tag:blogger.com,1999:blog-8587336.post1922134574426872822..comments2024-03-29T06:22:47.638-07:00Comments on CONTRARY BRIN: The (far) Future We Are Fighting ForDavid Brinhttp://www.blogger.com/profile/14465315130418506525noreply@blogger.comBlogger44125tag:blogger.com,1999:blog-8587336.post-517264717660385022008-07-01T06:53:00.000-07:002008-07-01T06:53:00.000-07:00Okay, I teared up at the end of Wall-E, and I thin...Okay, I teared up at the end of Wall-E, and I think this may be Pixar's best work of art to date. <BR/><BR/>Wall-E and EVE (the egg-shaped "love interest") are bit players in a much more interesting story. (And the casting of Fred Willard as the World CEO of Buy-n-Large was genius, in my opinion.)<BR/><BR/>It reminds me of "Rosenkranz and Gildenstern Are Dead", or the "View from the Gallery" episode of Babylon 5, where the main characters are not really the heroes of the story. That belongs to the captain of the AXIOM. <BR/><BR/>I suggest seeing this one in the theaters. <BR/><BR/>But, I don't understand David's reaction to Jumper, which I also watched "free" over NetFlix. I just switched my brain off during the opening credits and had a fine time riding the FX-coaster. When else am I going to see the inside of the Roman Coliseum for so little money, after all?Rob Perkinshttps://www.blogger.com/profile/15618647194288598056noreply@blogger.comtag:blogger.com,1999:blog-8587336.post-6934321315508809922008-06-30T21:17:00.000-07:002008-06-30T21:17:00.000-07:00If you see WALL-E, stay for at least the first hal...If you see WALL-E, stay for at least the first half of the credits.<BR/><BR/>There's a negligible "surprise" at the very end, but it's hardly worth waiting for. But the visuals behind the first part of the credits is a coda of sorts.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8587336.post-16846449948476677712008-06-30T21:00:00.000-07:002008-06-30T21:00:00.000-07:00I have hopes for Wll-E. Otoh... I'm glad I paid n...I have hopes for Wll-E. <BR/><BR/>Otoh... I'm glad I paid nothing to see JUMPER. Finally saw it by Netflix. Gawd what a wretched-evil horrid thing. Poor Alfred Bester. More on this soon.David Brinhttps://www.blogger.com/profile/14465315130418506525noreply@blogger.comtag:blogger.com,1999:blog-8587336.post-23157438686777350102008-06-30T20:00:00.000-07:002008-06-30T20:00:00.000-07:00If Robert Sheckley novel and "The Velveteen Rabbit...If Robert Sheckley novel and "The Velveteen Rabbit" had a love-child, it would probably look like <I>Wall-E.</I><BR/><BR/>That is one seriously amazing movie.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8587336.post-23821815532202863762008-06-30T17:35:00.000-07:002008-06-30T17:35:00.000-07:00@TravcYes, I did know that - and much of the termi...@Travc<BR/><BR/>Yes, I did know that - and much of the terminology is from agriculture!<BR/><BR/>I interviewed for a job at a major medical laboratory testing business to be their guy in charge of teaching stats, and their best and brightest had "never understood beta" so at least you remembered it! It is a killer for a business if you don't understand it, since you can run your test on ten samples, do your stats correctly, and find no reason to conclude there is a difference - while a HUGE difference is there, you just have massive beta error (inability to detect what is there). I have not infrequently averted massive dollars lost due to just this. (It is how stats are taught out of psych departments is the problem.)<BR/><BR/>Actually, our discussion gave me my topic for my monthly article and I take that Wired article to task. I'll post a link here once it is posted if anyone is interested. So thanks! Getting a topic is always the hardest step!<BR/><BR/>Evo algorithms are very cool - I'd like to test how they do with some data examples. Can you point me to references?Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8587336.post-85389155412532551142008-06-30T14:29:00.000-07:002008-06-30T14:29:00.000-07:00Come to think of it, is there any practical use fo...<I>Come to think of it, is there any practical use for a volumetric display?</I><BR/><BR/>For RPG games set in space, would be really useful with star maps and such.JuhnDonnhttps://www.blogger.com/profile/06795417373366495092noreply@blogger.comtag:blogger.com,1999:blog-8587336.post-26442584106572164232008-06-30T14:02:00.000-07:002008-06-30T14:02:00.000-07:00Here is that link in a better form... sorry.<A HREF="http://amygdalagf.blogspot.com/2008/06/if-it-aint-got-that-swing.html" REL="nofollow"> Here </A> is that link in a better form... sorry.Travchttps://www.blogger.com/profile/12790548845692414891noreply@blogger.comtag:blogger.com,1999:blog-8587336.post-7072180638067336122008-06-30T14:01:00.000-07:002008-06-30T14:01:00.000-07:00This is cool:http://amygdalagf.blogspot.com/2008/0...This is cool:<BR/>http://amygdalagf.blogspot.com/2008/06/if-it-aint-got-that-swing.html<BR/><BR/>Willis Conover, who was hugely influential using 'soft power' working for the VOA (hosting a world famous nightly music (jazz) hour) and setting up 'good will' tours of famous jazz musicians... Was a huge HP Lovecraft fan and corresponded with him for years. He also started a Sci-fi fanzine in his youth.