Ben: "Non-algorithmic programming" is an oxymoron ... unfortunately you (literally) don't really know what you're talking about ...
Perhaps the best way to understand the total limits of an algorithm and a Turing machine – limits about which Ben is completely blinkered – is to think in terms of pianos. A piano can in its own way be thought of as a universal machine if you see the keys as potentially controlling any kind of mechanical action whatsoever – as being able to control not just strings producing sounds, but, alternatively say,/ muscles producing movements, or letters-to-be-printed and so on So let us now visualise what an algo is - it’s essentially this: http://www.google.com/imgres?safe=off&biw=1644&bih=748&tbm=isch&tbnid=OuNVMNmhd1rAKM:&imgrefurl=http://www.jackofalltraining.com/992/history-of-video-games/ibm-punchcard&docid=ylBjcCSwu8c3VM&imgurl=http://www.jackofalltraining.com/wp-content/uploads/2013/01/ibm-punchcard.gif%253Fae39b2&w=583&h=267&ei=xL2PUaPKHce20QWN_4HABQ&zoom=1&iact=rc&dur=585&page=1&tbnh=152&tbnw=332&start=0&ndsp=15&ved=1t:429,r:2,s:0&tx=696&ty=311 it’s a punch card (or a piano roll) for a piano Note for the deeply unimaginative – it doesn’t matter whether an algo is not simply one,but an extremely complex set,of variable punch cards, it’s still basically a punch card And what that means is, if we in turn think of that card as directing the keys of a piano: the algo/punch card plays a SPECIFIC SET OF KEYS in a SPECIFIC ORDER. So it can play only ONE specific key at any one point. Aonly only ONE specific tune over a given period. Look at that punch card and that piano again – there are a WHOLE SET OF OTHER POTENTIAL HOLES AND OTHER POTENTIAL KEYS that could be played, but aren’t being used from A1 to Z30, say. But an algo can’t play ANY KEY at a given point, it can only play ONE - B4 say. - not A2 B3 H4 , , just flipping B4.. And over a given sequence of key presses, it can play ONLY ONE TUNE – just B4 H2 C3 D9 etc. . And if we see an algo as a set of punch cards, well it can play only ONE SET of tunes over time. What’s more it’s always an OLD tune. Or if it’s an evolutionary algo, a mash up of an old tune. That algo isn’t just a punch card, it’s a STRAITJACKET – literally, mechanically a straitjacket. This is the limitation of an algo – and a Turing machine – it can only play ONE tune/set of tunes , only play ONE note at any point. It has a totally limited repertoire. This is the limitation that the field of AGI is supposed to overcome. Now Ben thinks that this is the way things HAVE to be – there simply is no other possibility – the Turing machine – the algo/punch-card-based machine is basically the only kind of machine possible, the only kind of software possible. Punch cards are the only way to press keys. Not surprisingly since Ben has never done, or thought about, anything else besides algos – he’s somewhat conditioned. The only thing a machine/program can do apparently, to put that another way, is we may say, to ITERATE – to repeat the one command it has been given at any point. But that is obviously false – look again at that piano – it has many keys, not just one.. It is both mechanically and computationally possible to IMPROVISE - to play ANY key at a given point, and ANY SEQUENCE of keys over time, to play ANY TUNE. That means, by extension, to play NEW, DIFFERENT KEYS at a given point, to any that have been tried before in a given context. That’s what creative, improvising musicians actually do – *mechanically* – they try many different keys. They “play around” – try other possible keys/notes – try many alternative keys/notes - new notes in context, whose musical effect they have no means of predicting Note, for the imaginatively impaired, this does NOT have to be a random business. It can be constrained by goals, which mean that only keys suitable to those goals are chosen – for example, to be crude, predominantly lower keys, if you’re aiming for sad music. So, let us rephrase that : while it is mechanically possible to play ANY key, it is also *reasonably*/intelligently possible to play not just one but “MULTIPLE” (i.e. “any of many” but not necessarily all keys randomly) How is it computationally possible to do this? Ben apparently hasn’t heard of nondeterministic programming, and the command that allows you to try any of many alternatives – hasn’t heard of true mechanical trial and error – hasn’t realised that machines can indeed improvise. It’s computationally possible, it’s mechanically possible. You don’t have to play just one given note, one given key at a time – or have tunnel vision like Ben – you can play any note, any key. You can play new notes. So now you have seen the basic concept of the Tintner Machine. A Turing Machine is actually only *one* form of computer/robot NOT the only one as unimaginative Ben thinks. It’s a punch card version of a computer, a horribly STRAITJACKETED computer/robot/.piano. It’s a bloody useful machine, as we all know, as long as you just want to play one tune or one set of tunes, the same old set of tunes. But if you want to do something new, creative – something that hasn’t been done before - play new notes (i.,e. new in context of a given piece of music) and new tunes - it’s USELESS. No algo or punch card has ever played a new note, ever taken a single new step. It can only use certain very limited parts of its body, not new parts and only play certain very limited keys. A Tintner Machine is a creative machine – one that is still *programmed* – but programmed nondeterministically – to try any (or any of multiple) reaonsable notes –any action that it is capable of - to make as full use of its body, its piano as possible – – as full use of a computer/robot as possible -and explore new territories whether in the world of music or the world at large (unlike Turing machines and Ben in their virtual prisons). A Tintner Machine – a free machine - is perfectly possible –and is absolutely essential – if we want a machine that can take new courses of action – take new journeys in new fields – and construct new structures. If someone here wants to create a more formal version of a Tintner Machine, parallel but opposed to that of the Turing tape machine, there’s a historic opportunity here.One needs a very visual illustration, similar to the tape, of a machine/computer that can play any note, not just one. P.S. What Ben and others here will do, faced with a radically new idea, is perfectly predictable. They will resort to *logic* and legalistic *definitions* and bog things down in a morass of words, anything to resist thinking outside their paradigm/straitjacket. There is one main thing you need to hold on to here. Is it mechanically and computationally possible to play ANY key on a piano, not just one, at a given point? Yes, of course, it is. And is a machine that can invent an infinity of tunes – an infinity of possible courses of action – somewhat more useful than a machine that can only play one? Yes, of course it is. ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657 Powered by Listbox: http://www.listbox.com
