|
Maybe instead of making a new one he'll just shoot himself in the head like what a real nazi would do.
|
# ? Mar 11, 2016 19:25 |
|
|
# ? May 9, 2024 04:37 |
|
Parallel Paraplegic posted:There was a neat story on NPR about some waves that travel across your brain while you sleep that slightly reduce the "connectedness" of neurons, so that stuff you practiced and learned really well remains but all the little bullshit of the day basically fades into the noise floor, that seems like a pretty useful sleep thing. Ooh, interesting. Is that anything like how ANNs use random dropouts to prevent overlearning? Literally The Worst posted:he's only a neo nazi on paper He was a neo-Nazi on Youtube. Therefore, Youtube is made of paper. QED.
|
# ? Mar 11, 2016 19:26 |
|
Literally The Worst posted:he's only a neo nazi on paper Huuuuuuuuge Neo Nazi. On PAPER.
|
# ? Mar 11, 2016 20:40 |
|
Zemyla posted:Ooh, interesting. Is that anything like how ANNs use random dropouts to prevent overlearning? This one is just speculating, I haven't read up on the topic, but I think it shouldn't be too far off. Remember what regularization (of which dropout is one form) is for in machine intelligence (ML/AI). The most brain-like ML, Deep Learning, is a process by which a small number of extremely fast, linear, and 100% precise machines looks at a large set of input a few hundred million or so times, with one specific goal. The danger is that these nets will simply manage to store a set of specific details about this input, instead of detecting the underlying broad patterns. Regularization makes it harder to rely on details particular to your data set. E.g., you're trying to categorize pictures according to if they contain a cat. Maybe in the 1 million cat pictures you have, for 80% of them, the top left pixel is brighter than (125, 125, 100) or something like that, and for 80% of your 1 million pictures without cats, it's darker than that. Or maybe it's even less - that somehow, 100.000 of your cat pictures come from a cat lover who somehow has a single broken pixel in their camera that's always all-out black. So if you have one neuron that whenever it sees a black pixel at that spot, it screams CAT CAT CAT CAT CAT, that will actually improve your training accuracy by a lot. But you don't want the net to perform like that, you want it to somehow figure out that cats have 4 legs, a tail, fur, and so on. So you kill the net's ability to rely on details in a deterministic fashion. (This is not the best explanation of regularization, but I hope it suffices.) Now the brain is in many ways in a completely different situation. It's a noisy multi-purpose machine that learns quickly from few instances. In a process called "fast mapping"*, kids can learn the meaning of a word - such as "cat" - after sometimes as little as one exposure. But what it will never even be able to do is relying on the sort of generally irrelevant, but locally informative features I've tried to talk about above. Brains simply aren't good with details. Brains also don't see 1.000.000 static pictures 1.000.000 times. They see 5 cats for a few minutes each, while they're moving, interacting, and, importantly, while a bunch of stuff is going on at the same time. The neurons particularly tasked with figuring out how wide or narrow to draw the net that catches the entities in the world to whom the sound "cat" applies will not only receive labelled instances of cats and labelled instances of non-gets. They'll also be bothered by all the other neurons in the brain trying to do things like - recognize the intents of cats - prevent you from falling over while you're trying to move towards the cat to pet it as you have to cross uneven surfaces - listen to your parents talking in the background to figure out if the word "cake" is being used And the cat neurons will even be, to some extent, recruited to help in all of these tasks, to the extent that in real brains, there probably aren't any neurons that are "just" cat neurons; every neuron is a multitasker. That's a lot of noise. And I'm not even getting into more low-level, physiological noise, such as the fact that the neurochemical environment changes all the time, so that sometimes, the cat neurons' excitability will be low, sometimes high, depending on how awake you are - that's a total nightmare, computationally speaking, but at least it means you usually don't have to bother with actively regularizing. There also is overfitting in the brain - e.g. certain forms of optical illusions or stereotypes. But generally, even considering all the commonalities, the problems are very different. I guess Yud sees it roughly like that when he says in the interview that AI will not be based on brains much - I'm not quite so confident at that, because, for example, we could have figured out the importance of regularization earlier if we had thought about the benefits of noise. But okay. Okay. Sleep spindles. As Parallel Paraplegic mentioned, a possible function of sleep spindles might be to support one of the major conditions of sleep: a decoupling between the cortex and the environment. Generally, the brain is super responsive to everything that is going on, and different brain areas area continuously interacting with each other in extremely complex patterns. In sleep, this stops. The thalamus essentially separates the cortex from the outside world, and the cortex itself decouples internally - information-dense long-range interactions are really reduced. One benefit to this might be that it makes it much easier to, with this much lower noise level, move stuff from, to really really really oversimplify, the brain's general-purpose RAM in the hippocampus to the modality-specific long-term storage in the cortex. (IE., when you memorize something from just instance, that's dependent on your hippocampus capturing it, and the hippocampus can do that regardless of the sensory modality; vision, audition, smell ... But the hippocampus seems to be a suboptimal long-term storage, so in the long run, stuff becomes consolidated in modality-specific areas in the cortex. And that process may depend on reduced crosstalk. And this reduction of crosstalk could be a function of sleep spindles, by keeping the outside world and the inside world separate, while allowing limited interaction within the brain.) Whew. Okay. I hope this wasn't too technical. It certainly was oversimplified and probably in parts inaccurate, but sleep isn't my focus. * A major researcher on the topic of fast mapping is Paul Bloom, who in turn co-authored Steven Pinkers' language evolution papers in the 1990s, that, 2 decades later, resulted in Pinkers' turn towards biodeterminism and his anti-blank slate crusade. E: https://www.youtube.com/watch?v=DleXA5ADG78 this is a talk by Geoffrey Hinton on a similar topic. Cingulate has a new favorite as of 21:23 on Mar 11, 2016 |
# ? Mar 11, 2016 21:19 |
|
DStecks posted:On the actual subject of the dark enlightenment, apparently Aurini's YouTube channel got shut down.
