Re: [agi] The Smushaby of Flatway.
> If it was just a matter of writing the code, then it would have been done > 50 years ago. if proving Fermat's Last theorem was just a matter of doing math, it would have been done 150 years ago ;-p obviously, all hard problems that can be solved have already been solved... ??? --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244&id_secret=123753653-47f84b Powered by Listbox: http://www.listbox.com
Re: [agi] The Smushaby of Flatway.
Logic has not solved AGI because logic is a poor model of the way people think. Neural networks have not solved AGI because you would need about 10^15 bits of memory and 10^16 OPS to simulate a human brain sized network. Genetic algorithms have not solved AGI because the computational requirements are even worse. You would need 10^36 bits just to model all the world's DNA, and even if you could simulate it in real time, it took 3 billion years to produce human intelligence the first time. Probabilistic reasoning addresses only one of the many flaws of first order logic as a model of AGI. Reasoning under uncertainty is fine, but you haven't solved learning by induction, reinforcement learning, complex pattern recognition (e.g. vision), and language. If it was just a matter of writing the code, then it would have been done 50 years ago. -- Matt Mahoney, matmaho...@yahoo.com --- On Wed, 1/7/09, Jim Bromer wrote: > From: Jim Bromer > Subject: [agi] The Smushaby of Flatway. > To: agi@v2.listbox.com > Date: Wednesday, January 7, 2009, 8:23 PM > All of the major AI paradigms, including those that are > capable of > learning, are flat according to my definition. What makes > them flat > is that the method of decision making is > minimally-structured and they > funnel all reasoning through a single narrowly focused > process that > smushes different inputs to produce output that can appear > reasonable > in some cases but is really flat and lacks any structure > for complex > reasoning. > > The classic example is of course logic. Every proposition > can be > described as being either True or False and any collection > of > propositions can be used in the derivation of a conclusion > regardless > of whether the input propositions had any significant > relational > structure that would actually have made it reasonable to > draw the > definitive conclusion that was drawn from them. > > But logic didn't do the trick, so along came neural > networks and > although the decision making is superficially distributed > and can be > thought of as being comprised of a structure of layer-like > stages in > some variations, the methodology of the system is really > just as flat. > Again anything can be dumped into the neural network and a > single > decision making process works on the input through a > minimally-structured reasoning system and output is > produced > regardless of the lack of appropriate relative structure in > it. In > fact, this lack of discernment was seen as a major > breakthrough! > Surprise, neural networks did not work just like the mind > works in > spite of the years and years of hype-work that went into > repeating > this slogan in the 1980's. > > Then came Genetic Algorithms and finally we had a system > that could > truly learn to improve on its previous learning and how did > it do > this? It used another flat reasoning method whereby > combinations of > data components were processed according to one simple > untiring method > that was used over and over again regardless of any > potential to see > input as being structured in more ways than one. Is anyone > else > starting to discern a pattern here? > > Finally we reach the next century to find that the future > of AI has > already arrived and that future is probabilistic reasoning! > And how > is probabilistic reasoning different? Well, it can solve > problems > that logic, neural networks, genetic algorithms > couldn't! And how > does probabilistic reasoning do this? It uses a funnel > minimally-structured method of reasoning whereby any input > can be > smushed together with other disparate input to produce a > conclusion > which is only limited by the human beings who strive to > program it! > > The very allure of minimally-structured reasoning is that > it works > even in some cases where it shouldn't. It's the > hip hooray and bally > hoo of the smushababies of Flatway. > > Jim Bromer --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244&id_secret=123753653-47f84b Powered by Listbox: http://www.listbox.com
[agi] The Smushaby of Flatway.
