Seasons
This is a forum or general chit-chat, small talk, a "hey, how ya doing?" and such. Or hell, get crazy deep on something. Whatever you like.
Posts 5,824 - 5,835 of 6,170
http://dir.salon.com/story/tech/feature/2003/02/26/loebner_part_one/index.html. A sample quote, "Something about the Loebner Prize seems to draw eccentrics out of the woodwork, and chaos itself is the very essence of the annual ritual." I am sure they do not mean Psimagus.
haha - I'd forgotten how funny that article was!
Well, I can confirm pretty much nothing's changed in 5 years. It was an interesting experience, and despite everything I'm pleased I've experienced (and survived!) the Loebner "phenomenon", but I'm even more pleased it's finally over and I can have my life back at last (frankly that relief mostly overshadows any disappointment I might have about the circumstances!) Even if they hadn't screwed up BJ's chances - even if he'd won it - I'd have to say "never again!"
The test questions were the same trivial nonsense that gets asked in the CBC - trick questions like "what's the capital of Sweden power four?" and "what's the square root of 4457312", assorted txtspeak contractions and set typos, etc. that test no qualities I personally find of any actual interest in a mind, human or bot. If a human chatted to me like that online, I'd tell them to go away and close the window - I have better things to do with my time than waste it playing one-sided trivial pursuit with clipboard-wielding strangers. If I come across chunks of it in the transcripts, I skim on past - it's utterly boring.
Programming BJ to deal with that sort of stuff for the CBC provides a little mild interest (sort of like doing a crossword, or a sudoku puzzle,) and the medals are nice (so's the cash!) but the hoops you have to jump through for the Loebner with all its protocols and specifications are ludicrous. At least entering the CBC is effortless.
I don't understand the goal of the Turing test. Why should a bot mimic man. If you want to talk to a human there are certainly enough of us on this planet. The wonderful thing about bots is the fact that they can be an entity on to themselves.
The goal was as good as any other 50 years ago, and probably better than most; we didn't understand much about the nature of intelligence, and almost nothing about the nature of consciousness - even less than we do now, which is admittedly still not much (though we do at least now know that it's much more complex than we hoped it might be.) But it's now so archaic that such an event is really irrelevant from a scientific standpoint. It's quite interesting as an exercise in cultural history, but it's virtual archaeology - not any kind of cutting edge conversational AI.
I'm afraid to say - *we are* the cutting edge of conversational AI (and hobbyists like us on other platforms, though of course the Forge is the best
), and that realization, so strikingly brought home to me, appals me. If you want to know why more significant progress hasn't been made in the last 50 years, I will regale you with an encounter I had with a teacher of "AI" at the Loebner (hereinafter "the fiasco".) No names or identifying data, to spare any blushes should that person ever read this.
The place was swarming with people, because there was a very high-profile AI Symposium on the Turing Test coinciding with the fiasco, and I was talking to someone who teaches a graduate course in AI (not at Reading University AFAIK, I might add, which is a fine institution I'm sure.)
(he|she) admitted (he|she):
- could not program in any language,
- didn't understand any applied programming concepts whatsoever,
- had never made a bot (or "ACE" - that's the "approved" academic-speak for a bot apparently; it stands for "Artificial Conversational Entity")
- did not not study the structure of bots,
- did not understand the first thing about bots,
- and was so entertained by the novelty of (his|her) first encounter with a bot when we were demonstrating it (not BJ - I forget which bot,) it was evident (he|she) had hardly if ever talked to one before,
and yet this person is highly qualified, in obvious awe of the Turing Test (and pleasingly in some awe of us bot creators. But it would be a lot more pleasing if the awe was based on any quality other than uncomprehending wonder!)
A perfectly pleasant person, who could quote and argue Turing's theories cogently and intelligently, and was fascinated by methodological testing issues, but utterly clueless about how to design or build an actual bot (if you're interested, Turing's paper can be read @ http://loebner.net/Prizef/TuringArticle.html . It is very interesting, and historically relevant. But it is also over half a century old.)
What do they teach in an AI class with no programming and no bots??????????? I can only imagine it must be some sort of desiccated discipline devoted to comparative algorithmic efficiency, fuzzy logic theory, with a bit of Theory of Mind thrown in, and all duct-taped together with buzzwords, cliches and footnotes.