<BR/><BR/>It is too hard to properly summarize... just click through.Travchttps://www.blogger.com/profile/12790548845692414891noreply@blogger.comtag:blogger.com,1999:blog-8587336.post-16893290028281019192008-06-30T13:55:00.000-07:002008-06-30T13:55:00.000-07:00SteveO... I'm a biologist (at least for now) and s...SteveO... I'm a biologist (at least for now) and so the quibble over "most" science being explicitly manipulative or not is probably just a matter of perspective. BTW: You do know that lots of basic statistics were invented by biologists (have to deal with that messy data somehow)? ;) <BR/><BR/>Pretty much agree with you. I'd actually forgotten the alpha vs beta error terms, though not the concept. People who don't know that concept (and a lot of other statistics tidbits) should not be using results from data mining!<BR/><BR/>I'm an evolutionary algorithms geek... the circuit design you mention was done on an FPGA, and illustrates a very important point (I've used it as an example in many a talk). Evolutionary algorithms will find solutions for the actual environment you use, not what you think the environment is. If you want to evolve a solution to be robust to variations in the environment, you have to vary the environment it evolves in. Another example was with Karl Sims <A HREF="http://www.karlsims.com/evolved-virtual-creatures.html" REL="nofollow">Evolved Virtual Creatures</A>... when evolving for speed, there was a bug in the physical model simulator and 'creatures' developed which would just quiver at a particular frequency, overflow/whatever some variable in the simulator and *poof* be teleported to a new location. There was another solution evolved which was just a long stick, which would fall over and be counted as very fast (since it movement was measured from center of mass).<BR/><BR/>This problem with evolutionary algorithms is also a bit strength. If there is an aspect of the environment which can be exploited, evolution can find it even if you don't know it exists.<BR/>--<BR/><BR/>@David McCabe: I could certainly use a volumetric display. Currently I'm working on a method which evaluates a function in 4 dimensions... It is a pain in the ass to visualize.Travchttps://www.blogger.com/profile/12790548845692414891noreply@blogger.comtag:blogger.com,1999:blog-8587336.post-45733235852197674592008-06-29T19:38:00.000-07:002008-06-29T19:38:00.000-07:00Back to "Iran War as October Surprise": This New Y...Back to "Iran War as October Surprise": This <A HREF="http://www.newyorker.com/reporting/2008/07/07/080707fa_fact_hersh?currentPage=all" REL="nofollow">New Yorker article</A> goes into quite a bit of detail. The number of sources, especially the "Senior Congressional Leadership" (this can only be one of 8 people) that are giving information on this, combined with a lot of other information that is probably classified somewhere around "Drop Dead Before Reading", in itself is a message that there are a lot of people scared that Cheney and McCain have cooked up a plan to keep the GOP in power.<BR/><BR/>--DaveDave Rickeyhttps://www.blogger.com/profile/02567136316289610947noreply@blogger.comtag:blogger.com,1999:blog-8587336.post-83904598942311217512008-06-28T12:28:00.000-07:002008-06-28T12:28:00.000-07:00The joke is that "sinister" is Latin for "left", a...The joke is that "sinister" is Latin for "left", as in "left-handed". Now go look at that picture again.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8587336.post-26042721893357894952008-06-28T02:05:00.000-07:002008-06-28T02:05:00.000-07:00Congress gets wise, re-funds Fermilab: http://www....Congress gets wise, re-funds Fermilab: http://www.sciam.com/article.cfm?id=fermilab-bailed-out-by-coDavid McCabehttps://www.blogger.com/profile/16603857353437134459noreply@blogger.comtag:blogger.com,1999:blog-8587336.post-60147933093920718362008-06-28T00:21:00.000-07:002008-06-28T00:21:00.000-07:00Anonymous, I don't get it.Come to think of it, is ...Anonymous, I don't get it.<BR/><BR/>Come to think of it, is there any practical use for a volumetric display?David McCabehttps://www.blogger.com/profile/16603857353437134459noreply@blogger.comtag:blogger.com,1999:blog-8587336.post-54389935761636539242008-06-27T23:29:00.000-07:002008-06-27T23:29:00.000-07:00PROOF that there is something sinister about Obama...<A HREF="http://www.scouttufankjian.com/main.php" REL="nofollow">PROOF</A> that there is something <A HREF="http://dictionary.reference.com/browse/sinister" REL="nofollow">sinister</A> about Obama.