Cleverbot-the online AI converser
Discussion-the action or process of communicating about something

I'm no programmer or anything, but I wanted to discuss this piece of conversational AI and more specifically, how it "learns." Curious to know if any of you could explain to me in semi-Layman terms how it works.

With big enough database of recorded conversations, this method can be very convincing. I used to run a simillar bot on the omegle, started off with the whole Star Wars script ( ) which was the base for my recorded conversations, and added all of the responses from there. While my bot was nowhere as good, it was actually pretty dumb, it still gave some damn funny conversations, and as the bot was running for longer time it was getting more interresting words to use.

I guess what I mean to say is does it have the capability to learn how to change its programming? Like could you teach it programming and convince it to change itself given the right keyword or something?

no

it simply records all conversations, and matches all recorded - even those recorded "live" - when responding to new requests. Nothing special in this really. It simply echoes what humans said to it, and tries to do it in a smart way.


For example it has recorded a conversation between two people, and it was something like:
- "do you know a good joke?"
- "yes. php."

now it has recorded "do you know a good joke?" == "yes. php."

and if someone asks about good joke, it will respond in the same way. Of course it strips formatting, tries to match only words etc and is much smarter (eg. recognizes vovels, different writing, typos etc) than simplified example, and can follow conversations much deeper than question-response.

Plus it can have a thousand of "do you know a good joke?" matches. Then if someone asks this, gets pre-recorded response, the bot can ask for another joke and receive brand new reply that it will include in it's database.

This post was edited by nuvo on Aug 16 2013 05:31pm

so it never formulates anything, just copies? Hence some of the really weird random answers?

That's the bots basics yes. If you're looking for a semantic processer, as in something that attempts to understand your sentence and formulate a reply on its own, you should look at IBM's Watson. Cleverbot is more akin to the Google search engine: It matches search queries (in cleverbot's case, user input) to the most relevant answer in its database. As opposed to Google, however, it only serves you the top answer rather than a whole pile of them. Once the database is wide enough, fine tuning can be performed around meaningful keywords with certain mix and match algorythms specific to each, but I don't think the cleverbot's intent is to go that far in the demonstration.

Bottom line, it's a mimic of search engine technology, not a semantic processer.

This post was edited by flyinggoat on Aug 16 2013 08:29pm

as you know a computer works in algorithms, These Super computers don't just work in 0 & 1

0 = No
1 = Yes


this isn't done once but thousands upon thousands of time to calculate precise co-ordinates.

if a 1 = 1,000 + the 0 can be in half of those yes because they all circulate together

0=1=0=1=0=1=0=1=0=1=0=1 at the end of the sequence The Yes command restarts at the 0

do you offer lessons for fg?