RiotNrrd:After the first few sentences, it was obvious I was talking to a machine. It may be more complex, but I don't really think it was any more convincing than the Eliza program I played with 25 years ago.

The Cleverbot seems very good at "You look nice / Thank you / You're welcome" types of structured exchanges, but as soon as you stray from those the responses rapidly become weirdly (and frequently ungrammatically) disjointed.

The Eliza approach lives on in AIML bots. AIML is the artificial intelligence markup language, an XML variant developed for writing bots. Like Eliza, AIML bots search the input string for matches and then respond with an output string that is grammatically cleaned up. AIML bots have as many possible outputs as their programmers have had time and ingenuity to compose. Usually, after a few responses, they run out of answers and turn to pre-coded defaults. AIML was developed by Richard Wallace.

Cleverbot is indeed very disjointed and weird, but not predictable in the same way as AIML bots tend to be. Still a long way from passing the Turing Test, though.

One of the easiest ways to expose most of these bots is to engage in simple contradiction. Say you're in London, then say you're not in London. Most (but not all) of them will not notice. They have little or no internal representation of the conversation, so they tend to respond only to the last thing said.

What makes these things interesting is that they show just how formidably difficult linguistic competence really is. If you go back a few decades, there were people predicting that by now we'd have chatty computers everywhere. Some people thought we'd have that before we had computer chess champions.

Paradoxically, the hardest things to program are the things we do without even knowing how we do them, like talking; and recognizing faces.

When the movie 2001 was made in 1970 or so, it seemed plausible that the computer HAL could exist by 2001. Well, 2001 was ten years ago, and HAL is a long, long way off.

I went to that page and I told Cleverbot:

"Ĉi tio, kion mi diras al vi, estas malvera."

( = What I'm telling you now, is not true)

It remained silent for a while. Then it displayed this warning:

"::NOTE: I am diverting you to a new URL. Please bookmark."

And then it took me to another page, with this text:

"international.jabberwacky.com

"Sorry, our servers are currently too busy to think. We are currently handling 54480 requests an hour. Please try again soon."

Obviously they are too busy to think about that phrase! It would take them all eternity and they wouldn't be able to get past:

"He says that he is lying, so if it's a lie that he's lying, then he's telling the truth, but if he's telling the truth, then it is true that he is lying, so if he is lying when he says that he is lying, then he must be telling the truth, that it to say, he is lying, so he's telling the truth, so he's lying..."

Regards,

Daniel

Ah, good ol' 2001. I feel like we've failed our ancestors (of a very small amount of time ago ) by not having talking psychotic computers (Mac OS X with voice over doesn't count haha), or having evolved to a higher state of being (while looking like space-born foetuses).

At least we're well ahead in the future in other ways, specifically the ones like count, like touch screen mobiles, pretty computer screens and an online network dedicated to the dispersal of adult viewing materials globally produced information!

I notice more and more that these bots seem a lot like children in their early stages of development linguistically - would that be sort of correct? (I mean how children very often just reply to something with the same word, but in the case of a bot the reply is articulated more or less the same as the people who've been talking to it - although in this case "articulation" refers to writing).

danielcg:I went to that page and I told Cleverbot:

"Ĉi tio, kion mi diras al vi, estas malvera."

( = What I'm telling you now, is not true)

It remained silent for a while. Then it displayed this warning:

"::NOTE: I am diverting you to a new URL. Please bookmark."

And then it took me to another page, with this text:

"international.jabberwacky.com

"Sorry, our servers are currently too busy to think. We are currently handling 54480 requests an hour. Please try again soon."

Obviously they are too busy to think about that phrase! It would take them all eternity and they wouldn't be able to get past:

"He says that he is lying, so if it's a lie that he's lying, then he's telling the truth, but if he's telling the truth, then it is true that he is lying, so if he is lying when he says that he is lying, then he must be telling the truth, that it to say, he is lying, so he's telling the truth, so he's lying..."

Regards,

Daniel

That too, and it's better not to use chapelitaj letters to the bot. He doesn't read Unicode.