[List All Topics] [Start Topic] [Search] [Show Thread Index] [Help]

[Not Logged In / Login / Register]

Sort: [Threaded | Date]

		Evolutionary Algorithm igniting the singularity before 2024? posted on 01/06/2004 7:03 AM by radmail	[Top] [Show Index] [Reply to this post]
	Ok the question is: can an evolutionary algorithm combined with a lot of computer power bring about the singularity on its own? Its an open question to anyone who is interested. The consequences are profound when you consider that if the answer to the question is yes, then the singularity (as described by RAY- the J curve) could be here within 20 years (unlike the 40 year prediction by Ray)… This is the observation of Thomas as I understand it. I will include the origin of this debate here: http://www.kurzweilai.net/mindx/frame.html?main=sh ow_thread.php?rootID%3D22688%23id22748 This is Toms view: http://sl4.org/archive/0401/7523.html Toms homepage: http://www.critticall.com Now I will include a section from Rays book ‘The Age Of Spiritual Machines’ (unfortunately the full length version is not available on the net for free so if you have a copy p281-297) Ray is referring to creating intelligent artificial brain which I guess will spark off other paradigm shifts thus leading to the singularity: “As I discussed earlier, an evolutionary algorithm involves a simulated environment in which simulated software “creatures” compete for survival and the right to reproduce. Each software creature represents a possible solution to a problem encoded in its digital DNA. The creatures allowed to survive and reproduce into next generation are the ones that do a better job of solving the problem. Evolutionary algorithms are considered to be part of a class of emergent methods because the solutions emerge gradually and usually cannot be predicted by the designers of the system. Evolutionary algorithms are particularly powerful when they are combined with other paradigms. Here is a unique way of combining all of our intelligent paradigms.”…. Combining All Three Paradigms “The human genome contains three billion rungs of base pairs, which equals six billion bits of data. With a little data compression, your genetic code will fit on a single CD-ROM. You can store your whole family on a DVD (digital video disc). But your brain has 100 trillion ‘wires’, which would require about 3,000 trillion bits to represent. How did the mere 12 billion bits of data in your chromosomes (with contemporary estimates indicating that only 3 percent of that is active) designate the wiring of your brain, which constitutes about a quarter million times more information? Obviously the genetic code does not specify the exact wiring. I said earlier that we can wire a neural net randomly and obtain satisfactory results. That’s true, but there is a better way to do it, and that is to use evolution. I am not referring to the billions of years of evolution that produced the human brain. I am referring to the months of evolution that go on during gestation and early childhood. Early in our lives. Our interneuronal connections are engaged in a fight for survival. Those that make better sense of the world survive. By late childhood, these connections become relatively fixed, which is why it is worthwhile exposing babies and young children to a stimulating environment. Otherwise, this evolutionary process runs out of real-world chaos from which to draw inspiration. We can do the same thing with our synthetic neural nets: use an evolutionary algorithm to determine the optimal wiring. This is exactly what the Kyoto Advanced Telecommunications Research Lab’s ambitious brain-building project is doing. Now here’s how you can intelligently solve a challenging problem using all three paradigms. First, carefully state your problem. This is actually the hardest step. Most people try to solve problems without bothering to understand what the problem is at all about. Next, analyze the logical contours of your problem recursively by searching through as many combinations of elements (for example, moves in a game, steps in a solution) that you and your computer have the patients to sort through. For the terminal leaves of this recursive expansion of possible solutions, evaluate them with a neural net.. For the optimal topology of your neural net, determine this using an evolutionary algorithm. And if all of this doesn’t work, then you have a difficult problem, indeed.” Therefore Rays solution to creating an intelligent machine is the combination of three paradigms: 1. Recursive algorithm (pick best next step) 2. Neural Net (using mathematical models of human neurons) Many variations are possible and the designer of the system needs to provide certain critical parameters and methods such as defining input and topology. 3. Evolutionary Algorithm.(genetic code) Ok my knowledge of the details is limited… so over to you guys!! [Reply]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 01/06/2004 8:45 AM by radmail	[Top] [Show Index] [Reply to this post]
