AH: I most certainly am not avoiding the challenge Here’s the challenge for language – the challenge of its continous new elements/creativity. You’re doing semantic nets? http://en.wikipedia.org/wiki/Semantic_network Take two concepts from the net there: CAT and BEAR. How are you going to produce/understand/reject even a tiny fraction of the ways in wh. they can be related? And how, once you have a set of relationships for them individually, are you going to incorporate new ones - basically the frame problem, no? e.g. THE CAT STOOD ON THE BEAR THE CAT CUDDLED THE BEAR THE CAT KILLED THE BEAR THE CAT PUNCHED THE BEAR THE CAT CARRIED THE BEAR THE CAT SCRATCHED THE BEAR THE CAT WAS BIGGER THAN THE BEAR etc, endlessly.... .. How are your nets, or whatever approach to language, going to be able to deal with such endless word/concept/object combinations – perhaps the most classic example of language’s creativity? Just as there are no creative algos generally, so are there no creative NLP-or-other-computational approaches to language. Any AGI approach to language, I suggest, should start from the problem above – because essentially that is the problem that confronts the infant. [e.g. an infant may have to relate CUP - DOG – DADDY - MUMMY in any of endlessly diverse ways]. You don’t need to create absurdly complicated linguistic structures, as Ben & many others do – you can start with v. simple infant language. But if you can’t crack that problem – and you won’t – then the rest is pointless.
From: [email protected] Sent: Friday, October 19, 2012 3:57 PM To: AGI Subject: Re: [agi] ONE EXAMPLE I most certainly am not avoiding the challenge. The very thing you say I'm avoiding is the same thing you just said I am suicidal for attempting to do. There are *stages* to engineering, programming, or anything else that involves construction of complex artifacts. I am building the tools to work with first: the internal representation of information that I believe most closely parallels what we do in our minds. Once that's in place, it'll be possible to actually define what "thinking" means, and to generate testable hypotheses of how it works. You make demands as if we had already produced an AGI and were touting all it was capable of. We have not, and we all know it's going to take a *lot* of effort, clear & in-depth thinking, and probably time. It seems to me that what you keep harping about is that programs can't think outside their own box. You can see the box, and the program can't, and so you can't see it as creativity, just variation on a theme. But every system has a box. Some are roomier than others, but it's always there. The human brain included. We *feel* like we think outside the box, but that's just because we've never thought of some things *before*. We inherently have those capabilities built in, though, from the very beginning. No one, and no thing, can escape it's nature. Sorry if you think you're an exception. That's just part of your nature. -- Sent from my Palm Pre -------------------------------------------------------------------------------- On Oct 19, 2012 6:17 AM, Mike Tintner <[email protected]> wrote: MA:You've been kind of dancing around all the examples people have been feeding you. There has been no example of genuine generativity. Period. If you or anyone else feels that a given example is not being sufficiently discussed, I will gladly discuss it in depth. Do you personally think that there IS an example? Put it forward . This is the crucial issue of AGI – it should be discussed seriously. On my side, I haven’t yet, but I can gladly give you endless and spectrum-wide examples of how ALL REAL WORLD PROBLEMSOLVING CANNOT BE SOLVED BY ALGORITHMIC MEANS. That’s 90-odd per cent, I would guess, of all human reasoning. You guys similarly to the above, almost never discuss real world problemsolving – just logic, maths and programming problems. You can’t get much more blinkered than that. MA: you never answered my suggestion that an algorithms need only appear to be creative to pass the Turing test Huh? You want to try and fake it? Step back a moment and think of what a twisted, “dancing around” mindset you are embracing – wh. is the kind of twisted thinking that most AGI-ers seem to fall into when confronted by the non-generativity of algorithms. No. Algorithms will never in any shape or form pass any Turing test – i.e. be able to hold a real world conversation. Why? Because real world conversation is creative – it involves a continuous stream of new elements and new,non-formulaic combinations of concepts – and their real world referents. We continually talk about the “news” – from media to professional to personal – and it really is “news” – never before happened quite like that. Algorithms can’t handle the possible new combinations of just *two* or *three* concepts let alone a full-blown conversation. This is one reason why what Aaron & others are doing – trying for a language/conversational AGI – is suicidal. The *first* thing he should be doing is asking: how can I produce/process creative/generative conversation? But he like nearly all others avoids the challenge. From: Mike Archbold Sent: Friday, October 19, 2012 11:41 AM To: AGI Subject: Re: [agi] ONE EXAMPLE On Thu, Oct 18, 2012 at 2:07 AM, Mike Tintner <[email protected]> wrote: You’ve been predictable and produced a lot of personal waffle - but not ONE FUCKING EXAMPLE of a single creative thing – a single new element - that algorithms have ever produced. You've been kind of dancing around all the examples people have been feeding you. I think one thing you ignore is that people are kind of tiresome in their application of algorithms and are far from really having an elusive undefined creative element. eg., if so-and-so is a certain social class and does such-and-such, then I am supposed to do so-and-so or risk some undesirable-outcome....but some new social situation arises, unforseen, now a new rule is needed, so we apply whatever algorithms we seem to have that work.... anyway you never answered my suggestion that an algorithms need only appear to be creative to pass the Turing test. If it appears to be creative then it is creative, meaning