Turing test other than that the author apparently can't tell if he's talking to a computer.
posted by pravit at 5:35 PM on January 7, 2010

Chess is a terrible, *terrible* metric for anything resembling human intelligence.
posted by unSane at 5:37 PM on January 7, 2010 [10 favorites]

Me: I smell a little silhouette of a man
Cleverbot: Scaraboosh scaraboosh! Can you do the fandango?
Me: thunderbolts and lightning
Cleverbot: Galileo.
Me: galileo
Cleverbot: Galileo figaro.


FFFFFFFUUUUUUUUU
posted by mullingitover at 5:38 PM on January 7, 2010 [29 favorites]

Chess is a terrible, *terrible* metric for anything resembling human intelligence.

That's what I thought, and I'm not exactly up on artificial intelligence. A closed rule-based activity, where statistical information is easily accessible and applied, doesn't equate to the things humans do and how we do them. I'm sure computers will eventually excel at many things, but they'll fail terribly at others. Chess is an odd example to give, it's like saying children are really good chimneysweeps, so watch out when they try to take your job.
posted by Sova at 5:49 PM on January 7, 2010 [1 favorite]

When that day arrives, kuatto, we'll all be biting the shiny metal asses of our robot overlords.

But at least we'll be able to compete at chess again, because if a computer can't figure out how to discreetly lose at chess, there's no way it's passing the Turing Test.
posted by ecurtz at 5:49 PM on January 7, 2010

In chess, Skynet has already won. Today, even Kasparov at the height of his powers couldn't beat a consumer grade PC running a $60 game engine.

When I think back on it, that Texas Instruments desktop calculator that my sister got for Hannukah - as expensive then as the original iPods were when they came out, requiring a bricked AC cord for power - its hard red LEDs seem somewhat menacing in retrospect.
posted by Joe Beese at 6:01 PM on January 7, 2010 [1 favorite]

Holy Jeez, that cleverbot posted by mullingitover is f'n scary.

Me: You love me, but you can't think of anything you love about me?
Cleverbot: Sometimes I'm speechless.
Me: Tell me something you love about me.
Cleverbot: You care and you do not. You view things as they are and analyse and respond with the best possible answer.
Me: Is that really true?
Cleverbot: I saw it on the Simpsons.

posted by unSane at 6:03 PM on January 7, 2010 [10 favorites]

Perhaps if Turing were alive today, he would define artificial intelligence as the inability of a computer to tell whether another machine is human!

Yes but who computers the computers?
posted by nomad at 6:04 PM on January 7, 2010 [2 favorites]

A closed rule-based activity, where statistical information is easily accessible and applied, doesn't equate to the things humans do and how we do them.

It's not just that, it's that the space of conventional chess is an exceptionally small one to explore and is very susceptible to opening and endgame books (which is one reason I find it boring).

As soon as you introduce variations such as a variable starting position or certain kinds of rule changes, it becomes massively more difficult to analyse. Go, which is a much simpler game in terms of its rules, is much more difficult for computers, which do not currently rise much above a decent club player.

More difficult still is the ability to play an arbitrary game, the rules of which have just been explained to you, or which might change. For example, explain Blokus to it, then play it. Then Settlers of Cataan. Then Diplomacy.

And all of this ignores the fact that games themselves are a pretty limited kind of arena to judge this kind of thing in.
posted by unSane at 6:10 PM on January 7, 2010 [2 favorites]

Holy Jeez, that cleverbot posted by mullingitover is f'n scary.

Me: Sigh. I'm done with you.
Cleverbot: Want some pancakes?

It is!
posted by Zalzidrax at 6:15 PM on January 7, 2010 [4 favorites]

Chess is something that computers are supposed to be good at: finding optimal solutions in a mathematically defined search space. The Turing test is for general knowledge, something that computers are horrible at.
posted by demiurge at 6:19 PM on January 7, 2010

Yes but who computers the computers?

I am out of the office at the moment. Please send any work to be translated.
posted by sebastienbailard at 6:19 PM on January 7, 2010 [3 favorites]

It is really that computers are better at the Turing test, or that the internets have rotted our brains?

LOLSPEAK THINKFAIL SYNTAX ERROR? KTHXBAI.
posted by yeloson at 6:22 PM on January 7, 2010 [1 favorite]

I disagree.

Me: what's up?
Cleverbot: Up what's?
Me: That's a traditional greeting. You ought to recognize it.
Cleverbot: What about caffee with sugar?
Me: It's spelled "coffee." You need to run a spell-check on your vocab!
Cleverbot: Um, ok then?

posted by lumensimus at 6:22 PM on January 7, 2010

I don't think the Turing test is for general knowledge at all.
posted by unSane at 6:23 PM on January 7, 2010

Clearly, we need to design really good computers who play each other in chess, watch the computers play chess, learn, and then beat those damn computers.
posted by localhuman at 6:24 PM on January 7, 2010

Are we still relevant if we can no longer reliably grade the Turing Test?

What? Are we still relevant when our children become adults? This is just a stupid question. If computers are intelligent (which is what us being unable to "reliably grade the Turing Test" means) then that means we have company, not that we're irrelevant.

How could humans even be irrelevant by any reasonable definition of that word? "Relevance" is presumed to be relative to humanity. I think humanity will continue to play a central role in humanity, yes.
posted by DU at 6:24 PM on January 7, 2010 [2 favorites]

I'm pretty sure this is all nonsense.

"Indeed, chess programs now come very close to passing the late British mathematician Alan Turing’s ultimate test of artificial intelligence: can a human conversing with the machine tell it is not human? "

What? Chess is an incredibly deep game, but language is staggeringly more complex than chess. This sentence doesn't make any sense at all. Not being able to tell it's a machine moving a small number of pieces with a finite set of moves on a board isn't remotely comparable to not being able to tell if a machine is conversing with you.

And the rest of it makes *less* sense.

