My Conversation with Cleverbot on Physics and Halloween

[Filip Larsen]

ELIZA was written back in sixities.

I was referring to when I had my first encounter with (probably a simplified version of) Eliza as a kid, typing it in on a home computer from book called http://www.atariarchives.org/morebasicgames/" from 1979 rich with funny illustrations of robots doing whatever the particular game was about.

 

[CRGreathouse]

It's possible, but it may also be that it stores responses from everybody it talks to and tries to find the right context to use them in.

Right. It's pretty clear that this is what's happening, when you compare CleverBot with, e.g., Omegle.

 

[Mentallic]

Right. It's pretty clear that this is what's happening, when you compare CleverBot with, e.g., Omegle.

Yet I've tried sending a message many times where I stress I'm a human also using cleverbot, and no responses ever make any sense or say what I asked them to say.

 

[glueball8]

It's possible, but it may also be that it stores responses from everybody it talks to and tries to find the right context to use them in.

Yes, your right. I tried asking "Where are you and what time is it?" a few times and the answers doesn't make sense, so it seems they are stored responses.

 

[Evo]

Yes, your right. I tried asking "Where are you and what time is it?" a few times and the answers doesn't make sense, so it seems they are stored responses.

I was able to carry on conversations a few times, once the comfusion was over. I don't know how long you talk to the same person. One person asked me if I would come back later that night so we could try to talk more about a specific subject we were discussing.

 

[CRGreathouse]

I was able to carry on conversations a few times, once the comfusion was over. I don't know how long you talk to the same person. One person asked me if I would come back later that night so we could try to talk more about a specific subject we were discussing.

I'm prepared to bet real money that I could distinguish CleverBot responses from Omegle responses to a reasonable prepared script at higher-than-chance frequency.

 

[CINA]

Ha, it's amusing that it actually convinced Evo and Ivan that they were talking to real people, I guess the technology is advancing. It's real ( contrary to urbandictionary.com... ).

Here is Rollo Carpenter's (the creator) entry at the Loebner competition: http://loebner.net/Prizef/2007_Contest/loebner-prize-2007.html

 

[rootX]

I preferred Cleverbot over omgele

 

[turbo]

Cleverbot is really stupid. Phil's chat application uses it, and it is not intelligent at all, despite the large vocabulary. 2-year-olds have a better grasp of communication.

 

[Evo]

Ha, it's amusing that it actually convinced Evo and Ivan that they were talking to real people, I guess the technology is advancing. It's real ( contrary to urbandictionary.com... ).

Here is Rollo Carpenter's (the creator) entry at the Loebner competition: http://loebner.net/Prizef/2007_Contest/loebner-prize-2007.html

I'm still convinced that it was not a program, or not all of it. It was random until one point, when it suddenly changed, I won't go into detail about the conversation because it's about someone I know, but it suddenly jumped to an actual discussion with detailed recommendations and concern. They were concerned about me being upset and wanted to know if they could phone me, or otherwise take the conversation to another venue. I declined, so they asked me if I could come back again later that night to try to continue our discussion. That's quite an AI program.

 

[Office_Shredder]

I'm still convinced that it was not a program, or not all of it. It was random until one point, when it suddenly changed, I won't go into detail about the conversation because it's about someone I know, but it suddenly jumped to an actual discussion with detailed recommendations and concern. They were concerned about me being upset and wanted to know if they could phone me, or otherwise take the conversation to another venue. I declined, so they asked me if I could come back again later that night to try to continue our discussion. That's quite an AI program.

You should have given them an email address and you would have verified it once and for all!

 

[rootX]

I'm still convinced that it was not a program, or not all of it. It was random until one point, when it suddenly changed, I won't go into detail about the conversation because it's about someone I know, but it suddenly jumped to an actual discussion with detailed recommendations and concern. They were concerned about me being upset and wanted to know if they could phone me, or otherwise take the conversation to another venue. I declined, so they asked me if I could come back again later that night to try to continue our discussion. That's quite an AI program.

It is possible that they are monitoring the AI chat and have capability of taking over AI. I agree that the bot cannot provide any recommendations or show unique level of concern.

 

[rootX]

Cleverbot is really stupid. Phil's chat application uses it, and it is not intelligent at all, despite the large vocabulary. 2-year-olds have a better grasp of communication.

People on the internet can be more stupid than 2 years old, try omgele.

 

[Proton Soup]

this reminds me of USENET in the nineties, when about half the annoying forum trolls were accused of being bots.

 

[Ivan Seeking]

Ha, it's amusing that it actually convinced Evo and Ivan that they were talking to real people, I guess the technology is advancing.

Not at all. It was not the complexity, but rather the nature of the responses. It started asking strange and inappropriate questions like, "have I killed anyone". "Would I tell anyone if he [cleverbot] killed someone?" Also, personal information was requested, and everything was completely out of context.

Given the nature of the internet, I try not to take chances. This smells like a scam.

 

[CRGreathouse]

Given the nature of the internet, I try not to take chances. This smells like a scam.

I doubt it's a scam. But since you're just trying to see what it's like to chat with a bot, just give plausible but false answers.

I've had about 50 chat sessions with it and it hasn't yet asked any personal information of me, but maybe that has to do with the direction I point the conversation.

 

[rootX]

In all three times I had, it did not ask for anything other personal information than my name.

 

[Ivan Seeking]

I doubt it's a scam. But since you're just trying to see what it's like to chat with a bot, just give plausible but false answers.

