On Sat, Jan 2, 2021 at 10:32 PM <[email protected]> wrote:
> Ah so I was right, he means making an AI based on real physics [may] set > loose abilities it wouldn't otherwise have if we miss some stuff in the > code. > > That's what you should have said from the start. Very short 1-liner. > > BTW to your statement that a real plane transports you while a simulated > plane doesn't and so same for a virtual / real brain......erm no....even a > real, bloody meaty brain doesn't touch or move our world, unless you stick > some sensors and motors into the meat so it can listen and talk. > It's not expressed this way for a reason. It is more nuanced than a 1-liner. It is not merely a matter of 'missing some stuff in the code', although it's good to see the penny drop. The real point is in realizing there may be aspects of the operation of brain physics that are as impossible to 'code' as 'transport' and 'kidney filtration', 'blood pumping', 'combustion' and 'digestion'. No amount of coding will make up for these endogenous essential physics behaviours. Something endogenous to natural brain physics may be just as critically necessary. Throwing out all the brain's physics by using GP-computers guarantees losing function that you don't know that you don't know. Xchip restores the lost physics in a way that facilitates figuring out what is missing. Exclusively using GP-computers commits you, as a designer, to engineering that redresses the functional shortfall. That shortfall and engineering effort may actually be the fatal flaw that makes the project unviable without anyone knowing what it is. It can be either logically impossible to redress or physically impractical due to energy/compute resources requirements, for example, and you'd never know. The science is impoverished in this way, and has been that way since day 1. Introduction, Page 4: "For example, an artificial (i) kidney is a (ii) dialysis machine including actual filtration physics, not a (iii) GP-computed exploration of abstract filtration physics. And so forth. The natural brain is, prima facie, no different. Some as yet unknown aspect of its nonlinear, thermodynamically far-from-equilibrium complexity-physics may be just as essential to brain function as filtration physics is to kidney function. Nobody can yet formally claim to know what the brain’s essential physics is and what goes missing if it is not retained. To eliminate the (i) fundamental signaling physics of the brain in the context of the Figure 1(e)(iii) use of GP-computers, is to assume that there is no (i) endogenous brain physics essential to brain function. This kind of ‘physics-independence’ (not to be confused with substrate/material independence - see Supplementary 2-2(ii)) is unprecedented in the science of natural phenomena. This formally confirms a potential equivalence between (i) and (iii), in the case of the brain, as unprecedented in science and unique to Figure 1(e) neuroscience." Re your 2nd point: I hadn't thought it needed to be made explicit. I have added this to V3 of the paper: 5.4 Embodiment, embeddedness, Xchip and the final form of the science A final nuance is that Xchip can only be developed and proved as a robotics project. Inorganic ‘physics of embodiment’ (a robot body with a peripheral nervous system) is as mandatory as the claimed essential physics within its Xchip brain (central nervous system). Embeddedness within a test environment is also naturally mandated. The natural brain exists embodied and that body is embedded in an environment. Xchip inherits the same natural situation. Supplementary 2 depicts the developmental embodied/embedded AI test situation containing both an Xchip-based robot and a GP-computer-based robot. These two test subjects are compared/contrasted in a fully normalized neuroscience that includes the Figure 1(e)(ii) empirical component previously missing from the science. Further details will emerge in the future prototyping work. V3 of the paper is submitted awaiting moderator approval on TechRxiv. I think I may be done. As fractious as this process can be, it has actually improved the paper, and for that I am grateful! regards, Colin > *Artificial General Intelligence List <https://agi.topicbox.com/latest>* > / AGI / see discussions <https://agi.topicbox.com/groups/agi> + > participants <https://agi.topicbox.com/groups/agi/members> + delivery > options <https://agi.topicbox.com/groups/agi/subscription> Permalink > <https://agi.topicbox.com/groups/agi/T2f2a092379e757d2-Md89145079433f7a2b2133b48> > ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T2f2a092379e757d2-Mc20e1597b67bf02fa79790a7 Delivery options: https://agi.topicbox.com/groups/agi/subscription