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8587336.post-23812300037055444822008-06-27T17:14:00.000-07:002008-06-27T17:14:00.000-07:00Travc and Dave, what you are talking about is "seq...Travc and Dave, what you are talking about is "sequential excitation of flourescence."<BR/><BR/>http://stinet.dtic.mil/oai/oai?verb=getRecord&metadataPrefix=html&identifier=AD0780955<BR/><BR/>I worked with those guys at Bettelle labs and it looked like THE way to make REAL 3D displays... except the problem is you don't want a whole lot of BRIGHT pixels in a 3D display. What you want is DARK pixels that block what's behind them. It was great for a 3D movie about ghosts.<BR/><BR/>A similar method was tried (and I was involved) to do TRUE 3D prototype manufacturing. Today's methods are all half-assed 2D ratcheted systems, adding or cutting layers at a time. It shoulda worked. Dang.David Brinhttps://www.blogger.com/profile/14465315130418506525noreply@blogger.comtag:blogger.com,1999:blog-8587336.post-17839603892479008592008-06-27T14:09:00.000-07:002008-06-27T14:09:00.000-07:00Hey Travc,Yeah, I had to watch my post length so I...Hey Travc,<BR/><BR/>Yeah, I had to watch my post length so I had to simplify.<BR/><BR/>For example, astronomy is sure a science but it is non-manipulable (at least at macro scale - it is fascinating to me that cosmology and quantum physics have so much to say about each other). But there are gradations in science and other sources of knowledge. (and this is something I have thought a lot about) Astronomy is another level of science, where you still make hypotheses based on data and continue observations to see if it is confirmed or denied. Less powerful, but still valid - e.g. that smoking causes cancer in people. No one set up a randomized control experiment for that but given the amount of data we now have, we feel pretty confident that the correlation there is real.<BR/><BR/>I would not say "most" science is done this way, but although inefficient it can find answers when manipulation is not possible.<BR/><BR/>I like evolutionary algorithms, but just as in nature, they may or may not home in on an optimum. It is highly dependent on initial conditions and constraints. But there are many situations where that, or other control algorithms, can be useful and indeed really efficient. But again as you say, it is not explanatory, so it doesn't help you in setting up the initial conditions and constraints for the next iteration. And usually evolutionary solutions as set up today are not robust to departures from the chosen environment. (I remember reading about using evo-algorithms to evolve a programmable chip for a specific use. The algorithm found a really great solution, but it only worked on that one chip since it took advantage of a current leak that was accidental - still fascinating though!)<BR/><BR/>I still see explanatory science as more powerful that way, though again data sweeps provide interesting things to look into.<BR/><BR/>The thing about data mining for terrorists is that you are far more likely to make an alpha error (conclude someone is a terrorist) than you are to catch a real terrorist, and/or far more likely to miss a real terrorist (beta error) than catch them for the reason that the number of terrorists is small and even with a stated alpha error of 1% or something, a positive indication is almost certainly an alpha error - unless you reduce the power of detection down to where you won't catch anyone anyway. (Same reason that a single positive AIDS test has a high probability of being a false positive - if you are not in a high-risk population.) This is also why airport security is silly, since you are almost certain to miss a real terrorist and all the signals that you get are almost certainly false positives. So people ignore them and become complacent and have an even higher chance of missing a real signal.<BR/><BR/>By the way, the "keep everything the same and manipulate one variable" is a terrible way to experiment, even though that is what I was taught as an undergrad engineer too. This is about the least efficient way you can find out what is going on, and you cannot assess for multiplicative effects. What you really need is to manipulate multiple variables at once and use ANOVA for analysis. You can even use fractional factorials to vastly reduce the total number of experimental runs (e.g. 16 runs to test five factors at two levels and all two-way interactions, rather than 2^5 runs). If you are interested in learning more, I have a <A HREF="http://www.roi-ally.com/case_polymer.htm" REL="nofollow">boring case study</A> and a <A HREF="http://www.six-sigma-online.com/index.php?option=com_content&view=article&id=28:anova-by-any-other-name&catid=1:heretic-articles&Itemid=17" REL="nofollow">more interesting article</A> on the concepts. Confounding factors are fairly easily handled with "blocked" effects or with other experimental techniques. How's THAT for wonky? :)<BR/><BR/>To sum up, I am a stats geek, but blind search algorithms will not replace the scientific method - though they may be a tool in the process.