Re: An AI program that teaches itself
On 10/25/2017 5:37 AM, Bruno Marchal wrote: I am not entirely sure of this. I think that in the long term, the free-market can work, both for preserving resource and happiness. We might have a different feelings due to the fact that it does not seem to have work with us, but the reason is that we don't have a free-market, given that we have the prohibition laws. Even at the start, Henri Ford, who made his 300 first Ford car in Hemp, and using Hemp, defended the Hemp for building car by saying that it is a renewable resource, and that it would not perturb the current concentration of 0_2, C0_2. If the Market would have been free, most people would have used Hemp (which was the petrol before petrol) instead of petrol. Nonsense. Hemp was grown for rope. It was never a fuel. Henry Ford built a car whose body panels were made from plant cellulose, mostly from soybeans but including 10% hemp. But it was never shown to be economically viable or durable enough to replace steel. Notice that when GM built plastic bodied cars, the Corvette, Saturn, Fiero...they did not make the plastic from soybeans or hemp and the cars have not aged well. The plastic hardens and cracks. To sell something as toxic and disgusting as petrol, you *need* to abolish the free market, which is what happened. After that you do lose happiness, and you do destroy basically everything quickly, hopefully in a reversible way. Free-market is like evolution. It does not see anything in the long term, but can still lead to building things which can see in a longer and longer terms. Bruno I'm afraid you've become a crank on this point...as though marijuana the basis and measure of world capitalism. Brent -- You received this message because you are subscribed to the Google Groups "Everything List" group. To unsubscribe from this group and stop receiving emails from it, send an email to everything-list+unsubscr...@googlegroups.com. To post to this group, send email to everything-list@googlegroups.com. Visit this group at https://groups.google.com/group/everything-list. For more options, visit https://groups.google.com/d/optout.
Re: An AI program that teaches itself
Hi Telmo, With AlphaGo, it is curious that the heuristic half of the system is also the one that becomes a black box. When you have the time, you might elaborate on this. AlphaGo combines search trees and neural networks. The old-school approach to solving turn-based games such as checkers, chess, etc is by using the minimax search tree algorithm. The idea of minimax is simple: suppose you start with a give board state and it's your turn. You consider all of your possible moves, then all possible opponent moves from each move and so on, alternating between players. The goals is to maximize your expected outcome on each of your moves, while minimizing your expected outcome for every opponent move (thus the name). Intuitively, it tries to find the strongest possible play assuming the strongest possible opposition. The problem, of course, is combinatorial explosion. A first approach is alpha-beta prunning. When exploring a branch, once it finds that it already knows a play that is guaranteed to be better than the one being explored it stops (prunes) that branch. Then this can be further improved with heuristics. For example, one can have an heuristic function for chess that assigns a utility value to a board configuration, based on the pieces remaining on each side and their positions. By exploring the most promising branches first, more branches can be pruned earlier. Then it can be made even more aggressive by using the heuristic speculatively, cutting branches even if it's not certain (but just likely) that they will be weaker. One strategy for chess is to go as deeps as time allows, and fallback to heuristic pruning once there is no more time. The more powerful the computer and the more clever the implementation the deeper you can go, and this is how Deep Blue eventually defeated a grandmaster (plus a dictionary of openings and endings from human masters playing the game, chess textbooks, etc). AlphaGo replaces the heuristic function with neural networks: the protocol network and the value network. The value network learns to assign a value to board configuration. Ah! OK. It also uses a stochastic version of minimax, using a Monte Carlo technique. Instead of following all branches, or following them by some heuristically-determined order, it samples them. The sampling is guided by a probability assigned to each future state. The protocol network learns to assign these probabilities. OK. In the first version of AlphaGo, the protocol network was first trained to replicate the actions of human masters. Then, it was further improved by playing against itself. The first stage is supervised learning, the second is reinforcement learning (more similar dopamine-based learning, if you will). The new version was able to do it purely by reinforcement learning, with no reference to human-generated examples. It became a master by exploring the game from scratch. The neural networks used are convolutional networks, usually applied to image recognition. Instead of taking a large number of inputs (the entire board), they scan it. A smaller square starts on the top-left feeding the input of the network, and then it iteratively roams the board. It's a very clever, hybrid combination of AI techniques, combining the strength of old-fashioned symbolic methods with more recent statistical learning. Cool. What I meant is that the search-tree part is purely logic and deterministic, while the neural networks learn to have the right intuition about which board configurations look promising. The search-tree is easy to understand, while the neural network becomes a complex black-box. So the programmers of AlphaGo can inspect its data structures and explain what it hopes to achieve with a given move, but they are not capable of explaining you why the program bet on exploring certain ideas and not others. Thank, I understand now. I find this is akin to the bicameral model of the brain, which I know you like. Here the corpus callosum is simply the piece of code that plugs the output of the neural networks to the Monte Carlo search algo. It seems obvious from the above description that this cannot easily be extrapolated to creating computer programs (or performing self-modification), but it is also clear that this hybrid approach looks promising. I like it very much. There is a big fight in AI between its "tribes" (symbolic, connectionist / statistical, evolutionary). Yes, that fight is part of the process. Tomorrow, the "clever" machines will ask to see the archives of that fight. Their origin. I think that wonderful things will be built by combining all of these ideas, and using their respective strengths where appropriate. What is missing (but has no economical value) is to make such a machine with the only goal being to survive by itself, and multiply. It would be implemented by a reentry of all above into itself (a circular neural nets). Instead of learning games, it
Re: [everythinglist] - A comically knotty inflation, giving rise to our 3 dimension universe?