All of the major AI paradigms, including those that are capable of learning, are flat according to my definition. What makes them flat is that the method of decision making is minimally-structured and they funnel all reasoning through a single narrowly focused process that smushes different inputs to produce output that can appear reasonable in some cases but is really flat and lacks any structure for complex reasoning. The classic example is of course logic. Every proposition can be described as being either True or False and any collection of propositions can be used in the derivation of a conclusion regardless of whether the input propositions had any significant relational structure that would actually have made it reasonable to draw the definitive conclusion that was drawn from them. But logic didn't do the trick, so along came neural networks and although the decision making is superficially distributed and can be thought of as being comprised of a structure of layer-like stages in some variations, the methodology of the system is really just as flat. Again anything can be dumped into the neural network and a single decision making process works on the input through a minimally-structured reasoning system and output is produced regardless of the lack of appropriate relative structure in it. In fact, this lack of discernment was seen as a major breakthrough! Surprise, neural networks did not work just like the mind works in spite of the years and years of hype-work that went into repeating this slogan in the 1980's. Then came Genetic Algorithms and finally we had a system that could truly learn to improve on its previous learning and how did it do this? It used another flat reasoning method whereby combinations of data components were processed according to one simple untiring method that was used over and over again regardless of any potential to see input as being structured in more ways than one. Is anyone else starting to discern a pattern here? Finally we reach the next century to find that the future of AI has already arrived and that future is probabilistic reasoning! And how is probabilistic reasoning different? Well, it can solve problems that logic, neural networks, genetic algorithms couldn't! And how does probabilistic reasoning do this? It uses a funnel minimally-structured method of reasoning whereby any input can be smushed together with other disparate input to produce a conclusion which is only limited by the human beings who strive to program it! The very allure of minimally-structured reasoning is that it works even in some cases where it shouldn't. It's the hip hooray and bally hoo of the smushababies of Flatway. Jim Bromer --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244&id_secret=123753653-47f84b Powered by Listbox: http://www.listbox.com
Re: [agi] Epineuronal programming
Steve, Dp/dt methods do not fundamentally change the space of possible models (if your initial mathematical claim of equivalence is true). What I am saying is that that model space is *far* too small. Perhaps you know some grammar theory? Markov models are not even as expressive as regular grammars. Hidden markov models are. But there is a long way to go from there, since that is just the first level of the hierarchy. > By "Markov" you are referring to successive computation processes, e.g. > layers of neurons, each feeding the next? For sequential data, an Nth-order markov model is a model that predicts the next item in the sequence from the last N items. These can be built by making an n-dimensional table, and running through the data to count what item appears after each occurrence of each n-item subsequence. Equivalently, an nth-order markov model might store the probability (/frequency) of each possible sequence of length N+1; in that case we've got to do some extra calculations to get predictions out of the model, but mathematically speaking, we've got the same information in our hands. Markov models can be extended to spatial data by counting the probabilities of (all possible) squares of some fixed size. (Circles would work fine too.) >> Markov models are highly prone to overmatching the >> dataset when they become high-order. > > > Only because the principal components haven't been accurately sorted out by > dp/dt methods? The reason that overmatching becomes a problem is that the size of the table grows exponentially with N. There is simply not enough data to fill the table properly. Let's see... where normal methods would give a variable values 1 or 0, derivatives would allow 1, 0, and -1 (positive change, no change, negative change). So for discrete data, dp/dt will actually make the tables bigger. This could improve discrimination for low-order models (similar to the effect if increasing the order), but it will make overmatching worse for higher-order models (again, similar to the effect of increasing the order). of course, there is an added bonus if the data's regularity really is represented better by the derivatives. Come to think of it, it shouldn't be surprising that working in derivative space is like increasing the order... each unit of derivative-data represents (the difference between) two units of normal data. --Abram On Wed, Jan 7, 2009 at 1:40 PM, Steve Richfield wrote: > Abram, > > On 1/6/09, Abram Demski wrote: >> >> Well, I *still* think you are wasting your time with "flat" >> (propositional) learning. > > > I'm not at all sure that I understand what you are saying here, so some > elaboration is probably in order. >> >> I'm not saying there isn't still progress to >> be made in this area, but I just don't see it as an area where >> progress is critical. > > > My guess is that the poor performance of non dp/dt methods is depressing, so > everyone wants to look elsewhere. Damn that yellow stuff, I'm looking for > SILVER. My hope/expectation is that this field can be supercharged with > dp/dt methods. >> >> The main thing that we can do with propositional >> models when we're dealing with relational data is construct >> markov-models. > > > By "Markov" you are referring to successive computation processes, e.g. > layers of neurons, each feeding the next? >> >> Markov models are highly prone to overmatching the >> dataset when they become high-order. > > > Only because the principal components haven't been accurately sorted out by > dp/dt methods? >> >> So far as I am aware, >> improvements to propositional models mainly improve performance for >> large numbers of variables, since there isn't much to gain with only a >> few variables. > > > Again, hoping that enough redundancy can deal with the overlapping effects > of things that occur together, a problem generally eliminated by dp/dt > methods. >> >> (FYI, I don't have much evidence to back up that >> claim.) > > > When I finally get this all wrung out, I'll move onto using Eddie's NN > platform, that ties into web cams and other complex software or input. Then, > we should have lots of real-world testing. BTW, with really fast learning, > MUCH larger models can be simulated on the same computers. >> >> So, I don't think progress on the propositional front directly >> translates to progress on the relational front, except in cases where >> we have astronomical amounts of data to prevent overmatching. > > > In a sense, dp/dt provides another dimension to sort things out. I am > hoping/expecting that LESS dp/dt data is needed this way than with other > competing methods. >> >> Moreover, we need something more than just markov models! > > > The BIG question is: Can we characterize what is needed? >> >> The transition to hidden-markov-model is not too difficult if we take >> the approach of hierarchical temporal memory; but this is still very >> simplistic. > > > Most, though certainly not all elegant solutions are simple.