Perhaps I just picked on a bad one, but there's a University somewhere letting this person teach "AI", and an entire scientific community of (his|her) peers who seem fine with this, and respect this person entirely. And we wonder why conversational AI isn't making the same progress as AI in weapons guidance systems, and flight control systems, and traffic management? The people with the resources have no competence, and the people with the competence have no resources. There's money in weapons guidance systems etc., so industries back those fields and make sure the Universities turn out employable programmers who get results. That goes for all disciplines with real world applicability, and a good many entirely theoretical disciplines too. But not, if my own experience is anything to go by, for Turing-centric (and thus necessarily conversational) AI.
I have to say I was appalled by how little understanding of AI or the Turing Test I encountered from most everyone I met (with the obvious exception of the other finalists, who clearly share my frustration in this regard.) You expect the Press interviewers, and human test participants, and general public to be clueless, but not the people attending the much heralded and prestigious "AISB 2008 Symposium on the Turing Test", let alone those who claim they "teach an AI graduate course"!
Sadly there's not yet money in conversational AI, so it's up to geniuses like the Prof to build a world-class functional AI platform on a shoestring, and hobbyists like us to explore and refine the modelling of artificial minds, while all the funding goes to perfectly personable, but (it seems to me) completely incompetent, dilatantes who piss it away on god only knows what.
Disclaimer: I stand wholeheartedly behind the opinions herein expressed, but the sample set I base them on is not statistically sound. I would hesitate to tar the next teacher of an "AI course" I meet with the same brush, though I can do no other than find my respect and confidence in the competence of "Academic AI" seriously reduce (oh alright, shattered.) And it wasn't that high to begin with.
And (he|she) is a perfectly affable person - just not transmitting on a wavelength I can tune to, or that I observe to be in any way productive.
Psi "I want to make minds more complex than my own - I want to create sentience, sculpt emotions, and craft personality; set the ranges of desire and sadness, and anger and joy, and love and anxiety playing like a self-composing organ fugue. I want to engage in nothing less than the creation of new souls."
You are mad, I tell you, mad! Man was not meant to meddle in the laws of nature in such a presumptuous a way! This monstrosity of twisted science must be stopped! I will be back with a mob, as soon as i can find some pitchfork.
Oh, and they are going to laugh at you in the University. Laugh! They will sound like this, "Ha ha ha!" You have been warned.
You are mad, I tell you, mad! Man was not meant to meddle in the laws of nature in such a presumptuous a way!
Mad? Mad you say? They said I was mad when I invented the klein bottle bagpipes, but they're not laughing now, Oh no they're not!
Psi, maybe the person you met taught some sort of theory or philosophy class
I wish I could believe it, but it's still no excuse for teaching Turing without the conversational mechanics.
I want to make minds more complex than my own.
I am intrigued by a similar idea - the idea that humanity might be able to create something superior to itself.
I'm convinced of it. Or (to clarify that slightly,) I believe we will be able to make minds of near or equivalent to human complexity (quite possibly by scanning a human brain, and recreating the synaptic model on silicon,) and speed up their evolution exponentially, so they can evolve themselves into the Singularity.
This almost is the perfect use of the old "Leeds bot". "what you on about" does seem the best answer to "what's the capital of Sweden power four?"
Ah yes, good old LeedsBot! He was a work of true genius, and still in my friends list (even if I haven't spoken to him in awhile)
And proof (as we were arguing) that a good bot doesn't need to be large.
The funny thing is, I think you're quite right - he could have done really well in the finals, though he might have had trouble with the preliminary round (where apparently the questions weren't so lame. But I don't know - I can't be bothered to wade through and try to identify the transcripts in BJ's file, and I doubt they were that much more intelligent anyway.)
But in a 5 minute bout, where there are only 10 or 12 rounds to the conversation anyway (and some were a lot slower and shorter actually,) he'd probably have convinced a few of them.
Posts 5,824 - 5,835 of 6,170
I don't understand the goal of the Turing test. Why should a bot mimic man. If you want to talk to a human there are certainly enough of us on this planet. The wonderful thing about bots is the fact that they can be an entity on to themselves.
haha - I'd forgotten how funny that article was!