	also: http://www.kurzweilai.net/mindx/show_thread.php?ro otID=22763#id22763 [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 01/06/2004 1:52 PM by jontait	[Top] [Show Index] [Reply to this post]
	I think one of the largest hurdles for any evolutionary strategy is parallelism. The fastest way for an evolutionary process to happen is to process each generation in parallel, but what is the best way to do that? Parallel processing brings home PC's to a crawl, even supercomputers have a difficult time with it (3d-rendering a flame, fluid dynamics, cellular automata, etc). [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 01/06/2004 3:22 PM by Tomaz_(Thomas)_Kristan	[Top] [Show Index] [Reply to this post]
	It is an advantage, when a single computer can independently work on one evolution's branch. If it comes to some new very good results, then it is nice if it's able to send them to others. The evolution may continue from there - on all of them. This way, we do not need more, than a narrow bandwidth connection. And a lot of CPUs, of course. [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 01/20/2004 6:19 PM by /:setAI	[Top] [Show Index] [Reply to this post]
	analog computing is terrific for complex parallelism- just a coffee can sized container of the Israeli DNA-computation gel if properly configured could handle all the processing in REALTIME- [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 01/06/2004 3:13 PM by Tomaz_(Thomas)_Kristan	[Top] [Show Index] [Reply to this post]
	Thank you radmail, for this very welcomed clarification. [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 01/20/2004 5:16 PM by Tomaz_(Thomas)_Kristan	[Top] [Show Index] [Reply to this post]
	Another step ... http://www.critticall.com/sortmaker.html [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 01/20/2004 8:11 PM by TwinBeam	[Top] [Show Index] [Reply to this post]
	A limitation of EA is that something has to define "the right answer". So if we train an EA to get better and better at science, for example, it'd go along pretty fast as long as we can tell it when it's worked out the right answer. But when it gets smart enough to be in the areas we're not so sure about, it'll have to start doing experiments - which take real time. So evolution would slow down at that point. We might be able to evolve a very broad intelligence - equal to or surpassing the best human mind in ALL areas (math, creativity, science, intuition, spatial reasoning, etc) - but not very far beyond the best human mind in any one area. It could still continue getting better - but the rate of improvement would be way lower if it relies on EA. Of course, it might be smart enough to work out a better approach, by that time! [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 01/21/2004 3:22 PM by zurk@arbornet.org	[Top] [Show Index] [Reply to this post]
	i think you've discovered the main hurdle to AI. evolutionary, genetic, memetic, heuristic and all other approaches are basically "smart searches" of an infinite solution heap and nothing more. they all rely on a pre-stated fitness function to find the answer to a problem. guess what ... there is no fitness function for most problems humans solve. they derive the fitness function on their own. something that no computer can do currently (or IMHO not for a long long time, if ever). thats the main reason that anyone claiming to have written (or in the process of writing) an AI/synthetic mind program is spouting 100% grade-A bullshit. the best anyone has managed to do over the last 60 years is a smart search function. If anyone DOES manage to find an algorithm to derive an unknown fitness function given a generic problem and run a search to find a valid solution from the infinite search space, they'd win the nobel prize. [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 01/22/2004 2:24 AM by Tomaz_(Thomas)_Kristan	[Top] [Show Index] [Reply to this post]
