About OpenCog and "real time" ... OpenCog is being used to control an agent in a virtual world, and during 2013-2014 will be used to control a Hanson robot.... So in that sense it's semi-real-time
But it's not "hard real time" in the sense of, say, the adaptive control based stabilization system of a quadrotor drone... My feeling is that it's OK for AGI to have a system whose bulk is semi-real-time, interacting with specialized hard-real-time components as needed.... If you have an argument otherwise, I'll be happy to listen ;) ben On Tue, Oct 2, 2012 at 7:02 PM, Anastasios Tsiolakidis <[email protected]> wrote: > Thanks PM! The axiom is that an embedded intelligence, even one that can > spend 9 hours in bed and 10 hours in a chair, needs (occasionally at least) a > real time "instant" general intelligence apparatus, and I'd dare say even > writing my reply is proof of that, I've instantly formulated a reply in my > mind but it takes a while to type it out. It is quite a spanner in the works > for mathematical definitions of intelligence where some scary big calculation > happens in a "timeless space". Be advised that this is not even the outline > of an architecture, I simply wanted to highlight the axiom. It is about AGI > with performance "built in". A project worth pursuing if only to contrast the > results with non real time approaches such as OpenCog (perhaps the latter is > loosely real time). The evolutionary motto would be "survival of the > fastest", I can certainly imagine this being the case not only in warfare but > more importantly (!) in artificial life simulations. The Internet is an ALife > battleground and being the first one to hack-and-patch (closing the security > hole behind you so that the next agent will not be able to challenge you for > ownership of a resource) is going to be a very real time affair. > > I have not declared war on long term memory, but I am willing to compromise > it and approximate it, not unlike > http://en.wikipedia.org/wiki/Dataflow_architecture > > I guess this is a explicit as I can get, StreamAI would look a lot like a > huge FPGA with some self reconfiguration, or the simulation thereof. > > AT > I Am > > On 02/10/2012, at 08.35, Ben Goertzel <[email protected]> wrote: > >> On Tue, Oct 2, 2012 at 6:04 AM, Anastasios Tsiolakidis >> <[email protected]> wrote: >>> Allow me to briefly sketch an axiomatic approach to intelligence of >>> the highest order: real time, "stream" intelligence. >> >> Ummm... where are the axioms of your axiomatic approach? >> >> I'm a bit confused by what you're proposing. A system with no long >> term memory? ;) ... be more explicit please!! ... >> >>> As much as it is >>> desirable to spawn processes and logical inferences in Lisp and >>> Prolog-like systems in the pursuit of "better approximations of >>> reality", there is no escaping the many charms and attractions of the >>> more streamlined processes typical of biological nervous systems and >>> electronics: input-to-output in dt, as quickly as possible. >>> >>> StreamAI would be a strong contender in a lot of domains that use >>> intelligent agents, including artificial life, financial decision >>> making and warfare. But there is no reason, and possibly no path, to >>> fall into the narrow-AI trap if the design is from the beginning >>> multimodal and ambitious. There is no reason why a financial >>> streamAgent would not look for clues in the faces on Bloomberg TV in >>> addition to the Dow Jones ticker. One architectural problem that >>> could perhaps be solved in a general way would be how to incorporate >>> lengthier "thought processes" in StreamAI, two obvious candidates >>> would be "inlining" (expanding programs with some degree of >>> parallelism into parallel "hardware") and "externalizing", ie calling >>> an external prolog or whatever and "knowing what to do" while waiting >>> for an answer. I would argue that human chess expertise is first a >>> form of inlining, where a concept like "king safety" has been learned >>> together with "micro-variations", doing instant checks on the ability >>> of certain lines and squares been used for an immediate attack on the >>> king, and then "externalized", passed to the frontal cortex for a more >>> proper wider and deeper analysis tree. >>> >>> If you are thinking there is no reason to handicap your AGI quest by >>> putting all of your eggs in one stream, let me remind you that the >>> ever expanding range of stream processors, whether GPUs, FPGAs or >>> whatever are making StreamAI both possible and necessary. >>> >>> AT >>> >>> >>> ------------------------------------------- >>> AGI >>> Archives: https://www.listbox.com/member/archive/303/=now >>> RSS Feed: https://www.listbox.com/member/archive/rss/303/212726-11ac2389 >>> Modify Your Subscription: https://www.listbox.com/member/?&; >>> Powered by Listbox: http://www.listbox.com >> >> >> >> -- >> Ben Goertzel, PhD >> http://goertzel.org >> >> "My humanity is a constant self-overcoming" -- Friedrich Nietzsche >> >> >> ------------------------------------------- >> AGI >> Archives: https://www.listbox.com/member/archive/303/=now >> RSS Feed: https://www.listbox.com/member/archive/rss/303/14050631-7d925eb1 >> Modify Your Subscription: https://www.listbox.com/member/?&; >> Powered by Listbox: http://www.listbox.com > > > ------------------------------------------- > AGI > Archives: https://www.listbox.com/member/archive/303/=now > RSS Feed: https://www.listbox.com/member/archive/rss/303/212726-11ac2389 > Modify Your Subscription: https://www.listbox.com/member/?&; > Powered by Listbox: http://www.listbox.com -- Ben Goertzel, PhD http://goertzel.org "My humanity is a constant self-overcoming" -- Friedrich Nietzsche ------------------------------------------- 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