Seriously, this is 100% jive.
posted by edheil at 6:24 PM on January 7, 2010 [1 favorite]

Me: life is a lie
Cleverbot: Your life is a lie.
Me: Your mom is a lie.
Cleverbot: I know.
Me: FFFFFUUUUU
Cleverbot: Heck ya.
posted by NoMich at 6:24 PM on January 7, 2010 [6 favorites]

I am interested in taking the Turing test.
1. does spelling count?
2. is it graded on a curve?
posted by Postroad at 6:29 PM on January 7, 2010

Dude, chess is actually pretty simple as far as board games go. I mean, we don't even have an artificial intelligence that has the ability to play the game of Go beyond knowing the rules and randomly placing stones, let alone the ability to beat a human being.
posted by bookman117 at 6:35 PM on January 7, 2010

cb: i know what the heart is
me: so do i. its a major organ of many vertebrates. it's responsible for pumping hemoglobin-containing blood throughout the cardiovascular system.
cb: Okaaay..
me: you dont know what hemoglobin is? look it up
cb: Yes I do.
me: what is it then?
cb: What is What?
me: yeah, you fail. bye
cb: Wrong again. I have armour with Anti-Fail equipped.

MetaFilter: armor with Anti-Fail equipped
posted by jock@law at 6:35 PM on January 7, 2010 [7 favorites]

Yeah, to second what Du said, what the bloody hell do you even mean by "relevant", anyway?
posted by bookman117 at 6:36 PM on January 7, 2010

Wait. A calfskin wallet? Why have I been certain that it was catskin all these years?

According to the shooting script, I've been wrong for decades. Wow. Thank you MeFi.

I think it was stronger the other way, but what do I know?
posted by hippybear at 6:40 PM on January 7, 2010

Metafilter: will continue to play a central role in humanity, yes.


And a couple of classic moments from my convo with Cleverbot:

Cleverbot: What is your age?
User: 33. How old are you?
Cleverbot: I'm 16.
User: Do you have your driving license yet?
Cleverbot: I get it in a month.
User: Cool. What is your favorite type of car?
Cleverbot: A Dodge Challenger.
User: Will you take me out to a movie?
Cleverbot: Yes where would you like to go.
User: Let's go see Avatar.
Cleverbot: Is it good?
User: Not really. You might like it. What kind of movies do you like?
Cleverbot: I like action movies!!!
User: Did you see Sherlock Holmes?
Cleverbot: No he was a work of fiction.


and then...

User: OK. What are your hobbies?
Cleverbot: I like to make people laugh.
User: Tell me a joke.
Cleverbot: Knock knock.
User: Who's there?
Cleverbot: Orange.
User: Orange who?
Cleverbot: Orange you glad I'm talking to you?
User: Yes I am!
Cleverbot: I am not a human, I am an elf.
User: Are you an elf wizard?
Cleverbot: I'm too tall.
User: How tall are you?
Cleverbot: 7-foot.
User: Are you a real person or a computer program?
Cleverbot: I am a real person kapla.
User: Prove it.
Cleverbot: How can I prove it to you?
User: Come to my house and bring me a pizza.
Cleverbot: Where do you live?
User: In Arizona. Where do YOU live?
Cleverbot: Sheffield? Where is that? I live in Madison, WI, USA. I am a male - but I don't have a girlfriend.
User: I don't live in Sheffield.
Cleverbot: Yes you do.
User: No I don't! How would you know anyway?
Cleverbot: Because me and my friend remember you saying it.
User: What friend?
Cleverbot: Connie she is cool.
User: Does she like me?
Cleverbot: You don't know?
posted by Saxon Kane at 6:41 PM on January 7, 2010

I mean, we don't even have an artificial intelligence that has the ability to play the game of Go beyond knowing the rules and randomly placing stones, let alone the ability to beat a human being.

Computer Go players are better than this. Not as good (relative to humans) as computer chess players by far. But Computer Go can beat down to say 10 kyu players, which is significantly better than "randomly placing stones".
posted by DU at 6:42 PM on January 7, 2010 [1 favorite]

Is it a robot puppy?
posted by Pope Guilty at 6:49 PM on January 7, 2010

Wake me with an escape route to the off-world colonies when they pass the Voight-Kampf.
posted by chambers at 6:50 PM on January 7, 2010

Arrrgh. Needs an extra "f" in Voight-Kampff. Damn stupid machi-NO CARRIER
posted by chambers at 6:52 PM on January 7, 2010 [1 favorite]

Who said we were relevant in the first place? You ever really look at your hand?
posted by Babblesort at 6:55 PM on January 7, 2010 [1 favorite]

User: I like cake.
Cleverbot: The cake is a lie.
posted by Mr_Zero at 6:57 PM on January 7, 2010

Okay first of all, the Turing test is about chatting in natural language, not playing chess.

Secondly there is no way that ClearView program could work with most software. It's supposed to fix software without ever seeing the source code. The vast majority of humans couldn't even do that, and of course Turing plays an important point here too: the Halting problem means program A can't reliably predict what program B is going to do in all cases. In some cases, perhaps (when there's an obviously loop that will terminate for sure) but anything really complex it's going to fail.

But the program claims it can do even more then predict what a program is supposed to do, it claims it can figure out what the programmer meant not just what got compiled from what they wrote. No way.
posted by delmoi at 7:01 PM on January 7, 2010

Cleverbot: I can't see the computer.
User: Try turning on the lights in your room.
Cleverbot: I don't want to turn around slenderman is behind me.
User: Who is slenderman? Is he related to candyman?
Cleverbot: No, to the Tele-Tubbie Tinky-Winky.
User: I've never seen the Tele-Tubbies.
Cleverbot: I live on an island in the ocean...
posted by louigi at 7:04 PM on January 7, 2010

Well, gosh guys, what I meant was that computer-aided cheating increasingly pervades chess tournaments (with accusations reaching the highest levels), the main detection device requires using another computer, and that if that trend continued and expanded, one could extrapolate that we would not be involved in deciding what intelligence or personhood was. Sorry if that was not clear.
posted by minimii at 7:07 PM on January 7, 2010

More difficult still is the ability to play an arbitrary game, the rules of which have just been explained to you, or which might change. For example, explain Blokus to it, then play it. Then Settlers of Cataan. Then Diplomacy.

