I'll be back in 2013..... with Cleverbot !!

I’ll give it credit for being far more entertaining than previous conversation engines, but it is not remotely close to a Turing worth AI. I’d give it 5%. Maybe. On one of it’s good days.

Googling “Cleverbot turing” gives plenty of results about the issue. But the bottom line is, the test isn’t one guy looking at the responses of a few people pushing to break the system and revelling in the absurdities (as I’m sure many of us do). It’s a procedure involving many people, apparently, and that 59% is a result of that procedure. It doesn’t mean you can’t spot in 10 seconds that you’re talking to a robot (sometimes that’s how long it takes for it to break). It just means that after that procedure, that was the number it arrived at.

Also, that percentage has to do with (as I understand it) how many people thought they were looking at conversations between two humans (with, as I understand it, a two-human control, no less). It’s not something a judge rates abritrarily, it’s a statistic.

EDIT - Sorry, I think I’m being a bit confrontational. If so, please don’t take it personally or anything.

EDIT 2 - Another thing which I’m sure most people overlook:

Oh I didn’t think you were being confrontational.

Are you saying that the judging was people looking at a transcript of conversations, rather than the judges actually participating and interacting with it? Because that’s very dodgy, the human conversation partner might know how to hide cleverbot’s flaws.

The differing computational power could definitely make a difference sure. It would be interesting to try a fully powered one.

Ok, enough hearsay from me, here’s an actual quote:

So I was slightly wrong myself - a bit of was WAS an arbitrary rating. But there was a human control group, and the result is more statistics than otherwise.

And there are a lot of arguments along the lines of “Cleverbot passed as human because in most Chatrooms many people actually behave like that: nonsense passes for entertainment and sudden changes of topic aren’t uncommon”.

I suppose we should be grateful that Cleverbot actually spells things correctly.

Most of the time. I did find typos. Which, curiously enough, may re-inforce the illusion of a human. :slight_smile:

And re computational power, apparently Cleverbot makes 3 passes over its banks (whatever that means) before coming up with an answer. The powerful version, no doubt inspired by Deep Thought, made 42.

EDIT - And let’s not forget the human factor: I can totally envision human-human conversations where one of them tries to be as bot-like as possible.

You know the bot on IFMud that imitates an African grey parrot, every so often saying “AWWWK! Word on the street is [off-topic fact regarding something just discussed]”? For a long time, when reading ClubFloyd transcripts, I thought it was a person doing an incredibly obnoxious roleplay.

Oh, this is cute. It got me to admit it’s human.