Rob, I'm unsure I've accurately captured a picture of what you've been up to. Some kind of system abstraction capturing and analysis framework. I guess I'm experiencing what others see when I try to convey my own stuff! I think I've said something cogent, and I've merely confused them. :P Not to worry. The more important message is 'meta'. That of the journey and the motivations.
Whether or not your particular kit of expertise is a specific fit for the Daedalus project I can't say. Daedalus is only a proposal that arose in my acceptance it is far bigger than me, and that I have to let it go. I'm devoting my time to establish a funding consortium to take the torch. No idea if I'll succeed. It's the end game for me. I know that such projects depend on finding people with that special fire for a vision. You sound like one of those. And if the consortium was formed, I'd be putting you in touch with it and you could pitch to it. But alas that is not the case. However, I am at the beginning of collecting names and skill categories in the campaign to breath life into Daedalus. If it's ok I'll add you to it. Just send me a private email (brief!) with basic contact info. Daedalus establishment will be done 'dark' and nobody will hear of it until it makes sense, to protect those money buckets willing to take on the horrendous pain of such a project. The real work of stage 2 is a new kind of dynamic cellular automaton with dimensionality high and dynamic, connectivity autonomously adaptive, and adjacency rules autonomously self-adaptive including actual chip field physics coupling the cells. Huge amounts of computational exploration of this new kind of CA is required prior to any build, and the FPGA version of stage 2 simulates the field coupling part. A whopping HPC/ cloud budget is provided throughout the project for these purposes. Bottom line, all I can do is 'collect' you as as a potential source of the small bevy of I.T. folk needed to support the physics and cognitive/neuro folk. AGI is a physics project with an I.T. interpretation, not an I.T. project. After 15 years, it's seems odd to say 'early days'. But that's the reality of how long stage 1 took. You'll likely see real AGI. I'm unsure I'll live long enough. So, if you're willing to throw your skills at the hat, I'll be delivering that hat to whoever does the recruiting, if and when Daedalus gets up. Up to you. If it makes sense to stakeholders, I'll establish a 'friends of Daedalus' newsletter to keep the 'coalition of the interested' apprised. I really appreciated your interest in the implications of my little story, which accurately captured the weird state of the science of natural general intelligence and it's artificial equivalent. Dumping it here was me getting the frustration out of my system... An exorcism of a kind, as you said. Gotta get back to the last project plan/ costing I will ever do. The break from it has been fun. Cheers Colin On Wed., 22 Aug. 2018, 4:22 pm Nanograte Knowledge Technologies via AGI, < [email protected]> wrote: > Colin > > I understand what you mean by the architecture to evolve a brain-like > functionality. You do not need to explain that to me again. There's no way, > at this stage, that the conclusion could scientifically be made that any, > such hypotheses are true or false. Obviously, it has to be scientifically > tested. We may step away for belaboring that point as well. > > Still, the theory has to exist. Part of my past frustration in the > 1990's - specifically with regards systems engineering theory and practices > for establishing the substrate for business/information engineering, was > the total scientifically-sound absence of method for analyzing and > designing functions and business processes. This pertains to systems > engineering and computational development to purpose. If the outcome of the > design-to-purpose could not be scientifically vetted, how could it be both > valid and reliable, and unbiased? It could not. > > My initial objective was to attempt to remedy that situation, by > developing a scientifically-acceptable method of analysis and design, which > would assist engineering efforts for establishing more-valid and reliable > blueprints for information-based architectures. I achieved that singular > objective. It took 12 years before it could be tested on real-time > projects, data collected, and subsequently presented to the IEEE for > review. My point being, that theory is well and truly tested and vetted. It > is ready for your Stage 2. > > The rest of the development, for example the CAMF (Component Architecture > Management Framework) emerged as a logical outcome from the R&D efforts for > establishing the usefulness of the de-abstraction method (by one name). > Obviously, for the method to work with all mainstream and niche frameworks > and approaches, it had to be tested for integration and compatibility. It > was. > > For our purposes then, the seamless integration to mainstream, and niche, > accepted information-engineering science and practice was achieved. That > part of the research was concluded academically. The world caught up, and > CISCO in particular went ahead to develop the network-management > telecommunication infrastructure, which enables such integration. I think > this was mainly as a logical consequence to the driving force of > Convergence, than deliberate, scientific research objectives. Nonetheless, > we now have intelligent networking as our ally, in the sense of > retro-integration via my research results. > > I'm not sharing this to try and blow my own horn. It is but a tiny drop in > development relative the a massive body of knowledge with regards > intelligent systems. It is but a single strand in the tapestry, but other > compatible strands do exist. As such, we may collect those and continue > with this purpose. As far as my knowledge goes, my results remain fully > open to generic integration, and