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8587336.post-53586270841049894842008-06-27T13:49:00.000-07:002008-06-27T13:49:00.000-07:00I was hoping that the intersection of two beams wo...I was hoping that the intersection of two beams would put it over a threshold. But I have no idea how these things work. Perhaps it would be possible to excite it using the harmonic of two interfering beams?David McCabehttps://www.blogger.com/profile/16603857353437134459noreply@blogger.comtag:blogger.com,1999:blog-8587336.post-91512572307445315992008-06-27T13:24:00.000-07:002008-06-27T13:24:00.000-07:00@David McCabe... Cool. Though not quite display m...@David McCabe... Cool. Though not quite display material yet. What we really need is something which changes color only when excited by two different frequencies (3 would work even better). You really need to be able to excite a single point in space.Travchttps://www.blogger.com/profile/12790548845692414891noreply@blogger.comtag:blogger.com,1999:blog-8587336.post-69113397091438026292008-06-27T02:19:00.000-07:002008-06-27T02:19:00.000-07:00This sounds like a volumetric display to me:http:/...This sounds like a volumetric display to me:<BR/><BR/>http://www.thechemblog.com/?p=850David McCabehttps://www.blogger.com/profile/16603857353437134459noreply@blogger.comtag:blogger.com,1999:blog-8587336.post-17483841946538342602008-06-27T02:09:00.000-07:002008-06-27T02:09:00.000-07:00SteveO, you really do sound like an engineer (not ...SteveO, you really do sound like an engineer (not a bad thing IMO, since I have a BS in engineering myself).<BR/><BR/>How scary is it that 'law enforcement' types want to do data-mining? Lets see, people with little or no clue about statistics looking through mountains of data successively testing for one condition after another... oh, and these people have the ability to arrest you (or worse).<BR/><BR/>An interesting tidbit... My boss (population geneticist) actually served as an expert witness *against* DNA fingerprinting in several cases something like 10 years ago. Turns out, the FBI used incredibly faulty methods to estimate the frequency of a pattern's occurrence... so when they said "1 in 1,000,000" match or something like that, it was pretty much BS. Now they have fixed that error, and the (famous) lawyer my boss testified for works with the Innocence Project using DNA fingerprinting to exonerate people.<BR/><BR/>On the other hand, I think you are off-base dismissing non-manipulative analysis as 'not experimental' and therefore not science. A lot of systems really can't be manipulated in the controlled way you imagine. Testing for correlations which are consistent with one theory vs a different theory (and eventually inferring causality when a single theory is successful in a sufficient set of cases)... this is how most science is actually done.<BR/><BR/>Of course, I totally agree with you that there are lots of possible confounding factors that must be accounted for in such a setup... and it is much much less powerful than a 'keep everything else the same and vary only one parameter' idea of an experiment... but since you know statistics, you know that uncontrolled factors can be accounted for with sufficient data (with variation in the proper places).<BR/><BR/>This is a bit of a wonky point, but it is something I have thought a lot about.<BR/><BR/>BTW: You can do much better than neural-networks for most optimization problems. Evolutionary strategies (GAs are just a small subset) tend to work much better at providing robust good (though not guaranteed global optima) solutions. They can also handle much more complex cases with more constraints. <BR/><BR/>One real big problem with using learning systems to design or analyze is the 'black box' nature of the solution... sadly they are just as bad if not worse than raw statistics at providing a causal theory.Travchttps://www.blogger.com/profile/12790548845692414891noreply@blogger.comtag:blogger.com,1999:blog-8587336.post-44533923495235155652008-06-26T18:52:00.000-07:002008-06-26T18:52:00.000-07:00Zorgon: No, I hadn't read the tale on police advoc...Zorgon: No, I hadn't read the tale on police advocating themselves judiciary powers (mea culpa! Fixed now!)<BR/><BR/>It's risky to make generalisations from a case or three, but what I see is a bit of satiation in action. The coppers on the beat felt they needed more powers to administer 'rough justice' to do their jobs. The counter-terrorist unit did not.