|
# ? Mar 11, 2016 21:33 |
|
I clicked on that link and the images of teacupping alone tells me everything I need to know.
|
# ? Mar 11, 2016 21:45 |
|
Zemyla posted:Ooh, interesting. Is that anything like how ANNs use random dropouts to prevent overlearning? No, in my opinion, but I bet you could get a research grant on the topic anyway. If I were trying to warp my thinking to the point that this sounded plausible I would mumble something about how "both are (probably) regularization processes during which precise detail is lost in order to preserve the usefulness of the general meaning of the learned information", but I would probably have to vomit after I said it. My advisor was really into this whole "biologically inspired" AI fad, and I just never saw the point of it: it doesn't make better intelligent systems and it doesn't teach us anything about neuroscience. It feels a lot like exactly the sort of navel-gazing that AI people are prone to falling into for decades at a time.
|
# ? Mar 11, 2016 22:39 |
|
SolTerrasa posted:No, in my opinion, but I bet you could get a research grant on the topic anyway.
|
# ? Mar 11, 2016 22:59 |
|
Literally The Worst posted:he's only a neo nazi on paper On PA- Who What Now posted:Huuuuuuuuge Neo Nazi. On PAPER. gently caress
|
# ? Mar 11, 2016 23:00 |
|
Cingulate posted:I don't think the point of sleep spindles is regularization. But in the talk I linked to, Hinton says dropout resembles the fact that biological spikes are binary rather than analog and that the purpose of binary spikes is regularization. Haha, me either. But I got really used to twisting things around in my head until totally false comparisons between AI and biological systems sounded remotely plausible. I think at one point I told my advisor that my tweaks to MCTS resembled the mechanism by which babies learned to plan their physical movements. Thank god I took all that poo poo out of my actual papers.
|
# ? Mar 11, 2016 23:17 |
|
Hey SolTerrasa, what about the XML-like language Yud wanted to program FOOMAI in?
|
# ? Mar 12, 2016 00:13 |
|
Cingulate posted:Hey SolTerrasa, what about the XML-like language Yud wanted to program FOOMAI in? A quick google search for "yudkowsky programming language" says it was Flare.
|
# ? Mar 12, 2016 00:20 |
|
Nolanar posted:A quick google search for "yudkowsky programming language" says it was Flare. Oh I remember this horrible thing, he makes up a bunch of stupid words to describe it that make it really hard to read if you go in expecting it to be a, you know, useful programming language.
|
# ? Mar 12, 2016 00:44 |
|
For example:quote:2.10: Flare Level checking You see versions are "levels" and I've already thought of 9 of them and level 9 is actually an AI buddy instead of an interpreter! EDIT: quote:2.11.4: Voice comments Shame Boy has a new favorite as of 00:52 on Mar 12, 2016 |
# ? Mar 12, 2016 00:49 |
|
Parallel Paraplegic posted:Oh I remember this horrible thing, he makes up a bunch of stupid words to describe it that make it really hard to read if you go in expecting it to be a, you know, useful programming language. Parallel Paraplegic posted:For example:
|
# ? Mar 12, 2016 00:52 |
Cingulate posted:It's ... Urbit light? I wonder how the red pill types are taking finding out that they may have been participating in a trans narrative all this time.
|
|
# ? Mar 12, 2016 01:44 |
|
Nessus posted:Sisters, now. Generally by not thinking about it or Death of the Author, I'd expect.
