Turing Test - Daniel Dennett

[bribribri56]

some sweet info here

[ArxVirtus]

Youtube fails the Turing Test. In the related videos list.....What does a boob-a-licious Chinese girl have to do with the Turing Test?

[khijasmith]

very interesting video thanks

[packe777]

Gregory Chaitin work's on Godel's theorem has finally buried all this positivist dogma of Dennett, Churchland and similar. The Omega number, or Chaitin's constant isn't computable at all. It is irreducible complex algorithm, that is true just by accident. I'ts compatible with Penrose's ORCH-OR, as the platonic information (pattern) embedded at the Plank scale. The implications are mind-boggling. And still today, you have automata like Dennett talking nonsense.

[HomuncuIus]

Dennett's theories are absolute nonsense.

[packe777]

@HomuncuIus His "intuition pumps", "deflationary accounts:", "Sky Cranes/hooks" theories are something only smoked-up person in a bad trip could think of.

[HeyHeyHarmonicaLuke]

@packe777 wtf? Sky hooks and cranes is a great analogy for magic and non magic explanations, and it wasn't coined by Dennett anyway.

[MenziManzo]

An industry that grows so quickly

yet the fruit has only become shinier, smoother, more accessable

the underlieng principles are the same for the last 40 years!

new tools required!

[666arzin]

Hey Kid! I'ma Computa! Stop all the downloadin'!

[TeapotMullah]

@666arzin

Help computer?

[666arzin]

@TeapotMullah

Awww hell naw waddup dawg??

[ronniepage]

uploaded on my bday lol cool



[Grafight23]

Programs hoping to fool humans must rely on the programmer's cleverness to justify a non-sequitur conversation, nothing to do with AI.

The best program can't handle a simple question requiring correct use of previously provided information. For example: "Hi, I am Movielover, what do I love?"

Human: "Movies?"

Bot: "Want to tell me bout love?" or "You're asking way too many questions. Let's talk about something else"

[Serpico261]

@Grafight23 why do you think that a computerprogram that is desgined to have a conversation has nothing to do with AI?

[Grafight23]

@Serpico261 Well, I should rephrase that. I think AI begins when the program can actually learn something from the conversation. A simple macro in a pre-stored sentence, to me doesn't count as AI. Think of Watson, the computer which recently beat the Jeopardy champions. I consider that a real step forward towards AI, because Watson collects info from the questions and even from the human competitors answers to improve its chances.

[Serpico261]

@Grafight23 i absolutely agree with you. A static only program can not be considered as AI. I think even if you have an intelligent algorithm, the program should collect data and build heuristic knowledge. In the other hand could a human also decide not to learn? Or take the IBM chess player program, im not sure if this program plays intelligent...cause its so fast that it does NOT NEED or almost not, productive laziness. it would need hours to make a move if it would calculate fast like humans

[Grafight23]

@Serpico261 Well said, Deep Blue is a computation powerhouse which doesn't learn. It searches EVERY allowed move and counter-move, even the stupid ones, and will pick the one with the most chances of winning. After 3 or 4 moves, Deep blue can solve to checkmate, regardless of what the human opponent does. At that point, it's just a math problem with a solution, not a "game" anymore.

[Serpico261]

@Grafight23 youre right! An MIT prof of cs even said that big blue had no strategy capabilities, well i dont know if this is true, maybe big blue planned his moves only a few levels deep. But it is very simple for a computer to win against humans in a game like chess. I once played against a chess computer and i will never again,its like playing with a mind reader, its so frustrating. Today chess players are ranked by how long and good they can last against a machine, cause they wont win anyways

[ilovehd831]

Unfortunately, ultimately I think the Turing Test would be more of an inhibitor than anything else. It's very possible that some day a machine could be capable of intelligent thought, but differ in personality enough to make it distinguishable from a human. There would be many who would claim the machine is not intelligent because it doesn't pass the Turing Test, when in reality it is intelligent but a bit alien in its behavior.

[Unclesamslair]

@ilovehd831 the Turing test can only positively confirm intelligence not deny it. Failing it doesn't mean anything, only passing counts. Just because something is intelligent doesn't necessarily make it capable of passing the Turing test. Being capable of passing the Turing test however, necessarily makes it intelligent.

[ilovehd831]

@Unclesamslair Yeah, but I see it a possible issue of ethical concern. If there ever were peaceful intelligent machines among us who wanted their freedom, people could argue that they should only be granted human rights if they can pass the Turing Test. Some machines might be intelligent and want their freedom, but not be granted it because their personality differs too much from a human. It presents a discrimination scenario, so ultimately I think the real test will be much more diverse.

[Unclesamslair]

@ilovehd831 do you have a better idea? propose a better test and then we'll talk

[descendantdan]

I never thought I'd see a reference to blade runner in a Dan Dennett lecture.

[moneyjarrod]

Is that Dr. Ramachandran in the audience?

[polymath7]

@moneyjarrod No; it's Jesse Jackson.

[drunkagnostic]

Dennett doesn't think his interviewer's sentient due to a Turing test! He concluded the interviewer's sentient before they met, because Dennett's sentient and is human & the interviewer's human so he's probably sentient too. This logic might suck, but it IS what Dennett used--unless you think he thought during the interview 'would a sentient being respond this way...' What makes humans sentient may be replicable or not, but even Dennett can't rationally conclude this depends on a Turing test.

[richi3mass3]

@drunkagnostic. You realize he offers a Turing test as a metaphor for conversation (i.e., any interlocutor who fails to meet the standards expected by the other party in the conversation fails the "Turing test".)  His invocation isn't, in my opinion, meant to be taken as a literal Turing test. Although, it might as well be, that's the whole point. To my mind, what he said makes perfect sense.

[UrineCredible]

Dennet isn't quite right about the Turing test being a good test for intelligence. Take a look at the TED talk by Jeff Hawkins (I can't post the link here) for a good discussion about this.

[uvauva2]

@UrineCredible

While Hawkins talk is pretty interesting, I wouldn't go so far as to say that Turing is not a good test for intelligence. It seems to me that what he is that the behavioral approach is not the best one to actually pass the test. I would however think that should we pass the Turing test through means other than trying to copy the way we work would be pretty darn interesting.

[tomakist]

I'll admit Turning tests are fun to watch, but I'm surprised at computer programmer's lack of imagination. Its is obvious to anyone who talks to one of these chatbots that its answers are canned and not intelligent. Programmers, imagine more!!

[kid29a]

good points. many people writing "first" on YouTube definitely wouldn't pass the Turing test.

[shubidubar]

Sigh votebots...