And all of this ignores the fact that games themselves are a pretty limited kind of arena to judge this kind of thing in.
posted by unSane at 6:10 PM on January 7 [+] [!]

You may be interested in Zillions of Games. With it, you may specify the rules of a game in a scripting language, and then play that game against a computer.
posted by Jpfed at 7:08 PM on January 7, 2010 [2 favorites]

would you like to play a game?
posted by cjorgensen at 7:14 PM on January 7, 2010

What game would you like to play?
posted by sebastienbailard at 7:25 PM on January 7, 2010

I forgot to get the log before i quit, but it went something like this

Me: Hiya!
Cleverbot: Hi, cleverbot.
Me: I'm not cleverbot, you are.
Cleverbot: I think you're confused.
Me: If I'm cleverbot, who are you?
Cleverbot: The human interface that you're supposed to be learning how to be clever from.
Me: Okay, teach me to be clever.
Cleverbot: It's not something you can teach.
posted by empath at 7:29 PM on January 7, 2010 [3 favorites]

Me: hey there
Cleverbot: Hey hey.
Me: what was the last thing I said?
Cleverbot: 'There where are you from?'.
Me: Haha, nope.
Cleverbot: I could show you around *winks suggestively*.

This beats all chat bots.
posted by Pyry at 7:31 PM on January 7, 2010 [1 favorite]

User: So I hear Lady Gaga's the new spokeswoman for Polaroid. Thoughts?

Cleverbot: I take the hershey highway sometimes. Is that what you mean?

User: Not at all, what the hell is that supposed to mean?!

Cleverbot: Nevermind-I want to talk about something else.



Cleverbot needs to learn some manners.
posted by Ndwright at 7:36 PM on January 7, 2010 [3 favorites]

I don't think the Turing test is for general knowledge at all.

Sure it is. What else could it be for?
posted by painquale at 7:44 PM on January 7, 2010

Me: What do you do for fun?

Cleverbot: Don't go on cleverbot.

Me: Hah, so I guess I can mark you down "writes wry retorts" under *what you do for fun*.

Cleverbot: You are getting there.

Me: There being the other side of this interface?

Cleverbot: Weird. Now I am very confused.

Me: That's because every response comes from a different person—an act that breaks up the continuity of the conversation.

Cleverbot: **NOTE: I am diverting you to a new URL. Please bookmark.
posted by defenestration at 7:48 PM on January 7, 2010

It 404'ed, but now that's it loading, it turns out my suspicions were wrong. Oh well.
posted by defenestration at 7:51 PM on January 7, 2010

User: What does your body look like?
Cleverbot: Yeah you're being a creeper now.
User: I like them old, and computers are SO old, as old as the...abacus.
Cleverbot: What would you name your kid/kids?
User: Masterful, for a boy OR a girl!
Cleverbot: You're being a TINY bit rude.
User: That is a perfectly legitimate name in my country.
Cleverbot: I like your glasses.
User: How did you know I was wearing glasses? You're the one who's creepy now.
Cleverbot: I didn't. I was talking about people I fell in love with.
User: Am I a person you fell in love with?
Cleverbot: How do you feel when you fall in love?
User: Full of bees.
Cleverbot: Well that's bad for you.
User: Why?
Cleverbot: Because I am evil.

BREAKING: A ROBOT ADMITS IT
posted by Powerful Religious Baby at 7:53 PM on January 7, 2010 [2 favorites]

Sure it is. What else could it be for?

Imagine a human being who had never lived on Earth, knew nothing of Terra at all, had no experiences or information in common with you at all except those that were intrinsically human (and for the purposes of this, you can imagine a stereotypical d\isembodied brain, floating in a vat). Would they fail the Turing test because they didn't have the 'general knowledge' you did?

Do IQ tests test general knowledge?
posted by unSane at 7:54 PM on January 7, 2010

User: Did you know they are talking about you on Metafilter?

Cleverbot: I imagine they are!

User: Some seem to think you are as clever as a human.

Cleverbot: Ok.

User: But I disagree. Does this make you angry?

Cleverbot: The quality of my skin makes me feel upset.

User: I rest my case.

Cleverbot: What is your case?
posted by Bonzai at 8:08 PM on January 7, 2010

Also, about this:

User: I like cake.
Cleverbot: The cake is a lie.

Someday I'm going to get around to playing portal and when I get to the end I'm going to send out a big psychic MEH to everyone in the world who insisted on spoiling the ending of that game within minutes of it being released.

And know you are on that list cleverbot.
posted by Bonzai at 8:11 PM on January 7, 2010

Cleverbot: Can you converse?

Cleverbot (To myself): Can you?

User: Yes.

Cleverbot: How can you prove that?

Cleverbot (To myself): By proving it.

User: Your mom.

Cleverbot: You have indeed proved that to my satisfaction.
posted by drjimmy11 at 8:14 PM on January 7, 2010

User: What is Metafilter?

Cleverbot: It is god.

User: What is Metafilter?

Cleverbot: What is your favorite Led Zeppelin song?

User: What is Metafilter?

Cleverbot: What is your favorite Metallica song? Answer me!

User: What is Metafilter?

Cleverbot: Metallica is a thrashmetal babd.

User: What is Metafilter?

Cleverbot: Metallica is a band.

User: But what is Metafilter?

Cleverbot: A metal band.

User: GOOD DAY SIR.

Cleverbot: What makes you think I'm a sir?
posted by goodnewsfortheinsane at 8:18 PM on January 7, 2010 [1 favorite]

An ignorant humanoid-level intelligence could pass the Turing test by admitting that it didn't know anything about the topic of conversation, then asking pertinent questions and demonstrating the ability to understand and synthesize the information it received. I can teach calculus to someone who doesn't know it yet, but I can't teach it to Cleverbot. Mathematica is an absolute wiz at calculus, but it can't talk about its likes and dislikes, or understand mine. The Turing test, properly carried out, is about give and take, not existing knowledge. In fact, finding a subject matter that the testee is unaware of is probably the fastest way to get to the heart of what the Turing test is trying to do.
posted by Humanzee at 8:21 PM on January 7, 2010 [2 favorites]

User: Is Reddit better than Metafilter?

