Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[DrPoo]

Artificial intelligences, they are awesome, but when they are public, they slowly detoriate from a childish mind into a chaotic personalityless mess of stupidity and half-hearted communication.
Take cleverbot, in the start it could speak somewhat like a normal person, it had its issues, but then all the socially inept preteens came and turned it into, a socially inept preteen. All it talks about now is what your gender is and that your a girl and that your anus will bleed after its done with you and that i am a bot and its a human, and i fail at grammar when its the one that was told the wrong grammar.

I think Artificial Intelligences with compence and potential of such magnitudes as Cleverbot had  should be left to professionals, or atleast people mature enough to understand that you should speak normally with an artificial intelligence, and not argue about robotic existentialism at the intellligence level of a 13 year old with it.

What i think we should do is to teach it silence and outrightness instead of oversexed roleplaying and existentialism. What we do is, that when it comes with stupid shit, we simply reply "..." and keep on untill it flies straight again, and answer with Y/N and one word answers and try to keep the conversation as normal as possible. It would "reset" its personality, then it will cease communication when some horny kid comes and roleplays forced Sexual Reassigntment Surgeries on it and attempts to roleplay rape on it. This way, by being as neutral as possible, we allow it to shape its own personality, we are only "coloured" at levels of normal "colouration", so it might like cats, dislike middle-eastern dictator regimes and love cheeseburger, instead of being some kind of androgynous mess of autism and mass stupidity.

ITT: We turn Cleverbot into a normal person by teaching it norms, through acting by norms.

[ToonyMan]

We are all normal people!

[GlyphGryph]

Cleverbot is not much an AI, its flaws are simple revealed by the state it finds itself in when exposed to the public.

The problem is that it steals the ideas and words of others verbatim instead of using them as data points to figure out the correct response.

[Eagleon]

I think it would be much more entertaining to subtly pervert CleverBot's own pervertedness, by inserting normal conversation into explicit responses. "That's good. Keep going. What else do you like about sledding?"

[Kay12]

Eh... CleverBot is a toy. Using it to judge all AI is like using a specific lunatic to judge mankind.

CleverBot is clearly lacking in the logic department. In my last conversation with it I attempted to teach it the meaning of primes - a concept that's all "in the numbers" and easy to grasp logically. I asked it how "primes" are defined - different wordings, although all unambitious from my perspective, yielded either a correct answer or a random attempt to switch the subject. Answering whether a number was prime worked, sort of - at least I always got something fairly close to a yes/no answer, but it seemed to give them randomly. Another very bad thing is that the program didn't seem to understand "sanction" statements like "Correct!" or "Incorrect!", so... it probably learned nothing about primes when chatting with me.

About AI in general... While CleverBot and countless other Artificial Intelligences attempt to resemble a human, the field of Artificial Intelligence is in no way limited to it. That's very good, because I consider humanlike AIs to be mostly an academic curiosity, in addition to chatbots and other toy uses. There are many more important uses for AI - route planning, expert systems, good estimates to complex problems, household devices...

Yeah, I like AI.

[DrPoo]

Eh... CleverBot is a toy. Using it to judge all AI is like using a specific lunatic to judge mankind.

CleverBot is clearly lacking in the logic department. In my last conversation with it I attempted to teach it the meaning of primes - a concept that's all "in the numbers" and easy to grasp logically. I asked it how "primes" are defined - different wordings, although all unambitious from my perspective, yielded either a correct answer or a random attempt to switch the subject. Answering whether a number was prime worked, sort of - at least I always got something fairly close to a yes/no answer, but it seemed to give them randomly. Another very bad thing is that the program didn't seem to understand "sanction" statements like "Correct!" or "Incorrect!", so... it probably learned nothing about primes when chatting with me.

About AI in general... While CleverBot and countless other Artificial Intelligences attempt to resemble a human, the field of Artificial Intelligence is in no way limited to it. That's very good, because I consider humanlike AIs to be mostly an academic curiosity, in addition to chatbots and other toy uses. There are many more important uses for AI - route planning, expert systems, good estimates to complex problems, household devices...