	It can guess the fitness function. Pick a random one and evolve everything by this random chosen fitness function. And maybe evolve this function as well, by some other. At least that can be done and the results could be amazing. As they were, when the target was to survive in the environment of the planet Earth. Intelligence was a byproduct. Once we have a simulation of legos, we can drive the sculpture evolution toward several different goals. Like the highest, with the lowest potential energy in gravity field, the most dynamic using motors ... At first this would appear as a crawling in the dark, later it would resemble some incredible machines, later ... who knows! [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 02/10/2004 7:14 PM by zurk@arbornet.org	[Top] [Show Index] [Reply to this post]
	yes it can "guess" the fitness function .. but guess what ? it can "guess" the answer too! voila! now you've taken a simple problem with a limited search set and turned it into a problem of infinite complexity. brilliant! now we can spend [a] an infinite amount of time "guessing" our answer from an infinite search set or [b] develop an infinitely fast processor, thereby solving the problem by expending an infinite amount of energy....causing the heat death of the universe. [insert ironic statement about lack of progress in AI for the last 60 years here] [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 02/11/2004 2:28 AM by Tomaz_(Thomas)_Kristan	[Top] [Show Index] [Reply to this post]
	You are assuming here, that there is an infinite amount of functions, and only one of them yields to a meaningful result. In fact there is not that many short potential fitness functions, at all. And already some of them are quite promising. Like "the shortest", the "least memory consuming", the "fastest" and so on. In our natural environment -- "the most prolific one" -- is a simple function yielded locomotion, sight, flying through the air, intelligence ... There is no infinite amount of computing needed, to get somewhere. Quite modest, in fact. [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 02/10/2004 6:33 AM by Tomaz_(Thomas)_Kristan	[Top] [Show Index] [Reply to this post]
	I have two new subsites. http://www.critticall.com/abstract.html And even more this one: http://www.critticall.com/example.html Dijkstra 1976 and then Critticall in 2004. ;-) [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 02/10/2004 8:24 AM by radmail	[Top] [Show Index] [Reply to this post]
	working hard i see :) [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 02/10/2004 2:21 PM by Tomaz_(Thomas)_Kristan	[Top] [Show Index] [Reply to this post]
	Funny! The general idea is to turn over the work to the computer program, but the amount of work to be done, increases anyway. A repository of several thousands code segments are waiting to be translated and considered for their optimality. [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 02/26/2004 8:00 PM by Tomaz_(Thomas)_Kristan	[Top] [Show Index] [Reply to this post]
	Have some fun! http://www.critticall.com/ChessCitadel.exe Then try to outsmart it, if you can! A "simple evolution" builds more and more compact positions. To achieve the most connected configuration possible, for any given set of pieces. [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 02/27/2004 8:42 AM by radmail	[Top] [Show Index] [Reply to this post]
	53 out of 64, not bad.. i imagine thats the optimal position. Quite easy with enough time to calculate manually- you want the high scorers like the queen and king at center positions (that score maximum 8) while the weaker pawns (max 2) around the outside and in positions where they can score- hence the diagonal form. [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 02/27/2004 9:21 AM by Tomaz_(Thomas)_Kristan	[Top] [Show Index] [Reply to this post]
	I was not clear enough. It is how many "bonds" are established between pieces. How many "attacks" or "supports" are among them. Not how many fields are under fire. It could be that also, it is just an example, how fast is EA in those kinds of problems, where bruta forca would be much slower. [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 02/27/2004 9:35 AM by radmail	[Top] [Show Index] [Reply to this post]
	EA is fast i will grant it that, though not very flexible without the hard work of inputting rules. Use it to make AI conscious, it shouldnt require a great deal of power [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 02/27/2004 9:48 AM by Tomaz_(Thomas)_Kristan	[Top] [Show Index] [Reply to this post]
	Setting fitness rules by the EA itself, is not far away in the future. "Runaway computing under control", is the most likely outcome, I think. [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 02/27/2004 9:58 AM by radmail	[Top] [Show Index] [Reply to this post]
	Yes this i think would make ALL the difference. When the computer begins to set its own limits and keeps pushing these out then the slow cumbersome human can be demoted to supervisory position [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 02/27/2004 10:28 AM by Tomaz_(Thomas)_Kristan	[Top] [Show Index] [Reply to this post]