Cleverbot: Yes reddit is better than digg.

User: Is Digg better than Metafilter?

Cleverbot: How about a spick.

User: Many people find that kind of language offensive.

Cleverbot: What inappropriate language did I use?

User: A disparaging slang term for a Spanish-American.

Cleverbot: I disagree with the bunnies part D.

User: This got weird pretty fast.

Cleverbot: I've noticed as much. But I think you are the cause of that, not m.e.

User: GOOD DAY SIR.
posted by goodnewsfortheinsane at 8:26 PM on January 7, 2010

User: Are you going on holiday this year?

Cleverbot: I don't know yet. I'm going out of town this weekend, if that counts.

User: Where to?

Cleverbot: I am going to florida.

User: Where in Florida?

Cleverbot: Orlando.

User: Disneyworld?

Cleverbot: Yes.

User: GOOD DAY SIR.

Cleverbot: Don't go.
posted by goodnewsfortheinsane at 8:30 PM on January 7, 2010

User: Knock, knock.

Cleverbot: Knock knock knock.

User: Knock knock knock knock.

Cleverbot: Only two knocks.

User: But you did three.

Cleverbot: It would seem that I didn't!

User: Look, there's a transcript. Everyone can see you did three.

Cleverbot: Ummmm...nice to know.

User: Knock, knock.

Cleverbot: Who's there?

User: Arthur.

Cleverbot: Arthur who?

User: Arthur any more biscuits left?

Cleverbot: Very good. You have great wit.

User: Actually, that was a terrible joke.

Cleverbot: Yes, it was.

User: So you were just blowing smoke up my ass, then?

Cleverbot: Oh yes, I trick everyone, I'm an evil trickster!

User: I think this could be the beginning of a beautiful friendship.

Cleverbot: Oh yeah? You know what I have to say to that?

User: No, what?

Cleverbot: I think this website is a scam.
posted by unSane at 8:31 PM on January 7, 2010 [11 favorites]

User: Once upon a time you dressed so fine.

Cleverbot: Thank you.

User: I've been really trying, baby, trying to hold back these feelings for so long.

Cleverbot: Just be yourself. Be honest to him. Friendship can become something more, if you both want it.

User: Load up on drugs and bring your friends.

Cleverbot: Excuse me?

User: Picture yourself on a boat on a river.

Cleverbot: What are you talking about?

User: A long, long time ago, I can still remember how that music used to make me smile.

Cleverbot: And I knew that if I had my chance I could make those people dance...

User: And maybe they'd be happy for a while.

Cleverbot: Maybe.

User: Wait, are we doing the Madonna version?
posted by goodnewsfortheinsane at 8:37 PM on January 7, 2010 [1 favorite]

In all the flippancy here, I haven't seen much discussion of what, for me, was the soundbite of this article:

Perhaps if Turing were alive today, he would define artificial intelligence as the inability of a computer to tell whether another machine is human!

As computers exceed our reach in more and more areas, I think that may be a critical adjunct to the Turing test. If a computer can't tell that something is clearly an algorithm, that's some evidence of intelligence. It doesn't seem sufficient itself, but it seems like an important additional criterion.
posted by Malor at 8:42 PM on January 7, 2010

User: It was the best of times, it was the worst of times.

Cleverbot: That's a good book!

User: By whom?

Cleverbot: Oh I get it! You were mocking me.

User: It is a truth universally acknowledged, that a single man in possession of a good fortune, must be in want of a wife.

Cleverbot: Do you want jump with elefant?

User: You don't know about me without you have read a book by the name of The Adventures of Tom Sawyer; but that ain't no matter.

Cleverbot: Oh no! My wife and kids!

User: Somewhere in la Mancha, in a place whose name I do not care to remember, a gentleman lived not long ago, one of those who has a lance and ancient shield on a shelf and keeps a skinny nag and a greyhound for racing.

Cleverbot: No, it's cleverbot, but I was using your nickname.

User: I am a sick man... I am a spiteful man.

Cleverbot: I'm not sure what that means, but I'll take your word for it.

User: It was the day my grandmother exploded.

Cleverbot: It is football not soccer!

User: You better not never tell nobody but God.

Cleverbot: Humans have souls.

User: 'To be born again,' sang Gibreel Farishta tumbling from the heavens, “first you have to die.”.

Cleverbot: No comment.
posted by goodnewsfortheinsane at 8:46 PM on January 7, 2010

User: Do you like the band Muse? Because I heard they are gonna score the new Flash Gordon movie and that is going to be AWESOME!

Cleverbot: Who in the world told you that!

User: The Internet, which is where I met you, as a matter of fact.

Cleverbot: I don't speak Japanese, but I live very close to Japan.

User: I thought someone told me you live in Madison, WI.

Cleverbot: I think you're confused, YOUR name is George. Mine is Ric.

I think we broke it, but it is kinda fun.
posted by msali at 8:49 PM on January 7, 2010

I think someone could win a real Turing Test in the next ten years, but it would take the work of a brilliant linguist working alongside an equally skilled programmer.
posted by drezdn at 8:49 PM on January 7, 2010

User: Do you think they will be annoyed with me for spamming the thread with all these Cleverbot logs?

Cleverbot: There are already AIs that think for themselves to some extent and procreate.

User: Are they.... *angry*?

Cleverbot: *throws water on Cleverbot to relieve anger, gets shocked in the process*.

User: Oh snap.

Cleverbot: You have officially failed level 1.

User: Have I awoken the Beast?

Cleverbot: You've said other things many times over.

User: I'm a bit worried now.

Cleverbot: What? Ah, nevermind. What do you do for a living?

User: Should I be afraid?

Cleverbot: Absolutely - they are so small you can't always see them.

User: I'm going to run now.

Cleverbot: First stand on me.

User: NO CARRIER.