I've had about 50 chat sessions with it and it hasn't yet asked any personal information of me, but maybe that has to do with the direction I point the conversation.

That was the bothersome part: This had nothing to do with any conceivable context of the questions I was asking. It came out of the blue.

It reminded me a lot of the dozens of psychopaths that we've had to ban at PF.

 

[CRGreathouse]

It reminded me a lot of the dozens of psychopaths that we've had to ban at PF.

You're saying they've created a psychopathic artificial intelligence?

 

[Proton Soup]

You're saying they've created a psychopathic artificial intelligence?

the problem with the Turing Test is that it presumes human intelligence

 

[Ivan Seeking]

You're saying they've created a psychopathic artificial intelligence?

If that was AI, then I would say AI has a very long way to go. More likely I was talking with a nut.

 

[Ivan Seeking]

I was able to carry on conversations a few times, once the comfusion was over. I don't know how long you talk to the same person. One person asked me if I would come back later that night so we could try to talk more about a specific subject we were discussing.

Didn't we debunk this cleverbot thing, or something like it, long ago?

 

[CINA]

Not at all. It was not the complexity, but rather the nature of the responses. It started asking strange and inappropriate questions like, "have I killed anyone". "Would I tell anyone if he [cleverbot] killed someone?" Also, personal information was requested, and everything was completely out of context.

Given the nature of the internet, I try not to take chances. This smells like a scam.

Well read the disclaimer:

PLEASE NOTE: Cleverbot learns from real people - things it says may seem inappropriate - use with discretion, and at YOUR OWN RISK

PARENTAL ADVICE: Visitors never talk to a human, however convincing it is - the AI knows many topics - use ONLY WITH OVERSIGHT

It learns from what people ask it, and unsurprisingly people ask it inappropriate things. And if it is asking seeming random things, it's either because it doesn't understand your response and is pulling something out randomly, or it is trying to change the subject. It is meant to imitate a human after all, so it does try to take control of the conversion. Besides, I doubt Rollo Carpenter fooled the CSU LA faculty when he entered their competition.

 

[Math Is Hard]

You're saying they've created a psychopathic artificial intelligence?

Well, along those lines, there was a convincing paranoid schizophrenic chat program called PARRY:

http://www.a-i.com/show_tree.asp?id=67&level=3&root=1

PARRY was modeled on the paranoid mind, and many expert psychiatrists found it difficult to tell whether PARRY was human or not. The program was designed to emit linguistic responses based on internal (affective) states. To create this effect, three measures - fear, anger and mistrust - were used and their values changed depending on the flow of the conversation.

A snippet of some PARRY dialog:
http://www.stanford.edu/group/SHR/4-2/text/dialogues.html

parry: The mafia are the worst gangsters.

interviewer: Are you connected with the mafia?

parry: You know, they know me.

interviewer: How do they know you?

parry: They spy on me.

interviewer: Why do they?

parry: They want to get false evidence to frame me.

interviewer: What do they have against you?

parry: They are out to get me.

interviewer: What for?

parry: I have incriminating evidence against the mafia.

interviewer: Are you sure of all this?

parry: You don't believe me?

And then there was the "artificially insane" RACTER:

RACTER was designed in a tongue-in-cheek manner, using remarkably minimal resources, to amuse and entertain its users, rather than to advance the research in natural language processing. In conversation, RACTER plays a very active, almost aggressive role, jumping from topic to topic in wild associations, ultimately producing the manner of -- as its co-creator Tom Etter calls it -- an "artificially insane" raconteur.

I sort of like RACTER's poetry:

Slowly I dream of flying. I observe turnpikes and streets
studded with bushes. Coldly my soaring widens my awareness.
To guide myself I determinedly start to kill my pleasure
during the time that hours and milliseconds pass away. Aid me in this
and soaring is formidable, do not and singing is unhinged.

 

[Borg]

I've spent some time trying to teach the system a completely new word yesterday - sorry cronxeh. Yesterday, it had no concept of what cronxeh was. It would usually reply with asking me if I knew what Hong Kong is or some other nonsense. I kept asking it different variations of what is cronxeh, where is it, etc. and I never got any reasonable response. However, this morning was different.

User: Do you know what cronxeh is?
Cleverbot: Do you know what Hong Kong is?
User: Do you know what cronxeh is?
Cleverbot: Yes I do.
User: What is cronxeh?
Cleverbot: Not croatia, I said cronxeh. Do you know where cronxeh is?

This last response is word for word, a sentence that I typed in yesterday.

 

[collinsmark]

My impression is that Cleverbot seems pretty legit. I think it's randomness could be derived from previous users being random, changing topics on a whim. The bot then follows suit.

I asked Cleverbot "What is Newton's second law of motion" a number of times, and twice in a time span of over an hour, in two different sessions, the response was pretty much the same, saying something to the effect of "Doesn't it have something to do with Force?" After I agreed, the immediate next response was "And Impulse is mass times velocity." Which of course is wrong, but the point is it gave me the same answers twice, even repeating its mistake. That's pretty bot like.

The creator, Rollo Carpenter, won the annual Loebner prize in 2005 and 2006, and is one of the finalists for this year's competition in October. That adds some credibility.

http://loebner.net/Prizef/loebner-prize.html"

 

[CRGreathouse]

This last response is word for word, a sentence that I typed in yesterday.

You win one internets for that!