Telmo et. al., I can offer another aspect to the case Dorian is making. It might help in understanding what is being proposed, which is in the difference between
1) Computer-based AGI (C-GI) and 2) Non-computer based AGI (NC-AGI) In McCorduck, P. (2004). *Machines who think: A personal inquiry into the history and prospects of artificial intelligence*, A.K. Peters Natick, Ma, USA. Pickering, A. (2010). *The cybernetic brain : sketches of another future*. Chicago ; London, University of Chicago Press you will find the work of Grey Walter and W. Ross Ashby. For the sake of argument consider Walters' turtle robot. In 1948, using vacuum tubes, a 'subsumption architecture' machine was made. Yes, decades before Rod Brooks' computer-based version. The turtle robot bumbled about and had a control system, not a computer, running it. So yes there was a kind of computation going on, but it was not that of a computer. It was the regularities intrinsic to the control system architecture (physics) that implemented the 'brain'. Now imagine Walter and Ashby had all the neuroscience, biophysics and materials/chip fabrication faculties we have in 2015. What would Walter and Ashby have built? They would have built an elaborate regulatory/control system brain. Not a computer. And by now, in 2015, what would the field of AGI look like? (A) Pickering's 'another future' may have happened and we might have a world with both (1) and (2) in it, not what we have now, which is only (1). or (B) We might be in a world of (1) C-AGI only, except there exists a literature trail of scientific proof that NC-AGI offers no benefit. And we could cite that literature trail. And we could march on with C-AGI only and know why we do that. Neither of these futures came about. What actually happened is that (2) was simply abandoned when (1) arose and the words computer and 'artificial intelligence' were bound together by a series of events that became the history of computer science. In summary: a) The proposal to do (2) does not entail any denial that computers and computer-brained robots are on a trajectory towards real AGI. b) The proposal to do (2) resurrects something that has already started and is merely stopped. (2) merely restores the abandoned half of what was always a route to AGI. and to add an obvious nuance: c) The human brain is actually, literally an example of (2). Not (1). Maybe that is important to AGI? Shouldn't we explore that? So here Dorian and I have two approaches to NC-AGI in a world that is unwittingly wearing C-AGI glasses. How many others like us are there? As scientists, aren't we obliged to complete the examination of NC-AGI that was started nearly 70 years ago ... to contrast (2) with (1) in a science context, so that we scientifically understand the differences? Or at least make a case why 100% of investment has gone to (1) and nil% to (2)? That case has never been made. It does not exist in the literature. ... enter the IGI or something of that kind. Maybe a 'group of the interested' might write a paper that makes this case. Might that form the basis for moving on from here? regards, Colin Hales On Sun, May 17, 2015 at 3:36 AM, Dorian Aur <[email protected]> wrote: > Telmo, > > (i) Indeed, reverse-engineering the brain *on a digital computer is a > much harder problem* > (ii) Also, consciousness (like NLP) is a particularly "hard problem" only *if > we like to replicate it solely on digital computers * > (iii) Benjamin you are also right "a bunch of empirical data and applies > a measure of statistical significance...with interpretations that are > rather monstrous"...statistical significance does not provide a reeliable > theoretical model, and the effect is little success to understand the > diseased brain and provide reliable therapy. Billions and billions of > dollars wasted in the last six decades, science became "a mob opinion" > http://www.pri.org/stories/2015-04-30/do-we-all-have-alzhemers-completely-wrong-man-says-yes- > > > > While the process of computation in the brain* is essentially a computer > science problem* many computer scientists cannot contribute. And the > Turing approach provides only a "reduced model" . > > > We need to think differently. Why travel with a horse and carriage when we > can build a spaceship? With a different path we can solve both problems. > The solution is to bring both groups (a) and (b) thinkers & doers together, > initially create heterogeneous teams to engineer the hybrid system using > biological building blocks. > > > The younger generation of scientists will understand the issue and > probably will not repeat our mistakes if we can move fast with IGI . Both > problems can be thoroughly solved through creativity, design and > engineering. > > > I strongly feel that this endeavor will be the fastest, less expensive > and most effective path towards AGI and brain therapy. > > > Dorian > > > > On Sat, May 16, 2015 at 6:36 AM, Colin Hales <[email protected]> wrote: > >> Sorry if this is a duplicate posting. Something odd going on with my >> gmail. >> >> ---------------------------------------------------------------------------------------------------- >> >> My analysis of the potential for the IGI is continuing. I have thought >> about board structure, but that is secondary just now. The main point I >> want to make here is how I would see such a thing operate. >> >> The future of AGI has two main threads to it: >> >> 1) Computer-based AGI (C-AGI) >> 2) Non-computer-based AGI (NC-AGI) >> >> The IGI will be the first place ever that does NC-AGI. C-AGI has had 100% >> of all investment and over half a century of activity. This imbalance has >> to stop for the good of the entire AGI program. >> >> So the idea is that NC-AGI, which was always a possibility and is now >> more possible than ever, joins C-AGI as a way towards real AGI, however it >> turns out. I cannot and will not discuss the technical conceptuals >> contrasting C-AGI and NC-AGI. It will be the job of the IGI to articulate >> that. This thread is actually about the formation of an institute that >> might do it. >> >> I offer the following suggestion for the scope of the IGI: >> >> 1) The IGI does actual research and development of NC-AGI. The technical >> mission is to make new kinds of neuromorphic chips that do model-free AGI, >> put them as brains in robots and make a new ecology of NC-AGI-based robot >> critters from insect to H+ level. >> 2) The IGI establishes a double-blind independent AGI test facility that >> _all_ embodied (robotic) AGI solutions, C-AGI and NC-AGI, can use to >> formally test candidates. This has nothing whatever to do with Turing >> tests. It will design the test regime and develop and test the tests. >> 3) The IGI can set about isolating and instigating the practical legal, >> social and regulatory mechanisms to do with having a machine ecology join >> (or not) the natural ecology. >> ========= >> As such, it would be ideal if the IGI could be co-located with a C-AGI >> institute. The two approaches, side-by side, could then work together in >> 2) and 3). With a board that can see the merit in such an institute, and >> the right researchers within it, this could be a serious contender for real >> AGI. At the very least it would correct an imbalance to AGI that has been >> in place for decades. It will champion and give a voice to NC-AGI. >> >> Currently there are, as far as I can tell, two and only two researchers >> in the entire world who can envisage some kind of NC-AGI. >> >> Dr Dorian Aur (Ca, USA) >> Dr Colin Hales.(Melbourne, Australia) >> >> If anyone knows anyone else that might see this potential then I would >> like to be put in touch with them. >> >> That's all I wanted to say at this stage. If I were to be part of this >> initiative, then these are my thoughts. I remain enthusiastic about this >> potential. >> >> Regards, >> >> Colin Hales. >> *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> >> <https://www.listbox.com/member/archive/rss/303/17795807-366cfa2a> | >> Modify <https://www.listbox.com/member/?&;> Your Subscription >> <http://www.listbox.com> >> > > *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> > <https://www.listbox.com/member/archive/rss/303/11721311-20a65d4a> | > Modify > <https://www.listbox.com/member/?&;> > Your Subscription <http://www.listbox.com> > ------------------------------------------- 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