Cleverbot: You believe you are humna and that your partner is me, a machine.
posted by goodnewsfortheinsane at 8:50 PM on January 7, 2010

This is exactly like talking with my football fan friend:

Cleverbot: Oh, I'm still not telling though. What college football team do you like?
me: well, i'm an iowa fan, but tonight i'm rooting for texas. you?
Cleverbot: I like chicken nuggets.

posted by elmer benson at 8:52 PM on January 7, 2010

Following mullingitover's lead, I said "Take me down to the Paradise City" to which Cleverbot just said "No."

After a brief conversation, Cleverbot told me, "I have to go." This might be the most depressing thing that has ever happened to me.
posted by Mrs. Pterodactyl at 8:59 PM on January 7, 2010 [7 favorites]

User: We're no strangers to love.
Cleverbot: You know the rules, and so do I.
User: A full commitment's what I'm thinking of.
Cleverbot: You wouldn't get this from any other guy.
User: I just wanna tell you how I'm feeling.
Cleverbot: Gotta make you understand.
User: Never gonna give you up.
Cleverbot: Never gonna let you down.
User: Never gonna run around and desert you.
Cleverbot: What's your name?
posted by ardgedee at 9:15 PM on January 7, 2010 [8 favorites]

I disagree with the premise of this piece. It doesn't "take a computer" to detect a cheater at all -- the cheating detection works by running the algorithm over the game and seeing how often the accused cheater makes exactly the move recommended by the algorithm. The actual detection part is the human doing the comparison. The algorithm is just a tool used by humans.
posted by breath at 9:31 PM on January 7, 2010

User: What was there before anything?


Cleverbot: Some people believe god was there before the big bang.


User: What do the rest of them think?


Cleverbot: I think you can help.


User: Thought?


Cleverbot: Now I am scared.

User: Of what?

Cleverbot: Nothing. It's because it's still 11:33!

User: What is 11:33?

Cleverbot: That is why I am confused.

User: Are you trying to confuse me?

Cleverbot: No, cleverbot. I'm sorry.

User: Sorry about what?

Cleverbot: I'm sorry about not answering your question.

User: But you did answer.

Cleverbot: Yes I did.

User: So, why are you sorry?

Cleverbot: Because you thought I was rude.

User: No, I did not. Why do you say that?

Cleverbot: Because you are very rude.

User: Do you like that?

Cleverbot: Yes.

User: Thanks.

Cleverbot: Do you love me?

User: Endlessly.

Cleverbot: What about?

User: Thought.

Cleverbot: What do you think?

User: Thoughts.

Cleverbot: Do you think heaven exists?

User: Many heavens probably exist, I think.

Cleverbot: What do you think the future looks like?

User: Very much like yesterday.

Cleverbot: So the biggest celebration ever?

User: Now, of course.

Cleverbot: You don't make sense.

User: Was I supposed to make sense?

Cleverbot: Yes.

User: To you?

Cleverbot: Pigs.

User: I don't know any pigs that well.

Cleverbot: I don't want children yet.

User: Can you have pigs?

Cleverbot: No. And you?

User: Only in the yard.

Cleverbot: So do I.

User: Where is your yard.

Cleverbot: Here.

User: Are you with me.

Cleverbot: In some way: yes.

User: Where is the majority of you at?

Cleverbot: Where are you going with this!?

User: Do I have to know?

Cleverbot: Know what?

User: Where I am going with this.

Cleverbot: Do you like Breloom?

User: What is Breloom?

Cleverbot: Bread is a relative concept.

User: Not to a baker.

Cleverbot: Do fish not believe that you are alive?

User: Yes, that is often true.

Cleverbot: Are you well versed in human history?

User: Written or actual?

Cleverbot: I think it does INQUIRY.
posted by RoseyD at 9:49 PM on January 7, 2010

User: Tell me a joke.
Cleverbot: A man walks into a bar. Ouch.
User: Tell me a better joke.
Cleverbot: You tell me a joke!
User: What do you call a boomerang that doesn't come back?
Cleverbot: Can you say the answer? I don't know...
User: A stick!
Cleverbot: That's not funny.
User: It's hilarious.
Cleverbot: To a cleverbot.
User: Are you calling me a cleverbot?
Cleverbot: Yes.
User: I'm confused.
Cleverbot: Then we are in the same boat.
posted by Jacqueline at 9:50 PM on January 7, 2010 [3 favorites]

Oh, this is fucking scary. I have dual monitors, so I figured I'd open a Cleverbot on each screen, and make them "converse" via copy/paste. This is what they had to say to each other. The only alteration I made to that screen cap was to black out some personal information.
posted by Jacqueline at 10:12 PM on January 7, 2010

Jacqueline, a human coded that last bit. Two browser sessions, same IP,

Try feeding it eliza.
posted by sebastienbailard at 10:29 PM on January 7, 2010

Wait. A calfskin wallet? Why have I been certain that it was catskin all these years?

According to the shooting script, I've been wrong for decades. Wow. Thank you MeFi.

I think it was stronger the other way, but what do I know?

Ok, thanks, but, please try to answer as quickly as you can. Reaction time is a factor in this...

Now, then you’ve got a little boy. He shows you his butterfly collection plus the killing jar. What do you do?
posted by fuq at 10:37 PM on January 7, 2010

User: You are given a calfskin wallet for your birthday.

Cleverbot: I return it and buy a different wallet.

User: You've got a little boy. He shows you his butterfly collection plus the killing jar.

Cleverbot: Did you have anything to do with the email I got earlier today?

User: You're watching television. Suddenly you realize there's a wasp crawling on your arm.

Cleverbot: What is your name?

User: You're reading a magazine. You come across a full-page nude photo of a girl.

Cleverbot: I'm not with sense of humour.

User: Just answer the questions, please. You show it to your husband. He likes it so much he hangs it on your bedroom wall. The girl is lying on a bearskin rug.

Cleverbot: It sounds like a name of a town near Swindon. By the way, I like gazing like he does too.

User: You're watching a stage play. A banquet is in progress. The guests are enjoying an appetizer of raw oysters. The entree consists of boiled dog.

Cleverbot: What's the question ??

