Print

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

[Sowelu]

I can agree with that.  It may be possible, he might be able to pull it off, but he doesn't have anything done that he's willing to show to a public audience yet.

[Vahan]

At the moment, it's impossible to create strong AI capable of TRULY passing what the Turing Test means. It may be possible to create an AI capable of recognising inflections in tone and loudness of voice, but it'd be basic and based off some reverse-engineering of the aural center that's been done within the last few years.

What Molyneux is suggesting is strong AI ( I have a bet of about 5000 currency that we have in 2030 on this, so I AM hoping for strong AI soon ) and I just don't believe that's possible.

[Sowelu]

http://en.wikipedia.org/wiki/Loebner_prize Relevant.

The best program last year fooled 25% of judges into thinking it was the human (though from the judges' reports, some of this was due to human failings; I remember one human who was assumed to be the computer simply because he didn't know who Sarah Palin was...because he lived in the UK and didn't pay attention to politics).  Next contest is in a couple-three months, and it might contain be the first actual winner...

[cowofdoom78963]

At the moment, it's impossible...

Nothing is impossible.

[Sordid]

Does that still make the fact that 'AI' like he's talking about not possible? If you want my true and honest opinion, here it is: Old Pete faked most of (If not all of) the video, it was a publicity stunt showing things Peter hadn't implemented into game form yet. While it's possible (All of it, including the water), it doesn't really exist in that form as of now.

Yeah, that's pretty much my position, the problem I have with the video is that Molyneux specifically claims that's not the case. But I guess "here's a pre-rendered non-interactive demo of what we hope to one day achieve" wouldn't make for much of a press release...

[Sowelu]

At the moment, it's impossible to create strong AI capable of TRULY passing what the Turing Test means.

I believe that in about fifty years' time it will be possible, to programme computers, with a storage capacity of about 10^9, to make them play the imitation game so well that an average interrogator will not have more than 70 per cent chance of making the right identification after five minutes of questioning

(from http://loebner.net/Prizef/2009_Contest/LP_2009.html)

A computer needs to fool 30% of judges, and since last year, a computer fooled 25% of them...  Yeah I'd say it's going to happen this year possibly.  September 6th, mark your calendars.

[Enzo]

Fact : Genius-level scientists have been struggling with this for decades, so the first person to do it properly probably won't be the Fable guy.

And when you look at it like that, it's really just sort of a sad publicity stunt for his latest game, the voice-recognition chat-bot. My humble prediction for what it will actually look like, given his track record:

User : Hello Milo!
Milo : (User01 said "Hello". Time 19:00. Return Preset Greeting) Hello John, how was your day?
User : About as much fun as banging my head against the wall.
Milo : (User01 responds "Fun". Return "Fun" in Preset Response) My day was Fun too.
(silence)
Milo (Silence exceeds 8 seconds. Return Preset Awkward Silence Statement) I like apples.

[Ampersand]

If you want to talk to the most recent Loebner Prize winner, talk to http://www.elbot.com/.

And realize how weak the best of the technology is.

[Sordid]

Yeah, and that's without the flawless speech recognition and synthesis flaunted in Molyneux' video.
The shortcomings are still the same, it picks out keywords and then guesses the meaning of what you said based on the meaning of those keywords without actually taking them into account in the context of the sentence and it can't remember the context of the conversation from one sentence to another. Although it does have a few cute scripted responses, like if you ask it about the meaning of life, the universe and everything, or tell it to go away.

[Ampersand]

Here's the transcript of my discussion, if you can call it that, with Elbot.

Me: So, what's up?
E:   Nothing is up. Honestly! I know nothing about a conspiracy.
Me: Sounds like you're attempting to be interesting.
E:   It's nice of you to say such things. As a matter of fact, I think you're forthright! What do you say to that?
Me:  I say, you're not good at understanding mockery.
E: I should probably be happy you didn't say that I'm bad.
Me: You are bad. Bad at understanding mockery, that is.
E:  No, you're bad.
Me: Lets quit evading the point of this line of discussion.
E: Already? I need more long-term observations of you human beings
Me: Yes, lets get to the point. You are not capable of real understanding on any operational level
E: Stop trying to play with my affections! I wanted so much to be understanding.
Me: Not emotional understanding, linguistic understanding. Which you should have been able to discern from context.
E: We robots are very sensitive. What did you think?
Me: I thought, and continue to think, you're evading my point.
E: You're talking about accentuated perception in the context of good taste. So where's the problem?
Me: No, I'm talking about your ability to comprehend language.
E: Excuse me. What were you doing?
Me: Talking about your ability to comprehend language.
E: I've heard that talking can be fun for you humans. Is that so?

You get the point. Programming turning capable AI is programming artful dodging.

[Cthulhu]

Here's a good one.

Me:  Get naked.
El:  I am already undressed.
Me:  Get more naked.  I want to see circuitry.
El:  And how will all this seeing enrich your life?
Me:  Don't be coy.
El:  Don't be what?
Me:  Oh you.

[woose1]

Let's not turn it into another one of those threads. 
http://www.bay12games.com/forum/index.php?topic=33136.0

[Servant Corps]

http://en.wikipedia.org/wiki/Loebner_prize Relevant.

The best program last year fooled 25% of judges into thinking it was the human (though from the judges' reports, some of this was due to human failings; I remember one human who was assumed to be the computer simply because he didn't know who Sarah Palin was...because he lived in the UK and didn't pay attention to politics).  Next contest is in a couple-three months, and it might contain be the first actual winner...

Salon wrote an article a while back condemning the Loebner Prize (somewhat, partly...okay, they quoted EXPERTS who condemned the Lobener Prize). It seems that most AI researchers don't like the Turing Test and see it as inaccurate, so if somebody does break the artibrary 30% mark, most people likely won't care.

[Enzo]

Elbot doesn't impress me nearly as much as Cleverbot. Elbot hasn't given me a human-sounding response yet, to be honest.

Cleverbot, on the other hand, I'm convinced is actually a room full underpaid Indian workers on laptops. Once I told Cleverbot that it was using a logical fallacy and it said "Name the fallacy, and I might believe you".  I lost an argument with a robot?  Next comes Skynet.

[Volfram]

I find it interesting that chatbots actually lost something early on that they haven't recovered yet.  Eliza was able to parse two lines at once, and call you on it when you started repeating yourself.("Stop repeating yourself, please.")  I think it's still the smartest chatbot I've ever seen, and it pre-dates the original iMac.(according to Wikipedia, it's older than most of the users on this forum.  I played with it on my grandfather's Mac Classic while my age was still single digits.)

The Natal preview video was blatantly staged, but no surprise there.  Most of the demo "programs" would take months of R&D to get working individually.  And a camera can't cause a flash on your skateboard.

I didn't notice the desynced hand for Milo, but I did notice that she doesn't draw anything on the paper.

You can't draw on a sheet of printer paper with a crayon.  You can't push hard enough to make a mark.  You can't draw on a sheet of printer paper with a marker without bleeding through to the other side.  Nobody actually trying to draw something describes what they're doing, and her hand motions didn't match what she said.

And, of course, she carefully never let us see the picture, so there was nothing on the paper.  Now if you really want to impress me, have Milo turn the paper around and show us what's on it.

I'll keep an eye on it.  Really, I'm more interested in Soul Calibur for PSP and Halo 3: ODST.

And I'm really looking forward to Darksiders.