|
# ? Mar 12, 2016 01:47 |
|
Cingulate posted:Hey SolTerrasa, what about the XML-like language Yud wanted to program FOOMAI in? Yeah, here's the Path to Singularity which includes the spec for flare (as much of it as ever existed). http://www.yudkowsky.net/obsolete/plan.html I wrote extensively about this in the mock thread, if you want to go read about it that'd be where. It's really, really something. Notably, two people thought they'd try and donate some time to make it happen but Yudkowsky never delivered a real spec and nothing came of it. Surprise!
|
# ? Mar 12, 2016 03:07 |
|
The Ape of Naples posted:I clicked on that link and the images of teacupping alone tells me everything I need to know. "What does he mean by teacupping, I'll just click here and... oh. Oh, jeez. Way to go, Mulder."
|
# ? Mar 12, 2016 03:13 |
|
I can't help but post the most quotable parts again. This poo poo is amazing.quote:At present, I don't even have a publishable Flare whitepaper. I don't even have a finalized design. I am engaging in the sin of aggravated vaporware because I have been told, and convinced, that the timeline does not make any sense without knowing some of what Flare is and what it does. Please consider all discussion of Flare to have whatever degree of tentativeness and subjunctivity is required for that discussion to be an excusable act of speculation. quote:I don't have a Flare whitepaper available; I could probably get one together in, say, a month or so. Since I don't have a complete whitepaper, I was reluctant to say as much as I've said already. I don't want any crippleware versions coming out and depriving Flare of its incremental benefit. See, it's okay, because he has *notes*. If he ever *wanted* to, he'd just finish up those notes and then he'd be published. The reason he didn't do that is because he doesn't feel like it, he has more pressing demands on his time. quote:XML is to Flare what lists are to LISP, or hashes to Perl. (XML is an industry buzzword that stands for eXtensible Modeling Language; it's a generic data format, somewhere between generalized HTML and simplified SGML.) The effects are far too extensive to go into here, but the most fundamental effect is that XML is easy to extend and annotate, and this property extends into Flare programs and the Flare language itself. Yudkowsky had the ~brilliant~ thought that code should be machine readable. Wouldn't that be handy? (metaprogramming was invented before LISP. Go to school, kids, or else you too may end up reinventing the square wheel)
|
# ? Mar 12, 2016 03:17 |
|
SolTerrasa posted:I can't help but post the most quotable parts again. This poo poo is amazing. Today I learned that XML is just a buzzword and has no standards other than "looks kinda like HTML"
|
# ? Mar 12, 2016 04:17 |
|
The key point is, his idea of a 'language' read far more like what most actual coders would call a 'program.' Almost everything he put forward as big new advances that only his language would do were long-established software patterns (particularly the composite pattern). It was like someone fiddled around with Unity for a while, then decided "there should be a language that works like that," without once realising the loving stupidity that such an idea reveals.
|
# ? Mar 12, 2016 07:52 |
|
Somfin posted:The key point is, his idea of a 'language' read far more like what most actual coders would call a 'program.' Almost everything he put forward as big new advances that only his language would do were long-established software patterns (particularly the composite pattern). It was like someone fiddled around with Unity for a while, then decided "there should be a language that works like that," without once realising the loving stupidity that such an idea reveals. Cingulate has a new favorite as of 12:12 on Mar 12, 2016 |
# ? Mar 12, 2016 12:04 |
|
Urbit and it's constituents (Nock, I think, is the programming / scripting language it's built on?) are supposed to replace social media, and in more grandiose terms, the internet. The whitepaper Moldbug put up is really something. -- Also, what happened at the Chicago Trump rally yesterday? NRx twerps whinging about maoists breaking up things violently, St_Rev characteristically alleging Soros' involvement. If it turns out to be a Trump false flag that would be something.
|
# ? Mar 12, 2016 18:34 |
|
Trump had a rally planned, tons of protesters showed up to object, and he cancelled while claiming the police told him to (only for the police to say no, they did not) and things stayed relatively peaceful with just some shoving and a tiny handful of arrests.
|
# ? Mar 12, 2016 18:40 |
|
The alt-right are saying the opposite, though nothing specific. https://www.twitter.com/paxdickinson/status/708711629588926464 No doubt Trump is enjoying this Fututor Magnus has a new favorite as of 19:07 on Mar 12, 2016 |
# ? Mar 12, 2016 19:05 |
Night10194 posted:Trump had a rally planned, tons of protesters showed up to object, and he cancelled while claiming the police told him to (only for the police to say no, they did not) and things stayed relatively peaceful with just some shoving and a tiny handful of arrests. A guy in a Sanders shirt shoves a Trump dude after a bunch of yelling = liberalism discredited forever But remember, the media is liberally biased!