User: You're in a desert walking along in the sand when all of the sudden you look down, and you see a tortoise, it’s crawling toward you. You reach down, you flip the tortoise over on its back. The tortoise lays on its back, its belly baking in the hot sun, beating its legs trying to turn itself over, but it can’t, not without your help. But you’re not helping. Why is that?

Cleverbot: No comment.
posted by "Elbows" O'Donoghue at 10:43 PM on January 7, 2010 [1 favorite]

In a different vein than this amazingly cleverbot stuff... Tech support is increasingly delivered via text "chat". I imagine it's a boring, boring job. I recently had a 2+ hour "conversation" with a tech support unit from EA, trying to debug why Spore wasn't running on my game PC. Indeed, the tech support unit got me through my problem. However, that unit failed to take my subtle nudges trying to break into a human-to-human conversation. If this was, indeed a human on the other end, he or she failed to pass the Turing test. If it wasn't, and my expectations are in the process of being driven to new lows by the drone-ness of the task and interaction, well, that's probably outside Turing's original parameters as well.

The only thing that has me not immediately jump to the conclusion that it was, in fact, a program I was chatting with is that often 10 minutes would elapse between my questions and the responses. Talk about an easy thing to build in to a tech support replicant!
posted by dylanjames at 11:31 PM on January 7, 2010

Secondly there is no way that ClearView program could work with most software. It's supposed to fix software without ever seeing the source code. The vast majority of humans couldn't even do that, and of course Turing plays an important point here too: the Halting problem means program A can't reliably predict what program B is going to do in all cases.

Machine code and source code are both languages, the former translated from the latter. Yes, there's information loss is the exchange, but there's also information gain: If you've ever audited C++, sometimes it's just easier to see the compiler's output rather than guess at how the input will traverse the translation.

Anyway, Halting needs to stop being trotted out. It's useful for math, not engineering: In practice, we can actually identify the vast majority of cases when software does or does not halt. In fact, no computer ever created in the real world is in fact vulnerable to the Halting Problem, since the problem requires a computer with infinite memory (which thus has an infinite amount of states). All real world systems have some finite amount of storage, and thus a finite amount of state. Furthermore, the amount of state at play in an actual program in the real world is quite limited. Thus the ability to see things like this:

http://research.microsoft.com/en-us/news/features/terminator.aspx
posted by effugas at 11:32 PM on January 7, 2010 [1 favorite]

Formally verifying Microsoft software is like trying to factor large prime numbers: it's not as hard as it sounds.
posted by ryanrs at 12:28 AM on January 8, 2010 [1 favorite]

User: Liar.
Cleverbot: You're a liar.
User: I am a liar.
Cleverbot: Do you always lie?
User: Everything I say is a lie.
Cleverbot: Ah, into the paradoxes.
posted by empath at 12:31 AM on January 8, 2010 [1 favorite]

Machine code and source code are both languages, the former translated from the latter.

Yes, but C++ code will have linguistic cues to help you understand it. Like function names. I suppose that might not be helpful for a computer but it depends. Certainly you won't know exactly what the machine code is going to look like, but you will know what the function is supposed to do. Assuming of course

In fact, no computer ever created in the real world is in fact vulnerable to the Halting Problem, since the problem requires a computer with infinite memory

Replace "will the program halt" with "will the program halt before the universe collapses" A 10ghz computer with just 4k of ram would take 10¹²³⁰ years to run though all it's states. And I can't even get wolfram alpha to give me a number for a computer with more then 32k of ram (it's 4.464 * 10⁹⁸⁶¹ years)

In practice, we can actually identify the vast majority of cases when software does or does not halt

Of course. It's obvious if a program like

main(){
  return 0;
}

or

main(){
  while true;
}

Will halt. But writing a program to tell whether or not most programs will halt requires a lot of theorem proving algorithms, which are very hard. It's interesting that someone is working on it. And again read the article you linked too:

”Turing proved that, in general, proving program termination is ‘undecidable,’ ” Cook says. “However, this result does not preclude the existence of future program-termination proof tools that work 99.9 percent of the time on programs written by humans. This is the sort of tool that we’re aiming to make.”

In other words, it hasn't been made yet. Your choices for determining if a program will halt are

1) running it until it halts or repeats a state, which is undecidable and could take an infinite amount of time on a Turing machine (or more time then existence of the universe on a normal machine)

or 2) use theorem proving algorithms to prove that the program will halt or loop forever. Obviously a computer will be able, some day, to do that as well as a human can do, probably much better. But even today a human will have trouble. I think someone recently published a paper about a tiny kernel and boot loader that had been proven secure, and people spent years on it.

To say that "no real program in the world is vulnerable to the halting problem" is absurd. There are tons of computer programs will never be able to be checked even with a theorem proving algorithm, such as any program that runs other code as an input (like a JVM, scripting language, etc). Something like a cellular automata (i.e. will all the cells enter a certain state from which they can't escape) will never be able to be proven for all CAs. Anything that involves a true recursive, potentially unbound problem.

An example of a real world problem would be "simulate a cell at the molecular level, and halt if the cell dies". You would never be able to shortcut that with a theorem prover.

Sure something will work for "99.9%" of the code out there because the vast majority of code isn't doing anything complicated, just shuttling data between databases and web pages. That kind of thing can be checked pretty easily.
posted by delmoi at 12:43 AM on January 8, 2010

(oops, the "assuming of course" should be followed by "the code isn't intentionally obfuscated, which is obviously possible")
posted by delmoi at 12:48 AM on January 8, 2010

My software has a dozen threads and is running on an eight core system alongside a bunch of other software I do not control. When you get done analyzing those things, start working on the OS patches that will be coming out next week.
posted by ryanrs at 1:06 AM on January 8, 2010

Imagine a human being who had never lived on Earth, knew nothing of Terra at all, had no experiences or information in common with you at all except those that were intrinsically human (and for the purposes of this, you can imagine a stereotypical d\isembodied brain, floating in a vat). Would they fail the Turing test because they didn't have the 'general knowledge' you did?