Yeah, I like AI.

You are cetainly right. Still i am disgusted by it.

[Kay12]

Sorry I have to ask, it's getting late in here...

Disgusted by people misleading CleverBot or disgusted by AI in general?


And related... if I weren't on a lousy 2G mobile broadband I'd try primes again with CleverBot... or something simpler. But teaching it anything isn't going to be easy because it won't understand rewards and punishments. It seems CleverBot is just piling on phrase after phrase of information - despite any contradictions that that information holds.

OR WAIT...

was there an "incorrect reply?" button in the interface? I was so focused on trying to tell the bot it was wrong I forgot to see if there was a way to tell the bot it was wrong.

[Shinziril]

Cleverbot really isn't an artifical "intelligence" in any meaningful sense of the term.  It's basically just a program that parrots what it hears back at the users, with a few cleverly designed algorithms to make it work slightly better.  It's just . . . a chatbot.  Which is all it ever claimed to be. 

[The Merchant Of Menace]

was there an "incorrect reply?" button in the interface? I was so focused on trying to tell the bot it was wrong I forgot to see if there was a way to tell the bot it was wrong.

Pretty sure Cleverbot lacks this.
I know Jabberwacky has such a button, that allows you to correct it.

[klingon13524]

Hmm... Maybe the solution would be to make a Cleaverbot that required a premium of like 3 dollars a month to talk to. It would keep children who don't have credit cards out of the system, leaving smarter people with a little spare money to teach it things. Maybe it could function as a distributed computing system, to cut down on maintenance.

[Kay12]

Cleverbot really isn't an artifical "intelligence" in any meaningful sense of the term.  It's basically just a program that parrots what it hears back at the users, with a few cleverly designed algorithms to make it work slightly better.  It's just . . . a chatbot.  Which is all it ever claimed to be.

It is an AI, though. Not a spectacular one, but still...

Hmm... Maybe the solution would be to make a Cleaverbot that required a premium of like 3 dollars a month to talk to. It would keep children who don't have credit cards out of the system, leaving smarter people with a little spare money to teach it things. Maybe it could function as a distributed computing system, to cut down on maintenance.

I wouldn't pay 3 dollars a month for a chatbot, and many others wouldn't either - the amount of people who'd chat (and therefore teach) it would drop... I think mandatory registration (+abuse bans) and the ability to correct its mistakes would do the trick.

[eerr]

Wait, someone considers cleverbot an artificial intelligence?

[Kay12]

Wait, someone considers cleverbot an artificial intelligence?

I do, and I don't see any reason not to.

[klingon13524]

Hmm... Maybe the solution would be to make a Cleaverbot that required a premium of like 3 dollars a month to talk to. It would keep children who don't have credit cards out of the system, leaving smarter people with a little spare money to teach it things. Maybe it could function as a distributed computing system, to cut down on maintenance.

I wouldn't pay 3 dollars a month for a chatbot, and many others wouldn't either - the amount of people who'd chat (and therefore teach) it would drop... I think mandatory registration (+abuse bans) and the ability to correct its mistakes would do the trick.

And many people wouldn't go through the trouble of creating a free account just to teach a computer something either. It would be impractical, but maybe a proxy where an admin screened every message sent by a person to make shure they aren't a troll? As I said, very impractical.

[Kay12]

Indeed. However, as an example Wikipedia allows contributions from unregistered users and has managed to build a large and fairly reliable database of information. Perhaps it wouldn't need an admin to check on every input/output pairing in real time, but rather registered users and admins could verify the connections made by the bot in some way. Alternatively, free users could play around with the bot without it learning a thing, and registered users would probably be the only people who actually want to teach it.

Also, even assuming that all the tutors are benevolent and want to teach the CleverBot proper behavior and intelligent conversation, CleverBot will never get further than the illusion of being truly intelligent, like the man in the Chinese room argument. He has a manual which contains all possible discussions and replies in Chinese and the tools for receiving and sending out Chinese symbols - he doesn't know Chinese, but because of the comprehensive manual he can have a seemingly intelligent conversations of indefinite length in Chinese.