<BR/><BR/>The point being that that the guys who did not feel the *needed* the extra powers objected to being granted them.<BR/><BR/>(As to the guys on the beat: I can see they have a point, although I don't approve of their 'out of line' antics... which the article rightly portrays in increasing levels of outrageousness: it's a slippery slope down to the gladiatior ring). A better approach is to get a bit of legitimacy into what they do: get verifiable community support for their tactics. Present it to the justice department. Otherwise, they're ultimately no different to the people they are applying these tactics to.)<BR/><BR/>Yeah... it's some future: rapture either way!<BR/><BR/>Speaking of slippery slopes: geo-engineering can have consequences even more dire to the environment than doing nothing: iron sulphide may lead to increased acidification of the oceans (bye bye shellfish)<BR/><BR/>Even the sunshade idea (which is at least reversible) has been shown to have uneven effects: actually warming India!<BR/><BR/>Jubba the hutt in the oval office? <BR/>Bill Solo entombed in culture-war carbonite on the wall. Hil-Leia in chains at his ...er, feet. Selacious W Crumb chained to a desk furiously scrawling pardons for the eructations coming from the various guests feasting around the room.<BR/><BR/>... and a tall shadow in the doorway: '... a young jedi to see you , sire!'<BR/><BR/>Beware of trapdoors (and stretched metaphors)!Tony Fiskhttps://www.blogger.com/profile/14578160528746657971noreply@blogger.comtag:blogger.com,1999:blog-8587336.post-11313202167340877272008-06-26T17:03:00.000-07:002008-06-26T17:03:00.000-07:00Let me get this straight: we're struggling against...Let me get this straight: we're struggling against the forces of obscurantism and superstition and anti-rationalist warmongering so...we can all become cyborgs with collective intelligence travelling in hollowed-out asteroids near the speed of light to the other side of the galaxy?<BR/><BR/>Jeez. I think the average person would look at that and say, "Count me out. I'm voting for McCain!"<BR/><BR/>More good news:<BR/><BR/><A HREF="http://www.huffingtonpost.com/frank-schaeffer/dr-dobson-has-just-handed_b_108989.html" REL="nofollow">Why Dr. Dobson Just Handed Obama Victory</A><BR/><BR/><I>"[T]he new generation of evangelicals is sick of being labeled as backward rednecks because of their association with fossils like Dobson. There are many evangelicals like Cizik too who are not all about homophobia, nationalism, war-without-end and American exceptionalism or the Republican Party. ...They believe that the [sic] America has a responsibility to do something about global warming, poverty, AIDS, human trafficking and other issues. They see through Dobson and the other so-called pro-life leaders, who have actually done nothing to reduce abortion. In fact Dobson has increased abortions because of his "abstinence only" crusade.<BR/><BR/>As a result of his power grabs and bullying of other evangelicals, not to mention his telling people how to vote and pointing them to the failed W, Dobson & Co. have zero credibility with a growing number of otherwise conservative evangelicals who happen--this year--to be looking favorably at Senator Obama's holistic Christian-based world view."</I><BR/><BR/><A HREF="http://www.wired.com/science/planetearth/magazine/16-07/ff_geoengineering" REL="nofollow">Can A Million Tons Of Sulphur Dioxide Save the Planet?</A><BR/><BR/><I>"Geoengineering schemes sound like they're pulled straight from pulp sci-fi novels: Fertilize the oceans with iron in order to sequester carbon dioxide; launch fleets of ships to whip up sea spray and enhance the solar reflectivity of marine stratocumulus clouds; use trillions of tiny spacecraft to form a sunshade a million miles from Earth in perfect solar orbit. They all may seem impractical, but among a small but growing set of climate scientists, one idea that Wood and Teller started pushing in the late 1990s (before Teller's death in 2003) is gaining acceptance: Inject sulfur dioxide into the stratosphere to reflect a portion of the sun's rays back into space, thus cooling the planet.</I><BR/><BR/><A HREF="http://gizmodo.com/5019710/spray-on-skin-gun-shoots-stem-cells-to-heal-your-open-wounds" REL="nofollow">Spray-On Skin Gun Shoots Stem Cells To Heal Your Open Wound</A><BR/><BR/>Not a real device -- yet -- but an example of the kind of thing we're going to see soon that will revolutionize medicine.<BR/><BR/><A HREF="http://www.washingtonpost.com/wp-dyn/content/article/2008/06/25/AR2008062501942.html" REL="nofollow"> The Obamacons Who Worry McCain</A><BR/><BR/><I>""The Republican Party is a dead rotting carcass with a few decrepit old leaders stumbling around like zombies in a horror version of 'Weekend With Bernie,' handcuffed to a corpse."</I> <BR/><BR/>Syd Mead's visualization of Doha, Qatar in the late 21st century.