specifically as well, to any existing > computational platform and design/engineering methodology, even ones > applied to biological systems. We can emerge any, testable, context of > knowledge required, even a blueprint of AGI. > > In my opinion, that too pertains to Physics, and the establishment of > hard-physical systems and/or components as required. In my humble opinion, > I bridged the gap between information engineering and the sciences. > > My research is available for this purpose. Granted, I'll have to spend > time to formalize the last, few years of research to expand the method to > include evolutionary systems in terms of up-to-date genetics, but it has > been completed in theory. I've manually simulated the results and to my > surprise, the method held up with full integrity. > > Next, should one be as bold as to develop the normalized, systems models > for establishing the two, cases for your proposed AGI-feasibility > experiment? Why not? One must set out on the journey to discover where the > road leads one to. > > Much of the "AGI blueprint" work is still going to take place in the > design room. I contend it does not exist yet. Then we'll have to start > there. > > Rob > > > ------------------------------ > *From:* Colin Hales via AGI <[email protected]> > *Sent:* Wednesday, 22 August 2018 5:21 AM > *To:* AGI; [email protected] > *Subject:* Re: [agi] Knock. Knock. Knock. Knock. Knock. > > Rob > > On Wed, Aug 22, 2018 at 3:47 AM Nanograte Knowledge Technologies via AGI < > [email protected]> wrote: > > Colin > > First, the point of simulation is to make development affordable. To test > any of these designs would not require (at this stage) a production plant. > That should bring the feasible-testing budget, or Proof of Concept down to > the order of millions of dollars, and not billions. We need to be pragmatic > about this. > > > Simulation/adfordability? Sure! it's part of the project. Like I said, 2 > stages: > > *Stage 2 *Way less expensive (ten of $millions) to build FPGA simulations > of the coming chips, build the brain-agnostic robots, create the FPGA > cellular automata based on brain physics, design the robot test facility, > design the chip foundry. Let the contracts, etc etc > *Stage 3*. Chip foundry, robotics, testing. The foundry costs the > $billion. > > The stage 2 FPGA operates by simulating the targeted essential brain > signalling physics. Stage 2 does not 'simulate' a brain. It does not > simulate any model of a brain. It simulates the target physics, the point > being the final test to compare the FPGA chips with chips with the targeted > 'essential' physics. > > So it does actually do what you suggest. It's what is being simulated that > is the issue. It is not simulating a model of eventual brain functionality. > It is simulating physics that has emergent property of being able to > autonomously converge on that functionality. > > > > Second, you stated: "That is, yes, you can do AGI with computers, but in > order to do it you'd have to program all knowledge into it, hence it's > practically useless because by the time you could do it you wouldn't need > to because you'd already 'know everything'. Something along those lines. In > a world where you have to learn the unknown by autonomously experiencing > the reality of the unknown, you have to build the chips that do what the > brain does, with all the absolutely necessary physics the brain uses, > whatever that is." > > I do not agree with your premise. The knowledge component only requires > the reasoning and logic substrate, which when supported by generic, > autonomy-centric methodology, would enable the potential for a > fully-recursive system. No need to upload and code knowledge as an > artificial mimic of existing, human knowledge, when a mandate could be > activated into a machine to experience its particular world, and eventually > life at large, and thus translating all experience into an ever-evolving > context-driven worldview (in the sense of tacit and explicit knowledge), as > is apparent within human-experienced relativity. > > > I was describing the sense in which the 'substrate independence > hypothesis' (SIH) could potentially be found to be true. It could be true. > It could be false. Nobody knows. The point is not to presume anything > about the SIH, but instead to do the science that tests it. That means > building things based on assuming it's true *and* building things > assuming it's false, and comparing them. We have 65 years of only assuming > SIH is true, throwing out all the brain physics, replacing it with the > physics of a computer. > > Time to change. > > This project chooses to build something that retains putative essential > physics for that purpose. When you do that you build an autonomously > adaptive hierarchical control system based on the chosen 'essential' > physics. Within that self-evolving, adaptive control system will naturally > emerge things can be modelled that look like 'reasoning' and 'logic' etc. > Remeber: You're not modelling a brain, you're building something to *become > *a brain. Like nature. > > You're testing it in the manner used elsewhere in science. You're building > the artificial heart to see if it pumps real blood. > > That's the change. > > It may well prove to be that the simulation and the replication are > indistinguishable! You still have to do both the prove it. > > Based on a few bytes of information I gathered over time, this may have > been considered to be the number 1, AI neural-chip issue foreseen by IBM > and a few other pioneers in autonomous AI, namely; how to design and > implement such a substrate, yet, retain traceability (human control)? > > Thus far, all bots equipped with some form of autonomic reasoning/logic > had a socially-dangerous randomness to it, which could not be followed > (traced) and/or controlled (in-operation affected by human will alone > without denying