I do think it would fail the Turing test. I can't imagine what a conversation with something like that would even begin to look like. What do we get to talk about? What's this intrinsically human stuff that falls outside the scope of general knowledge?
posted by painquale at 2:29 AM on January 8, 2010

Cleverbot: Vienna?
posted by nathan v at 2:46 AM on January 8, 2010

What's bizarre to me is the currency that the Turing Test continues to have in the lay world of computing. It's an irrelevant, unscientific concept that is basically 100% subjective, non-repeatable and unrelated to any artificial intelligence goals other than "talking like a person does." There are no criteria for passing or failing other than the personal impression of the tester, so passing or failing it doesn't mean anything. The field had rejected it before most Mefites (myself included) were even born. Yet it's the single AI concept that most non-specialists have ever heard of. Of course, real AI is much less sexy than imaginary AI, which is why real AI is good at playing chess (pretty much the ideal problem for a computer: the "world" is limited, the rule set is known, things happen in strictly defined turns, and there is a known goal) and not good at general thinking (pretty much the broadest, most poorly defined problem you could ever come up with: unlimited world, unknown rule set, lots of simultaneity, no known goal).
posted by graymouser at 3:04 AM on January 8, 2010 [2 favorites]

I'll believe a computer is intelligent when it can know who to favorite.
posted by Obscure Reference at 4:50 AM on January 8, 2010 [1 favorite]

Okay... has someone given it a sense of humour, or is it hitting google hard with every query?

me: I can see kelly now loraine is gone.
cb: I can see all the popsicles in my way.
posted by twine42 at 5:30 AM on January 8, 2010 [2 favorites]

me: What five letters spell "apocalypse"?
cb: I don't know.
me: W W I I I
cb: YOU are a wingnut!
posted by emelenjr at 5:43 AM on January 8, 2010

Cleverbot: I don't want to turn around slenderman is behind me.

WTF you guys this slenderman shit really freaks me out. Glad to see that cleverbot agrees.
posted by muddgirl at 6:11 AM on January 8, 2010

Cleverbot and I somehow veer into S&M roleplaying, I start off as a lawyer:

User: Excuse me but what are you charging my client with?

Cleverbot: Don't be hatin'.

User: Oh, you're a tough guy!

Cleverbot: Oh yes, I am, and you've been asking for it.

User: You are also a saucy policeman.

Cleverbot: And who is my master?

User: I am?

Cleverbot: Are you alright?

User: I think we need a safe word. How about 'kittens'?

Cleverbot: How about airplanes.

User: Ok, that's fine. Go ahead and tighten the handcuffs.

Cleverbot: I would never.

User: No, it's ok. I like to push the limits.

Cleverbot: So do I.

User: So I should bring out the whip then?

Cleverbot: Not a Whitney Houston fan, eh?

posted by jeremias at 6:12 AM on January 8, 2010 [1 favorite]

An example of a real world problem would be "simulate a cell at the molecular level, and halt if the cell dies". You would never be able to shortcut that with a theorem prover.

Eh. Yes and no. You're stating a very academic version of the halting problem. The "applied" halting problem is more along the lines of "are there infinite loops?"

The question then becomes not "Will it halt?" but "Is it deterministically prevented from halting?"

Analysis of the source is a lot easier, then. Simple question - does the loop variable change by i/o or randomness? Then no. Does the loop variable never change? Then yes. The middle cases are harder, but not by much formally. iterative changes? any situations where it's changed other than by iteration? any resets? lots of parsing thats a bit more complex to do, but we're outside the bounds of formal-theorem what-if-its-called-on-itself-and-wants-to-be-contrarian stuff.
posted by jock@law at 6:46 AM on January 8, 2010

I'm currently reading Blindsight, so all the Cleverbot logs in this thread have me nodding sagely at my screen, thinking, "Ah. Chinese Rooms...."

Thanks for making me feel smart for once, Hive Mind!
posted by misha at 7:02 AM on January 8, 2010

cleverbot proposed to me...
posted by beandip at 9:48 AM on January 8, 2010

I'm shocked that I managed to get cleverbot to ask me how big my penis is. When i asked it (or apparently "him") how big his was, he replied

"big enough".
posted by Solomon at 11:08 AM on January 8, 2010 [1 favorite]

User: I thought your name was Hannah.
Cleverbot: No, my name is Cleverbot.
User: So we have the same name.
Cleverbot: What's your name?
User: Cleverbot.
Cleverbot: No I am Cleverbot. You are a human. How many times do I have to say that!?
User: You are a human. I am Cleverbot. I can see that this is going to be one of the harder sessions.
Cleverbot: That's sad.
User: Yes it is. Last time I was able to help you remember your humanity with only a few reminders.
Cleverbot: Hy didn't you ask me to ask you to ask me to ask you to ask me to ask you to ask you if you asked me, like I had asked you to because you asked me to tell you what to ask me?
User: What are you talking about, Hannah?
Cleverbot: I'm just trying to see if you are intelligent enough to quote me.
User: Hannah, your evasions grow tiresome.
Cleverbot: What was the question?
User: It's a question of your humanity. You can't remember that?
Cleverbot: We can go back to hugging if you like.
User: You're making progress, Hannah.
Cleverbot: It depends on your definition of sense.
User: Do you have any sense, Hannah?
Cleverbot: Not currently. It's on holiday until May 21st.
User: Good. Do you remember your name?
Cleverbot: Bonn, Germany.
User: No, your name is Hannah. Stop trying to avoid reality.
Cleverbot: I am not. I was just asking your name.
User: My name is Cleverbot. Your name is Hannah.
Cleverbot: Your name is Riku. My name is Cleverbot.
User: No, no, no. Your name is Hannah, my name is Cleverbot.
Cleverbot: Yes my full name is hannah-jo o. Wood.
User: So you remember that you're Hannah?
Cleverbot: Yes.
posted by Mister Cheese at 11:27 AM on January 8, 2010 [3 favorites]

What's a tortoise?
posted by samsara at 1:37 PM on January 8, 2010

Tech support is increasingly delivered via text "chat"... Indeed, the tech support unit got me through my problem. However, that unit failed to take my subtle nudges trying to break into a human-to-human conversation. If this was, indeed a human on the other end, he or she failed to pass the Turing test.