On 24 Oct 2017, at 04:35, 'Chris de Morsella' via Everything List wrote: Would be a neat explanation for the engine driving the epoch of inflation... also wonder what the implications would be for the multiverse hypothesis that relies upon a mechanism of eternal inflation (leading to an infinity of bubble universes), if inflation is instead an extremely short lived phenomena driven by the latent energy of cosmic knots. All speculative, but nevertheless also thought provoking. I especially like how it would provide a mechanism for why we experience a 3-D + time geometry and not some other number of dimensions. -Chris Why is our universe three dimensional? Cosmic knots could untangle the mystery Why is our universe three dimensional? Cosmic knots could untangle the mystery Next time you’re untangling your earbuds, remember that knots may have played a crucial part in kickstarting our universe, and without them we wouldn’t live in 3D. That’s the strange story pitched by physicists in a new paper, to help plug a few plot Quite interesting. I suspect knots and braids (and Temperley Algebra) are intermediate between number and quantized geometries, and provide indeed the constraints for the "spatial" dimensions, a bit like I suspect the finite simple group to determine the particle symmetries. Programming a quantum topological universal machine/number is essentially encoding information in braids, which I think should be enough for the digital unitary matrices. Note that that site crashed my old computer, but not my small portable fortunately. Some site have no respect for the old machines! Bruno -- You received this message because you are subscribed to the Google Groups "Everything List" group. To unsubscribe from this group and stop receiving emails from it, send an email to everything-list+unsubscr...@googlegroups.com. To post to this group, send email to everything-list@googlegroups.com. Visit this group at https://groups.google.com/group/everything-list. For more options, visit https://groups.google.com/d/optout. http://iridia.ulb.ac.be/~marchal/ -- You received this message because you are subscribed to the Google Groups "Everything List" group. To unsubscribe from this group and stop receiving emails from it, send an email to everything-list+unsubscr...@googlegroups.com. To post to this group, send email to everything-list@googlegroups.com. Visit this group at https://groups.google.com/group/everything-list. For more options, visit https://groups.google.com/d/optout.
Re: An AI program that teaches itself
On 23 Oct 2017, at 15:49, Telmo Menezes wrote: On Sat, Oct 21, 2017 at 3:58 PM, John Clarkwrote: On Sat, Oct 21, 2017 at 12:33 AM, Brent Meeker wrote: The problem is that, like most real problems, improving computer code has no simple one-dimensional measure of "better". Go games are won or lost. A computer program that does the same thing as another but is smaller and executes faster is objectively better ; and although there is no guarantee small fast programs usually have fewer bugs than large slow programs, and the bugs they do have are easier to find and fix. This is not necessarily the case. In engineering practice it is common to use the expression "premature optimization". The idea is: don't try to make programs as fast as you can, because this hurts readability and maintainability. Only optimize for speed when you absolutely must. There are biological equivalents, the idea of "evolution of evolvability". Some species hit local maxima and strongly optimize for a dimension, but this also places them in a dead end. Less optimized solutions might have the property of being more easily evolvable beyond the local maxima. This is why modern scientists use Python instead of C whenever they can. Python is one order of magnitude slower than C. And if you complain that speed size and robustness are 3 dimensions not one then try making the most money. That's the great thing about the Free Market, one dimension rules them all. The above is also a problem with the free market. The free market is incredibly efficient in utilizing resources to spread the maximum amount of gizmos to the maximum amount of people. It is not necessarily optimally efficient in preserving resources for what really matters in the long term, or creating incentives for individual happiness, or anything long-term to be honest. I am not entirely sure of this. I think that in the long term, the free-market can work, both for preserving resource and happiness. We might have a different feelings due to the fact that it does not seem to have work with us, but the reason is that we don't have a free- market, given that we have the prohibition laws. Even at the start, Henri Ford, who made his 300 first Ford car in Hemp, and using Hemp, defended the Hemp for building car by saying that it is a renewable resource, and that it would not perturb the current concentration of 0_2, C0_2. If the Market would have been free, most people would have used Hemp (which was the petrol before petrol) instead of petrol. To sell something as toxic and disgusting as petrol, you *need* to abolish the free market, which is what happened. After that you do lose happiness, and you do destroy basically everything quickly, hopefully in a reversible way. Free-market is like evolution. It does not see anything in the long term, but can still lead to building things which can see in a longer and longer terms. Bruno You seem to love astrophysics -- I do too, but you are surely more knowledgeable. Who pays for the astrophysicists and their equipment? Would the free-market ever do that? Maybe once there's a clear path to profit. Elon Musk is banking on that, but would Elon Musk take the leap without the previous efforts by NASA and other such agencies? I think this is equivalent to the local maxima problem that I allude to above. Best, Telmo. John K Clark -- You received this message because you are subscribed to the Google Groups "Everything List" group. To unsubscribe from this group and stop receiving emails from it, send an email to everything-list+unsubscr...@googlegroups.com. To post to this group, send email to everything- l...@googlegroups.com. Visit this group at https://groups.google.com/group/everything-list. For more options, visit https://groups.google.com/d/optout. -- You received this message because you are subscribed to the Google Groups "Everything List" group. To unsubscribe from this group and stop receiving emails from it, send an email to everything-list+unsubscr...@googlegroups.com. To post to this group, send email to everything-list@googlegroups.com. Visit this group at https://groups.google.com/group/everything-list. For more options, visit https://groups.google.com/d/optout. http://iridia.ulb.ac.be/~marchal/ -- You received this message because you are subscribed to the Google Groups "Everything List" group. To unsubscribe from this group and stop receiving emails from it, send an email to everything-list+unsubscr...@googlegroups.com. To post to this group, send email to everything-list@googlegroups.com. Visit this group at https://groups.google.com/group/everything-list. For more options, visit https://groups.google.com/d/optout.