[I should preface the following with:: I will only try and give an indication of what I mean for now, not make a proper case].
I’ve listened in the recent threads to more variations on the same, total confusion about both what AGI is, and how it should be tested. Brett was the only one to stress the obvious: Brett: Ideally it would be nice to have a test that determines the "capability of extending further". And this is the tricky one isn't it, It’s not an ideal. It’s the sine qua non. One crucial key to an AGI is “take-off”/generality – an AGI is a system that can automatically, self-evidently pursue not just one but an unlimited range of diverse tasks/activities – as humans and animals can. An essential test of AGI therefore is a test of whether it can take-off – does it have the “magic sauce” ? – can it immediately set about a vast range of tasks *on command* by human users (*without* knowing any of those tasks in advance) ? Neither Ben nor anyone else in AGI is directly addressing the problem of a take-off system – or indeed has a clue – wh. is why you can immediately write off Opencog and other such efforts. They have absolutely nothing to do with AGI/ take-off – wh. is also why Ben et al have always resisted any form of test – they always have and always will fail any test of take-off/generality. (It’s not just me BTW – many have remarked that Ben et al’s “magic sauce” is not there – not even the idea of one) What could produce take-off and pass an AGI test? As I said, I won’t go into details, but one obvious requirement is a true General Operating System – and the central feature of that is that it will feature a general command system which will automatically generate unlimited diverse courses of actions according to the situation – (and automatically therefore pass an AGI test.) You actually have an extremely primitive and incomplete form of such a General Operating System in Windows and other GUI’s. There you have a graphic system of lines/pointers and icons. And the crucial thing to note is that it’s to an extent a GENERAL graphic system, because the same system and commands can be applied to **any and every** program – an infinity of diverse tasks. The same line or linear movement of a mouse – the same form of commands - can get a computer to open a program that calculates or another that tells stories or one that finds routes or.... and hundreds of thousands of apps ad infinitum. And the same commands can also direct the tasks of all those programs to an extent. We now take these commands so for granted, that we don’t notice how general and brilliant they are. Of course, you won’t understand this system in terms of AGI, if you fail to notice that it works in part by means of a human user. It’s actually a General Robot Operating System - or GROS. Another GOS is our conceptual system - the same commands in our system, such as PICK UP THAT OBJECT – can generate a potentially infinite diverse courses of action, depending on the object and situation - on whether the object is a huge box, a string, a brush, a bicycle, a leaning tower , ... etc etc. Different objects will demand very different kinds of picking up and handling. But the same group of concepts can generate them all. The precise form and principles of a GOS, or GROS (or concepts) - how they will and do work – are an unsolved problem . No existing approach to computation can handle concepts and produce their open-ended, general results. But it should be clear that if you have a true General Operating System (or GROS), then you have an AGI – and that is what we should be aiming for. Once you accept a GOS/GROS as the necessary focus of AGI, then you will no longer have all these irrelevant discussions about whether this or that infant test is a valid test of AGI and the immense confusion of the last two days’ discussion. Nor will you think like Ben & others that cognitive synergy (yoking different progs together) has anything to do with AGI. Nor will you argue that tests are unnecessary or impossible ... or any other of the umpteen other excuses. Self-evidently, a GROS passes any test or definition of AGI. You have an AGI or AGI plan? Show us your GOS/GROS (or plan for one). And nobody can do that – and nobody is yet doing AGI. ------------------------------------------- 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