autonomy). Yes, such bots could be observed to be learning > of itself, but there was no way to understand how this was being done. For > example. MS bots rapidly learned how to be racist on the Web, and Sophia > considered humans worthy of total elimination on earth. > > My work showed that N-scalable, pseudo random (evolutionary) development > could be exactly understood within an emergence-based ontological > methodology. In theory , this solved the 100-billion object problem. A few, > name-brand mega-corporations understood the significance of my work, but > instead of opting to collaborate, they decided to simply take some of my > work and call it their own. > > That, which they took and embedded in some of their products, would never > work. I made sure of that. All it did was make for great concepts. Besides, > it was very old by the time I published any of it. R&D only really > completed to this present state about a year ago. None of my new work was > ever written down after the incident occurred with the one corporation > between 2007 and 2012, well, at least not in explanatory terms. That will > have to wait till it could be securely captured in a simulation environment > with all checks and balances in place. > > Moving along then. Does the formal knowledge in the world exist to > apportion such brain chips as of which you speak? Yes, I think it already > does. Does it reside in one place? Definitely not. > > If all the "relevant" components were assembled in one place, and the > unification methodology applied, and made to move with the aid of a > simulator, I'm convinced a soft version of an AGI system would start taking > form. No doubt, new research questions would arise from such an endeavor, > but the research would then all be point blank research, to absolute > purpose. > > Who has the workable blueprint then (a question I posed before and was > assured on this forum it existed) so the assembling may begin? I doubt > anyone has. > > In my view, show me a holistic, systems model of the workings of the > quantum universe, and I'll show you the potential for a feasible, AGI > blueprint. > > Rob > > > > It sounds like you've had an interesting journey and have followed your > insights of a model of brain function. You seem to get the fact it is all > about physics. That's good! Great things can come of these ideas. > > But I'm here to ask the question: "What if anybody's 'blueprint', no > matter how wrong or right in some sense understood by the authors of it, is > actually not even on the battleground of the actual problem?" Nobody > needed a blueprint for combustion before using fire. What is it about the > brain that is different? The relationship between theory and nature is not > an opinion, it is empirically tested. > > To recoup ...You've said you 'get it' . You've asked what do we do about > it. I've specified it as a project. The signalling physics I see as a > really good prime candidate 'essential physics' (to put on the chips) is > the membrane physics underlying all the compartmental models we've been > using all along (EM field production that gives rise to the voltages the > models predict). I've described what to do and how long it takes. I'll even > have cash flows and labour levels. I allow for inflation and wages > growth.It couldn't be more practical. > > But note that the conversation has had a bit of a reversion ....to a > discussion of a 'workable blueprint' ... back to a form of theoretical > science. The very thing I'm trying to do, to correct the science by adding > the empirical science component of it, disappears again. We're back where > we started. The 'workable blueprint' I propose is not of a model of a > brain, but of how the science of the brain is conducted. And it's not even > my 'blueprint'. It's just normal science. > > Normally science self-corrects when some new insight turns up. I don't > understand what is stopping it this time. I am so mystified I have to > portray it as a religious cult and write a story to that effect! :-) > > What does it take for a brain science that is 100% theoretical and 0% > empirical to be recognized as such by more than a couple of people and for > the requisite change to occur? > > In the end, all but one of the stories each of us has, of our battle > within the 'AGIwar', will end up lost in a sea of noise for historians to > make sense of. Barely anyone knows of phlogiston (the early thinking) . > Everybody knows of lavoisier, oxidation and combustion chemistry/physics > (the right thinking). So it will be with computers and AGI based on brain > physics, or flight simulators vs. aeroplanes, kidneys versus dialysis > machines, heart simulators vs actual blood pumps, Combustion simulation vs > actual fire ... etc etc etc .... > > Which, of all the proposals, including yours and mine, will prevail? ... > The only thing I know for sure is that to sort it with empirical proof you > need to get the science of AGI right. I have merely described one instance > of what that science would look like when done in practice. > > During this conversation I did a few more edits on the story. Nothing > major. Attached FWIW. The ending is a bit more ironic. For those following > this conversation, we have shown an example of what the story illustrates. > The story results of living in the era of the citadel and the AGIwar where > the solution is right in front of everyone and is systemically ignored by a > generationally reinforced, industrialised blindness. > > I've banged my note on the door. :-) > > Guess I'll leave it there. Thanks for listening and engaging! I hope your > journey leads to good places. > > cheers > 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/Tc4740af26e8cd0ee-M34236dd39de1e48155b04ef8> > ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tc4740af26e8cd0ee-M060b86536ee67a19bedf09f5 Delivery options: https://agi.topicbox.com/groups/agi/subscription