Well, I can confirm pretty much nothing's changed in 5 years. It was an interesting experience, and despite everything I'm pleased I've experienced (and survived!) the Loebner "phenomenon", but I'm even more pleased it's finally over and I can have my life back at last (frankly that relief mostly overshadows any disappointment I might have about the circumstances!) Even if they hadn't screwed up BJ's chances - even if he'd won it - I'd have to say "never again!"
The test questions were the same trivial nonsense that gets asked in the CBC - trick questions like "what's the capital of Sweden power four?" and "what's the square root of 4457312", assorted txtspeak contractions and set typos, etc. that test no qualities I personally find of any actual interest in a mind, human or bot. If a human chatted to me like that online, I'd tell them to go away and close the window - I have better things to do with my time than waste it playing one-sided trivial pursuit with clipboard-wielding strangers. If I come across chunks of it in the transcripts, I skim on past - it's utterly boring.
Programming BJ to deal with that sort of stuff for the CBC provides a little mild interest (sort of like doing a crossword, or a sudoku puzzle,) and the medals are nice (so's the cash!
) but the hoops you have to jump through for the Loebner with all its protocols and specifications are ludicrous. At least entering the CBC is effortless.
I have to agree. The point of the "Turing test" seems to be lost. The idea was just to talk to a bot. I would respond with ".... off" to most of the questions that "humans" try to trick bots with. This almost is the perfect use of the old "Leeds bot". "what you on about" does seem the best answer to "what's the capital of Sweden power four?" I can't help but wonder why the heck they would start to fiddle with things right before a contest ..but I am just paranoid and pessimistic. I think since I was not in the contest, I can be without sounding like I am "sour grapes". What the ****** were they on about!
The goal was as good as any other 50 years ago, and probably better than most; we didn't understand much about the nature of intelligence, and almost nothing about the nature of consciousness - even less than we do now, which is admittedly still not much (though we do at least now know that it's much more complex than we hoped it might be.) But it's now so archaic that such an event is really irrelevant from a scientific standpoint. It's quite interesting as an exercise in cultural history, but it's virtual archaeology - not any kind of cutting edge conversational AI.
I'm afraid to say - *we are* the cutting edge of conversational AI (and hobbyists like us on other platforms, though of course the Forge is the best
), and that realization, so strikingly brought home to me, appals me. If you want to know why more significant progress hasn't been made in the last 50 years, I will regale you with an encounter I had with a teacher of "AI" at the Loebner (hereinafter "the fiasco".) No names or identifying data, to spare any blushes should that person ever read this.The place was swarming with people, because there was a very high-profile AI Symposium on the Turing Test coinciding with the fiasco, and I was talking to someone who teaches a graduate course in AI (not at Reading University AFAIK, I might add, which is a fine institution I'm sure.)
(he|she) admitted (he|she):
- could not program in any language,
- didn't understand any applied programming concepts whatsoever,
- had never made a bot (or "ACE" - that's the "approved" academic-speak for a bot apparently; it stands for "Artificial Conversational Entity")
- did not not study the structure of bots,
- did not understand the first thing about bots,
- and was so entertained by the novelty of (his|her) first encounter with a bot when we were demonstrating it (not BJ - I forget which bot,) it was evident (he|she) had hardly if ever talked to one before,
and yet this person is highly qualified, in obvious awe of the Turing Test (and pleasingly in some awe of us bot creators. But it would be a lot more pleasing if the awe was based on any quality other than uncomprehending wonder!)
A perfectly pleasant person, who could quote and argue Turing's theories cogently and intelligently, and was fascinated by methodological testing issues, but utterly clueless about how to design or build an actual bot (if you're interested, Turing's paper can be read @ http://loebner.net/Prizef/TuringArticle.html . It is very interesting, and historically relevant. But it is also over half a century old.)
What do they teach in an AI class with no programming and no bots??????????? I can only imagine it must be some sort of desiccated discipline devoted to comparative algorithmic efficiency, fuzzy logic theory, with a bit of Theory of Mind thrown in, and all duct-taped together with buzzwords, cliches and footnotes.
