ALICE Takes Medal At AI Competition

jeffy124 writes: "The Artificial Linguistic Internet Computer Entity (ALICE) has won the bronze and the top marks at the Loebner AI Challenge, a competition based on the Turing test. Silver and gold remain unawarded as silver requires convincing half the judges the AI program is a human, and the gold requires speech interaction rather than text. ALICE repeated as this year's bronze by scoring best among all the entries. She failed to convince half the judges she was human, so she has to stick to bronze. The event took place last Saturday at the London Science Museum."

Comments

Who judges these things?

[pointym5]

Though I agree with Minsky that these things are silly, they'd be at least a little bit more meaningful if some or all of the judges did not know they were judging such a thing. If you took volunteers and told them they were (for example) staffing a career counseling intranet chat system, and had them interact with a blind mix of real people and machine systems, then I'd be more impressed by machines convincing judges that the machines are people.

Re:Who judges these things?

[nickol]

Originally, the Turing test was just as you proposed. However, there is another problem : some testers will mistakenly decide that humans are computers. This means there should be some statistics gathered. So the criteria should be something like this - the probability of recognizing computer as human must be the same as the probability of recognizing human as human. However, this form of the test will lead to the following paradox : testers, informed about the possibility of high artificial IQ will eventually tend to make more mistakes when talking to humans. On the other hand, if all testers will be unavare of existence of the machine IQ, they will treat every correspondent as human.

Re:Who judges these things?

[Christopher Thomas]

If you took volunteers and told them they were (for example) staffing a career counseling intranet chat system, and had them interact with a blind mix of real people and machine systems, then I'd be more impressed by machines convincing judges that the machines are people.

AI programs have already passed this test, repeatedly, if anecdotal evidence counts for anything.

Another poster has already mentioned the AOLiza page. My favourite conversation featured the victim remarking aloud that AOLiza's comments were repetitive, AOLiza asking, "and what does this tell you?", and the user still not cluing in...

A former co-worker of mine told me of another example from the early BBS days. A friend had set up a hacked version of Eliza as "Bob, the Assistant Sysop" before chatbots were common on BBSs. He got a few comments along the lines of "Bob's a nice guy, but he keeps asking me if I have any problems...".

A skeptical audience is harder to fool.

Re:perspective

[ColdGrits]

OT, but here we go.

(1) The deathtoll is around one THIRD of your inflated figure.

(2) Please are NOt "dropping like flies" from Anthrax - the number of fatalities you could count on the fingers of one hand.

(3) If you are not interested in the stuff mentioned on slashdot, then why waste your time reading it?

Perhaps 'tis you who ought to "get some perspective", maybe?

My Experience

[Anonymous Coward]

While it is true that I have to deal with many dolts that are less intelligent than A.L.I.C.E. I still wouldn't call A.L.I.C.E intelligent in any way.

If we assume A.L.I.C.E. to have an intelligence level of 0, then the dolts would be in the negative numbers. Still, no sign of intelligence here. Beam me up, Scotty.

Re:Loebner prize not well respected

[nyjx]

One of the main counterarguments to this is that if the AI purists are "right" then it should be basically impossible to fool a human for very long with a basic pattern matcher. I.e. your going to need some real reasoning, domain knowledge and (probably) learning to fool humans (silver medal).

These bots clearly have value on their own since they can be configured to talk about particular subjects and already act as a first base customer service interface (smarter than most tech support...).

Re:THE CHINESE ROOM

[koreth]

I find Searle's "refutation" to be more of an affirmation. (And I notice that the first URL you provided is pretty anti-Searle once you get into it.) For those who don't want to cut and paste those URLs, the condensed version of his argument is that if you lock an English speaker in a room with a bunch of sophisticated rules about which Chinese characters to write when presented with particular Chinese characters as input, the person's replies could pass a Turing test administered by a Chinese speaker, but that doesn't mean the person in the room speaks Chinese.

One of the classic refutations of the Chinese room experiment is the systems argument: it's true that the person doesn't understand Chinese, but the system made of the person plus the stacks of instructions does. Searle's response to that is to say, "Okay, then suppose the person memorizes the instructions" -- the fact that everything now happens in the person's head still doesn't cause them to understand Chinese.

To me this misses the point of the systems argument; the argument isn't about where the understanding is stored, but whether it exists. If you look at consciousness as a multi-layered entity, in this case the consciousness of the person is one layer below that of the person-plus-instructions. This additional consciousness uses the person in the same way that the person's uses brain cells: the cells can't be said to understand anything, but they make up a larger whole that exists as an emergent property above and beyond the sum of its parts.

Searle's argument also assumes that "understanding" is an almost mystical property that can't be reduced to a fixed set of rules. Which to me is just ridiculous unless you assume the existence of a noncorporeal soul; if the brain is really the seat of consciousness then consciousness does reduce to a fixed set of rules (laws of physics and chemistry) applied to an extremely complex system. His response to that argument (something about water pipes, if I recall) again misses the point by such a huge margin that it's barely worth mentioning.

I have no problem with the possibility that what I think of as "me" is simply a byproduct of a complex physical process, and that similar byproducts can exist in other complex systems. Doesn't mean we'll be able to detect them or communicate with them any more than a cluster of six brain cells can communicate with us.

Re:Who is the poor human that failed the Turing te

[JustinKSU]

Alice: No I have not?
Me: No you have not what?
Alice: Are you serious? I don't think I ever have what. What's it like?

"I don't think I ever have what" Do they have grammar check on this thing?