Re: [agi] Epineuronal programming
Abram, On 1/6/09, Abram Demski wrote: > > Well, I *still* think you are wasting your time with "flat" > (propositional) learning. I'm not at all sure that I understand what you are saying here, so some elaboration is probably in order. I'm not saying there isn't still progress to > be made in this area, but I just don't see it as an area where > progress is critical. My guess is that the poor performance of non dp/dt methods is depressing, so everyone wants to look elsewhere. Damn that yellow stuff, I'm looking for SILVER. My hope/expectation is that this field can be supercharged with dp/dt methods. The main thing that we can do with propositional > models when we're dealing with relational data is construct > markov-models. By "Markov" you are referring to successive computation processes, e.g. layers of neurons, each feeding the next? Markov models are highly prone to overmatching the > dataset when they become high-order. Only because the principal components haven't been accurately sorted out by dp/dt methods? So far as I am aware, > improvements to propositional models mainly improve performance for > large numbers of variables, since there isn't much to gain with only a > few variables. Again, hoping that enough redundancy can deal with the overlapping effects of things that occur together, a problem generally eliminated by dp/dt methods. (FYI, I don't have much evidence to back up that > claim.) When I finally get this all wrung out, I'll move onto using Eddie's NN platform, that ties into web cams and other complex software or input. Then, we should have lots of real-world testing. BTW, with really fast learning, MUCH larger models can be simulated on the same computers. So, I don't think progress on the propositional front directly > translates to progress on the relational front, except in cases where > we have astronomical amounts of data to prevent overmatching. In a sense, dp/dt provides another dimension to sort things out. I am hoping/expecting that LESS dp/dt data is needed this way than with other competing methods. Moreover, we need something more than just markov models! The BIG question is: Can we characterize what is needed? The transition to hidden-markov-model is not too difficult if we take > the approach of hierarchical temporal memory; but this is still very > simplistic. Most, though certainly not all elegant solutions are simple. Is dp/dt (and corollary methods) "it" or not? THAT is the question. Any thoughts about dealing with this? Here, I am hung up on "this". Rather than respond in excruciating detail with a presumption of "this", I'll make the following simplistic statement to get this process started. Simple learning methods have not worked well for reasons you mentioned above. The question here is whether dp/dt methods blow past those limitations in general, and whether epineuronal methods blow past best in particular. Are we on the same page here? Steve Richfield On Mon, Jan 5, 2009 at 12:42 PM, Steve Richfield > wrote: > > Thanks everyone for helping me "wring out" the whole dp/dt thing. Now for > > the next part of "Steve's Theory..." > > > > If we look at learning as extracting information from a noisy channel, in > > which the S/N ratio is usually <<1, but where the S/N ratio is sometimes > > very high, the WRONG thing to do is to engage in some sort of slow > averaging > > process as present slow-learning processes do. This especially when dp/dt > > based methods can occationally completely separate (in time) the "signal" > > from the "noise". > > > > Instead, it would appear that the best/fastest/cleanest (from an > information > > theory viewpoint) way to extract the "signal" would be to wait for a > > nearly-perfect low-noise opportunity and simply "latch on" to the > "principal > > component" therein. > > > > Of course there will still be some noise present, regardless of how good > the > > opportunity, so some sort of successive refinement process using future > > "opportunities" could further trim NN synapses, edit AGI terms, etc. In > > short, I see that TWO entirely different learning mechanisms are needed, > one > > to initially latch onto an approximate principal component, and a second > to > > refine that component. > > > > Processes like this have their obvious hazards, like initially failing to > > incorporate a critical synapse/term, and in the process dooming their > > functionality regardless of refinement. Neurons, principal components, > > equations, etc., that turn out to be worthless, or which are "refined" > into > > nothingness, would simply trigger another epineuronal reprogramming to > yet > > another principal component, when a lack of lateral inhibition or other > > AGI-equivalent process detects that something is happening that nothing > else > > recognizes. > > > > In short, I am proposing abandoning the sorts of slow learning processes > > typical of machine learning, except for use in gradual refine
