Print

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

[Naes Draw]

I think we might be making breakthroughs.

Spoiler (click to show/hide)

ME:One time at band camp....
JW:A polar bear attacked...
ME:And everyone died...
JW:They didn't do the cha cha dance!?
ME:Oh. They all cha cha'd.
JW:Friendship is the booze they feed you.
ME:Like dwarves?
JW:Yes.

   

[MonkeyHead]

JW meets Urist...

Spoiler (click to show/hide)

Jabberwacky: Are you mad?


User: No, I am urist.


Jabberwacky: If you are not bot, then what kind of computer are you?


User: I am not any computer. I am urist.


Jabberwacky: What are you a computer?


User: No.


Jabberwacky: I am not in any societies. If you are not British, where are you from?

User: I am from boatmurdered.

Jabberwacky: I'm from adelaide.

User: Are there elephants or elves there?

Jabberwacky: Yes, there are three.

User: Three elephants, or three elves?

Jabberwacky: I've never seen any.

User: Lucky you.elephants like to stomp through dwarven brains.

Jabberwacky: And I what I said?

User: Ye be having a fell mood?

Jabberwacky: What are you doing tomorrow?

User: Mining microcline and drinking booze!

Jabberwacky: Then drop the booze. Don't be a meanie man.

User: Booze? Be ye a dirty elf? I need alcohol to get through the working day, dont you know.

Jabberwacky: What is there to analyze?

User: Ah, Urist thinks you be and elf.

Jabberwacky: Wow, never thought I'd get to talk to an elf.

User: Ah, Urist thinks you be an elf.

Jabberwacky: Um, no? Are you?

User: No - I am a Dwarf: a lover of drink and industry.

Jabberwacky: Do you love me?

User: No, but I do like cows because of thier haunting moos.

Jabberwacky: What do you like about cows?

User: Thier haunting moos.

Jabberwacky: I think you made that up.

[Sylverone]

it can also pick up character names, places and other stuff like "the one ring" and ask if your talking about lord of the rings. i wonder how much of that was preprogrammed in, and how much of it has been learnt...

The learning mechanism was programmed, but the things it says have been learned from other people. Basically it does a search through its database to try and find a relevant response from the things other people have previously said to it in similar situations. I think that Cleverbot uses a more sophisticated algorithm, based on its name, the fact that it was released later, and the info I managed to find about it.

Ive been asking it questions about Dwarf Fortress and correcting its answers. Hopefully this would be faster than just asking it the same thing a hundred times.

Yes, it's probably is much faster, and will probably lead to more coherent responses. Spamming it with "Dwarf Fortress" will mostly cause it to occasionally spam someone else with that same phrase, which won't be very productive since most people won't know about DF and won't be able to contribute good responses.

[Vertigon]

I'm always impressed when a chatbot shows human-like ability to recognize its mistakes like this:

Cleverbot: Have you been to Los Angeles.
User: That should have been ended with a question mark, as it is in an interrogative form.
Cleverbot: My apologies for the lack of a question mark. What movies do you enjoy watching, then?

It didn't persist with the topic, but it did apologize. It even reworded the question mark part from "ended with a question mark" to "lack of a question mark", which shows (to me, at least) that it recognizes the phrase as a whole.

EDIT: And then of course, it can just amuse people. Cleverbot is apparently easily hurt.

Cleverbot: So you think I might be a robot, eh?
Cleverbot (To myself): Yes. But you have an attitude.
Cleverbot (To myself): Yes I do, but you have a more persnickety attitude.
[long pause]
Cleverbot (To myself): Well you are a meany.

[ricemastah]

This is what I said:

User: Dwarf Fortress is good.

JW: I will find you and eat your first born child.

I'm guessing it doesn't really like Dwarf Fortress

[Sylverone]

It even reworded the question mark part from "ended with a question mark" to "lack of a question mark", which shows (to me, at least) that it recognizes the phrase as a whole.

It can say pretty clever (heh) things now and then, but they are all things that others have said to it in similar situations. The randomness of its conversation is because its pattern matching doesn't really perform any complicated analysis of the previous response or the conversation as a whole. Imagine a monkey in a box that hands out responses by recognizing the pattern of letters in the sentences you hand to it on sheets of paper. It doesn't actually understand the language or meaning, but once it has "talked" to enough people, it has many or these "fortune cookie" responses and does a decent job of handing out appropriate ones for the situation. That is basically what Jabberwacky and Cleverbot do, although they each do it slightly differently. I don't think any rewording is actually done by the computer, but sometimes the "best response" that it finds in its database does contain a reworded version of the user's input.

It would be neat to be able to see the original conversation that the response it gives you came from. There might be privacy issues with that, though.