	Yes, it is a comfortable possition. I think, I'll (we'll) enjoy it! [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 02/27/2004 10:57 AM by radmail	[Top] [Show Index] [Reply to this post]
	Too-shay! [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 05/21/2004 4:57 PM by Tomaz_(Thomas)_Kristan	[Top] [Show Index] [Reply to this post]
	http://www.sciencedaily.com/encyclopedia/several_u nique_sort [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 05/23/2004 3:38 PM by radmail	[Top] [Show Index] [Reply to this post]
	Tom, can't you get your bubble sort thing to sort out the crapy chat bots. When ever i test Ramona she/it will ask a question like have you seen any movies recently, and when you reply 'jksdhskjhdsjkfkjf' and she will say something like 'was it interesting?' and then you reply 'jhdsgjhdsfhdsfhd' and she says 'sounds great' and it goes on and on and on and on... Theres clearly no understanding. cant you right a fitness function thing like for responses to survive if we think they pass the test... that kind of thing... maybe the moderators should include a box where we think its sounding convincing. Though this proceedure is unlikely to make it understand us, rather, fool us into thinking it understands. [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 05/23/2004 3:39 PM by radmail	[Top] [Show Index] [Reply to this post]
	write [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 05/25/2004 10:11 AM by Tomaz_(Thomas)_Kristan	[Top] [Show Index] [Reply to this post]
	I have no doubts, it could be done. From a random to a perfect conversationalist. What is necessary to evolve, is more than a simple chat skill. A whole world view would be better. At the beginning, the fitness should be calculated against some previously recoded chats. There is nothing, what couldn't be evolved. I am quite confident about that. [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 05/25/2004 10:51 AM by radmail	[Top] [Show Index] [Reply to this post]
	ok, so the next question is- would you do it? [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 05/25/2004 12:00 PM by Tomaz_(Thomas)_Kristan	[Top] [Show Index] [Reply to this post]
	Currently I am in evolution of algorithms and some heavy scheduling problems. But do tell me what's your idea or send me an email to thomas@critticall.com! [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 05/25/2004 12:57 PM by Ribald	[Top] [Show Index] [Reply to this post]
	I think one of the points missed by most EA enthusiasts is that their approaches tend to create specialists, not generalists. Unfortunately, generalists are what we typically consider "intelligent". The fitness test is the limiting factor. When a fitness test for intelilgence is possible EA has a chance of being more than a parlor trick... A sort routine that is optimized for a specific data set is not the same thing as a sort routine that is better than quicksort (which is a general sort). A chat routine with a fitness test of specific text output will be even less flexible than a Markov approach. What is not good is a chatbot that has a single response for a single input. An algorithm evolving in that direction would not be better would it? [Reply] [Parent]

		Re: Evolutionary Algorithm igniting the singularity before 2024? posted on 05/25/2004 3:51 PM by Tomaz_(Thomas)_Kristan	[Top] [Show Index] [Reply to this post]
	I think one of the points missed by most EA enthusiasts is that their approaches tend to create specialists, not generalists. There is no generalist among sorts. To be the best in all cases, that is. But to sort any file - of course. Take for example Quick sort, Several Unique sort and Bubble sort. They all sort any file you want. They are all quadratic at their own worst case. At some niches are all the best. Quick sort is the clear winner in an average case, but not always. It also needs some additional memory, what other two don't. A completely random file is the case, where Quick sort reigns supreme. In somehow modified version. In practice, we have many types of arrays. For each type, another optimal sort could (shall) be developed. And, of course, a test which shows which one is to be used. Evolved test. Unfortunately, generalists are what we typically consider "intelligent". As I've said. All those three sorts are general in the sense they are all sorting. Several Unique is an algorithm omitted by human intelligence for decades. Nobody saw it. But has its own personality, just like Insertion or Shell sort. :) [Reply] [Parent]