<BR/><BR/><A HREF="http://gadgets.boingboing.net/2008/06/24/syd-mead-dreams-the.html" REL="nofollow">Link.</A><BR/><BR/>But where's Jabba the Hut?<BR/><BR/>Oh, that's right, he's in the Oval Office...Unknownhttps://www.blogger.com/profile/10994509912655287453noreply@blogger.comtag:blogger.com,1999:blog-8587336.post-61274614383897345982008-06-25T16:45:00.000-07:002008-06-25T16:45:00.000-07:00Or, to occam-ize it...Thou shalt understand scienc...Or, to occam-ize it...<BR/><BR/>Thou shalt understand science, and keep it wholly.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8587336.post-2506024030656700172008-06-25T16:43:00.000-07:002008-06-25T16:43:00.000-07:00About the Wired article re: the scientific method....About the Wired article re: the scientific method...<BR/><BR/>This is something about which I can speak with some authority. I am an engineer by training, teach some fairly advanced statistics at the graduate level, and have used them in solving industry problems for over ten years.<BR/><BR/>Blind data mining or data sweeps can find interesting possibilities to test. We have used this method on huge databases of lots of process parameters that no one ever looks at, and have identified interesting relationships. But there is no explanatory power in these models for the same reason that eliminating ice cream sales on the beach will not eliminate death by drowning, though the two are highly correlated. This (correlation is not causation) is mentioned by the article, but then handwaved away without explaining anything.<BR/><BR/>The other issue is more subtle. In any large data set you will find relationships that are unlikely, but given a large enough data set, you expect to find those due to chance. Similarly, when flipping a fair coin, you expect to see runs of heads and tails every once in a while, and should actually get suspicious if you don't.<BR/><BR/>So if I do a huge data sweep and find no relationship, then I think that someone has been monkeying with the data. If I find relationships, I can then test them experimentally. (Note that there is a strict definition of what an experiment really is. Basically, you have to manipulate the system and compare it to the non-manipulated system. Correlation is not an experiment, though it is based on data.) In no case would I take a statistically significant relationship from an after the fact analysis and say that represents something real. I take my hypotheses from the data sweep, select an appropriate sample size, manipulate the process experimentally, and THEN make a conclusion about the relationship. That still sounds like the scientific method to me. (By the way, Popper's overly simplistic representation of the scientific method gives anti-science groups ammunition, but that is another story....)<BR/><BR/>I have also worked with neural nets trying to deduce (in this case) optimum mixtures. They don't really "understand" interactive (multiplicative) effects, so they tend to "average out" these effects and never find optima and settle on OK mixtures that might or might not be robust to variation.<BR/><BR/>So to sum up, in my experience, an empirical model followed by an experiment, leading to a provisional explanatory model is far more useful in real life (e.g. business). The missing element from the article was the experimental aspect - the manipulation of the variables in the system that actually assesses whether your numerical and/or explanatory model has any validity.<BR/><BR/>I just don't see in reality how you can get away from that, or the scientific method. I mean, let's say we meet an alien race that doesn't believe in atoms. Granted it will be tough to convince them to change their paradigm, but the ONLY way I know to even have a chance is through the scientific method.<BR/><BR/>(By the way, my own personal explanation for the Fermi Paradox is that the scientific method is a very rare discovery even amongst otherwise intelligent critters. I think there are probably lots of pre-Roman-era civilizations out there that are stable and will never discover science vs. dogma.)Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8587336.post-60416316639573577872008-06-25T15:21:00.000-07:002008-06-25T15:21:00.000-07:00To comments on the wired article(Provocative Wired...To comments on the wired article(<BR/>Provocative Wired article: The End of Theory: The Data Deluge Makes the Scientific Method Obsolete.)<BR/><BR/>1. Correlation isn't causation.<BR/><BR/>2. You still need models to find correlations/do statistics.<BR/><BR/>3. The examples given in the article still make predictions that can be evaluated. Prediction: pages with high Google rank are contain relevant information for that search term. Prediction: Venter's statistical blips represent a new species.Anonymousnoreply@blogger.com