only that somebody has not picked up on the "trick." Also, evolution, built into the universe, runs off accidents and survival of the fittest which creates the appearance of being creative. SO you are left with your own subjective assessment only of what does and does not constitute creativity. And how do you know your own assessments of creativity are not themselves bounded by some algorithm? It should strike you as extraordinary that no one can produce one example – nada. ..unless you’re prepared to look at the obvious. The whole of technology so far – esp algorithmic technology – has been about machines that produce routine, predicable, “old” courses of action and products. Algorithms – all zillions of them – have never produced a single new element. You can’t produce one fucking example because there isn’t one. AGI will be a revolution – a whole new epoch of technology - because it will be about machines that can produce NEW courses of action and products – with NEW combinations of elements – and do so endlessly with endless diversity and endless surprises and unpredictability. Like you – only hopefully you will start producing something newer than excuses... From: Aaron Hosford Sent: Thursday, October 18, 2012 5:17 AM To: AGI Subject: Re: [agi] ONE EXAMPLE Funny you should say that when he just said *you're* sticking to a primitive definition of "algorithm". You can't imagine anything *people* do (in particular anyone on this list, since that's a convenient group to pick on) that's new or creative. Maybe it's *your* creativity that's limited, in that your imagination can't follow where our imaginations tread. I can easily imagine a program (not an algorithm, mind you, but a collection of data structures and algorithms interacting with the real world in all its glorious complexity and surprisingness) from which creativity emerges. How did all those fonts come about? The randomness of the real world, interacting with the pattern recognizers and learning mechanisms that live in the human mind. Nobody thought them up from scratch. I'm a musician and songwriter in my spare time, which requires creativity. The worst insult to a song writer is to say that it sounds just like another song (unless all he cares about is getting paid, in which case formulas work great). When I write songs, I start by picking up the guitar, and fiddling around randomly until I hear something interesting come out. Then I reverse engineer what I just accidentally produced, and I start thinking about how to generalize the "feel" of it so I can produce more that goes with it. I try things out, and build onto it, not by thinking ahead, but by stumbling in the right direction and either backtracking if it sucks or holding on to what I've done if it sounds good. This goes on throughout the entire process of writing a song. My experience as a song writer serves as a general guide to determine the direction I'm going in and reduce the number of bad ideas and false starts I have to try out before I stumble onto a good one, but ultimately writing a song comes down to accumulating a lot of awesome mistakes together according to a strict measure of what sounds good to me. This, I suspect, is exactly what other artists and creators go through when they create anything at all. There's nothing particularly hard about implementing any of this in a computer program aside from determining the measure of goodness, which we humans have built in due to evolution. To make it general across multiple domains and not just one, we would have to also build in a way to detect the space of possibilities, such as that for a guitar, there are such-and-such notes, or for a canvas, there are x and y coordinates related to each other by a Euclidean distance metric. This is also do-able, albeit probably a lot more difficult. How much experience do you personally have with creating things, that you can sit in judgment of us and say we don't know what creativity much less how to build it? Are you a musician? An artist? A programmer? A writer? A philosopher? What? On Sun, Oct 14, 2012 at 9:31 AM, Mike Tintner <[email protected]> wrote: Frankly, Jim, definitions are for wankers. The last resort of s.o. who doesn’t want to get anywhere. Your ideas about algorithms’ powers are fictional. There isn’t an algorithm in the world that isn’t mindblowingly limited – that just “builds” Lego houses and no other kind of structure, or “cooks” one set of dishes and nothing else. Take just about any verb you like – “travel”, “fly”, “calculate”, “compute,” “translate,” et al – and an algorithm will only be able to do one hyperspecialised version, compared to the infinite possibilities. Show us something actual and general/creative, with new elements, that algos can do - or please stop wasting air. From: Jim Bromer Sent: Sunday, October 14, 2012 3:13 PM To: AGI Subject: Re: [agi] ONE EXAMPLE Mike, How many times does it take to get this idea across to you. You are confusing a primitive definition of algorithm - which might be currently acceptable to many people - as a fundamental notion of the characterization of a computer program. Jim On Sun, Oct 14, 2012 at 10:02 AM, Mike Tintner <[email protected]> wrote: P.P.P.S. Just to ram this home - UNLESS you do something like I’ve suggested, (and I know none of you have) - a) tackle a proper creative problem (and what better than geometrical/math ones for you?) - (you don’t have to come anywhere near solving it, just have a go at it), and then b) try and algorithmise/systematise your thinking - unless you do that, you will NEVER understand AGI. If you do, note what is the “set of elements”/options to be thought about here? (there never is one). AGI | Archives | Modify Your Subscription AGI | Archives | Modify Your Subscription AGI | Archives | Modify Your Subscription AGI | Archives | Modify Your Subscription AGI | Archives | Modify Your Subscription AGI | Archives | Modify Your Subscription AGI | Archives | Modify Your Subscription AGI | Archives | Modify Your Subscription ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-c97d2393 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-2484a968 Powered by Listbox: http://www.listbox.com