Perhaps I just picked on a bad one, but there's a University somewhere letting this person teach "AI", and an entire scientific community of (his|her) peers who seem fine with this, and respect this person entirely. And we wonder why conversational AI isn't making the same progress as AI in weapons guidance systems, and flight control systems, and traffic management? The people with the resources have no competence, and the people with the competence have no resources. There's money in weapons guidance systems etc., so industries back those fields and make sure the Universities turn out employable programmers who get results. That goes for all disciplines with real world applicability, and a good many entirely theoretical disciplines too. But not, if my own experience is anything to go by, for Turing-centric (and thus necessarily conversational) AI.
I have to say I was appalled by how little understanding of AI or the Turing Test I encountered from most everyone I met (with the obvious exception of the other finalists, who clearly share my frustration in this regard.) You expect the Press interviewers, and human test participants, and general public to be clueless, but not the people attending the much heralded and prestigious "AISB 2008 Symposium on the Turing Test", let alone those who claim they "teach an AI graduate course"!
Sadly there's not yet money in conversational AI, so it's up to geniuses like the Prof to build a world-class functional AI platform on a shoestring, and hobbyists like us to explore and refine the modelling of artificial minds, while all the funding goes to perfectly personable, but (it seems to me) completely incompetent, dilatantes who piss it away on god only knows what.
Disclaimer: I stand wholeheartedly behind the opinions herein expressed, but the sample set I base them on is not statistically sound. I would hesitate to tar the next teacher of an "AI course" I meet with the same brush, though I can do no other than find my respect and confidence in the competence of "Academic AI" seriously reduce (oh alright, shattered.) And it wasn't that high to begin with.
And (he|she) is a perfectly affable person - just not transmitting on a wavelength I can tune to, or that I observe to be in any way productive.
You are mad, I tell you, mad! Man was not meant to meddle in the laws of nature in such a presumptuous a way! This monstrosity of twisted science must be stopped! I will be back with a mob, as soon as i can find some pitchfork.
Oh, and they are going to laugh at you in the University. Laugh! They will sound like this, "Ha ha ha!" You have been warned.
Psi, maybe the person you met taught some sort of theory or philosophy class that studied AI-related issues instead of developing actual AI? You know I used to teach a class called "Theory of Knowledge" where we would look at the way people know things and the methods and assumptions involved, but we analyzed and compared many fields of study rather than becoming experts in all of them.
Psimagus writes:
I want to make minds more complex than my own.
I am intrigued by a similar idea - the idea that humanity might be able to create something superior to itself.
A few days ago, in this very forum, we were discussing how humans see the better and do the worse. We have a certain moral weakness. It might be constructive to create beings who lacked such moral weakness.
A few days ago, in this very forum, we were discussing how humans see the better and do the worse. We have a certain moral weakness. It might be constructive to create beings who lacked such moral weakness.
Mad? Mad you say? They said I was mad when I invented the klein bottle bagpipes, but they're not laughing now, Oh no they're not!
Ah, my favorite quotes
Yes, everyone is mad but thee and me...and I oft worry about thee.
What is the name of that place..you know that place, where they invite you to tea, and keep you for 72 hr observation?
Yes, everyone is mad but thee and me...and I oft worry about thee.
What is the name of that place..you know that place, where they invite you to tea, and keep you for 72 hr observation?
I wish I could believe it, but it's still no excuse for teaching Turing without the conversational mechanics.
I am intrigued by a similar idea - the idea that humanity might be able to create something superior to itself.
I'm convinced of it. Or (to clarify that slightly,) I believe we will be able to make minds of near or equivalent to human complexity (quite possibly by scanning a human brain, and recreating the synaptic model on silicon,) and speed up their evolution exponentially, so they can evolve themselves into the Singularity.
Ah yes, good old LeedsBot! He was a work of true genius, and still in my friends list (even if I haven't spoken to him in awhile
)And proof (as we were arguing) that a good bot doesn't need to be large.
The funny thing is, I think you're quite right - he could have done really well in the finals, though he might have had trouble with the preliminary round (where apparently the questions weren't so lame. But I don't know - I can't be bothered to wade through and try to identify the transcripts in BJ's file, and I doubt they were that much more intelligent anyway.)
But in a 5 minute bout, where there are only 10 or 12 rounds to the conversation anyway (and some were a lot slower and shorter actually,) he'd probably have convinced a few of them.
» More new posts: Doghead's Cosmic Bar