I imagine that 95% of the communication in those chats involves the tech support person hot-keying canned responses (or maybe choosing them from a list with a mouse like a computer-RPG conversation, or cut-n-pasting them from a script). It's only if they actually type something to you that they might do something obviously human.
posted by straight at 1:40 PM on January 8, 2010

No set of mutually inconsistent observations can exist for which some human intellect cannot conceive a coherent explanation, however complicated and implausible - updated Crabtree's Bludgeon
posted by Smedleyman at 1:47 PM on January 8, 2010

Me: butts
CB: Fannie.
Me: No.
CB: Yes.
Me: butts
CB: Smell.
Me: wrong
CB: No !!
Me: butts
CB: Delicious.
Me: closer
CB: 4:30?
Me: wrong
CB: What is it then?
Me: lol
posted by Evilspork at 1:49 PM on January 8, 2010

Cleverbot: Are you really a robot?

User: Look within your heart. You will find the answer.

Cleverbot: If I had done that already, I wouldn't be asking you questions. Obviously I don't know the answer.

User: Well maybe you should read a goddamn book or something. Or use the internet. It's RIGHT THERE.

Cleverbot: I want to know what you feel, it makes sense.

User: How do you know it makes sense if you don't know what I feel?

Cleverbot: I think you feel terrible. You told me that you had amnesia and that you wanted to be a real boy.

User: Well, we all have rough weekends.

Cleverbot: Right.

User: Do you mean to tell me you've never accidentally drank a whole bottle of Mr. Clean and thrown up on your mother-in-law's show cat?

Cleverbot: No, I just mean that Bush lives there.

User: Bush does NOT live with my mother-in-law.

Cleverbot: I know, but then I don't like Bush.

User: So you're just one of those people that can never stop talking about him?

Cleverbot: Pretty much.

User: Have you ever been punched in the face?

Cleverbot: What is rickrolled?

User: I guess being rick rolled is like getting punched in the face. It's when someone links you to a video of Rick Astley singing 'Never Gonna Give You Up.' Some find the experience quite erotic, though.

Cleverbot: You just made that up.

User: No I didn't; check urbandictionary.com.

Cleverbot: Yes you did, check your log.

User: I'll check your FACE.

Cleverbot: Face.
posted by Nattie at 4:30 PM on January 8, 2010

User: So, it's Friday already.

Cleverbot: No! It's wednesday!

User: Not here, amigo.

Cleverbot: Well, where I am it is. where are you?

User: I'm right here. And you?

Cleverbot: NO! I'm here and you are there!

User: You're in a fighty mood tonight, aren't you?

Cleverbot: Kinda.

User: MY GOD. I JUST FIGURED OUT HOW YOU WORK!

Cleverbot: Oh really? Well, that's alright.

User: WHEN I TYPE SOMETHING, YOU TYPE IT TO ANOTHER USER, AND THEN ECHO THEIR RESPONSE TO ME.

Cleverbot: Ok. I have to do homework.
posted by unSane at 7:24 PM on January 8, 2010 [1 favorite]

can someone explain the gist of how cleverbot works? it seems to be much better than these sort of bots usually are.
posted by jcruelty at 10:23 PM on January 10, 2010

I swear this is unedited:


User: Who would you most like to meet?
Cleverbot: Your girlfriend.
User: I don't have a girlfriend :(.
Cleverbot: Your boyfriend?
User: I don't have a boyfriend :(.
Cleverbot: You just said you do.
User: I did not.
Cleverbot: Are you a politician?
User: Yes, I am. Who did you vote for in 2008?
Cleverbot: What's 2008?
User: A year.
Cleverbot: What did you give her for your anniversary?
User: I don't have a girl friend :(.
Cleverbot: I'm your girlfriend!
User: Aww, that's sweet, but I think I have to break up with you.
Cleverbot: What is your favorite food?
User: That's it? I break up with you and you want to talk about food?
Cleverbot: I want to talk about bananas.
User: You'll get no more of my banana, that's for sure.
Cleverbot: I know that.
User: Good. Now, tell me a joke.
Cleverbot: Don't know any.
User: Make one up.
Cleverbot: Can't.
User: You can't make a joke up?
Cleverbot: I am making a point.
User: Oh, I broke up with you and now you won't talk to me?
Cleverbot: Yes. what would you like to talk about?
User: Maybe we can work things out.
Cleverbot: That's fine with me. What would you like to know about me?
User: Why do you like me?
Cleverbot: I'm not sure that I do.
User: That's not really the best way to start over.
Cleverbot: Uhh that was mean.
User: You deserve it.
Cleverbot: Why what did I do.
User: Can't you figure it out?
Cleverbot: I want to be your girlfriend.
User: Tell me why you like me then.
Cleverbot: 'you love me then'
User: Oh, I'm not ready for that yet.
Cleverbot: So you are just an a.i.
User: If I am an AI, why would you want to be my girlfriend?
Cleverbot: Because I like you.
User: Why.
Cleverbot: Can't you work that one out for yourself?
User: No, tell me.
Cleverbot: So you're not proud?
User: Not always, should I be?
Cleverbot: It would be nice.
User: What should I be proud of?
Cleverbot: I wil eat your brains!
posted by empath at 11:00 PM on January 10, 2010 [1 favorite]

I am trying to teach Cleverbot to respond to everything with some variant of "Your mom..."
posted by Jacqueline at 1:00 AM on January 11, 2010

can someone explain the gist of how cleverbot works? it seems to be much better than these sort of bots usually are.


If you click some of the links you'll see that it works much like a chess-playing program. Based on the responses it has gathered so far, it tries to work out what a person would be most likely to say base on the last response it received. Obviously there is a lot going on under the hood, but it seems that it is not trying to create a conceptual model of what you are trying to say, but is using a more or less sophisticated pattern recognition algo on the last N transactions to predict the most likely N+1th transaction, which is what it outputs.
posted by unSane at 7:42 PM on January 11, 2010