|
|
# ? Mar 12, 2016 22:31 |
|
Criticizing is purging to these people, so of course mild violence is assassination
|
# ? Mar 12, 2016 22:33 |
|
Nessus posted:Beating up protestors or random minorities = youthful high spirits Several news sites are doing the whole "we totally hate Trump but this is a failing on us liberals's part!" thing, but at the same time planning his assassination by being critical of him.
|
# ? Mar 13, 2016 00:17 |
|
Fututor Magnus posted:Several news sites are doing the whole "we totally hate Trump but this is a failing on us liberals's part!" thing, but at the same time planning his assassination by being critical of him. Don't be too mean or else you might stumble onto a Marxist dialectic!
|
# ? Mar 13, 2016 07:18 |
|
Fututor Magnus posted:The alt-right are saying the opposite, though nothing specific. quote:The media is openly stoking a poisonous atmosphere of hate towards @realDonaldTrump & his supporters with the intent of provoking violence. Yes honey, the media is the one stoking hatred, sure
|
# ? Mar 13, 2016 14:56 |
|
Scott on tumblr going through the usual rounds of criticising straw leftists (no sources but some nonsense from facebook?) for not properly criticising an "assassination attempt" which he agrees wasn't an assassination attempt.
|
# ? Mar 13, 2016 20:10 |
|
As A Leftist who presumably represents all leftists everywhere i would like to say that that guy who shoved that other guy totally doesn't represent us and i don't want Trump to die*, happy? *Because if he dies it would make it a lot harder to show all his supporters how wrongy wrong wrong wrong they are for being so loving wrong about everything
|
# ? Mar 13, 2016 20:16 |
|
Parallel Paraplegic posted:As A Leftist who presumably represents all leftists everywhere i would like to say that that guy who shoved that other guy totally doesn't represent us and i don't want Trump to die*, happy? As a leftist I gotta tell you that you do not represent all leftists. I wouldn't kill Trump, but I'd probably make sure to paralyze him from the neck down and cut his vocal cords.
|
# ? Mar 14, 2016 06:06 |
|
Parallel Paraplegic posted:As A Leftist who presumably represents all leftists everywhere i would like to say that that guy who shoved that other guy totally doesn't represent us and i don't want Trump to die*, happy? The term is 'be martyred' hth
|
# ? Mar 14, 2016 06:20 |
|
I just his rear end behind bars if he continues this charade, or better yet completely browbeaten by circumstances that leave him to show just how worthless he is, pure public humilation, and tossed aside like a day old newspaper.
|
# ? Mar 14, 2016 07:54 |
|
Annointed posted:I just his rear end behind bars if he continues this charade, or better yet completely browbeaten by circumstances that leave him to show just how worthless he is, pure public humilation, and tossed aside like a day old newspaper. but please, after he gets nominated and it's too late to turn back
|
# ? Mar 14, 2016 09:23 |
|
I for one hope Trump continues doing what he does for ever in eternity: unite the left, stoke hatred of the rich, make Fox consider feminism, break up any sense of unity on the US right, unearth deep layers hypocrisy and lead bigots into the light of day, and finally, ensure either a moderate-left Jew or a woman is going to lead the free world into the 3rd decade of the 2nd millennium. I also hope he gets to live a hundred years and, if he desires so, die of a heart attack while having sex with Ivanka or Elena or whatever her name is (the wife, not the daughter). Albeit hopefully in a trailer, after all his wealth has been redistributed. May Allah's blessing light shine upon the man as on any other human soul. And please please please never let anything violent happen to him in the next decade or so.
|
# ? Mar 14, 2016 09:47 |
|
Cingulate posted:I for one hope Trump continues doing what he does for ever in eternity: unite the left, stoke hatred of the rich, make Fox consider feminism, break up any sense of unity on the US right, unearth deep layers hypocrisy and lead bigots into the light of day, and finally, ensure either a moderate-left Jew or a woman is going to lead the free world into the 3rd decade of the 2nd millennium. I also hope he gets to live a hundred years and, if he desires so, die of a heart attack while having sex with Ivanka or Elena or whatever her name is (the wife, not the daughter). Albeit hopefully in a trailer, after all his wealth has been redistributed. May Allah's blessing light shine upon the man as on any other human soul. So in short the clown that tears the right into nothingness? You know Cingulate you're not half bad. I can't even remember the reason why I didn't like you that much.
|
# ? Mar 14, 2016 09:51 |
|
|
# ? May 9, 2024 04:37 |
|
Annointed posted:So in short the clown that tears the right into nothingness?
|
# ? Mar 14, 2016 10:14 |