2008-05-16.log

--- Day changed Fri May 16 2008
-!- nsh [n=nsh@87-94-155-173.tampere.customers.dnainternet.fi] has quit [Remote closed the connection]00:16
fennthe reason biotech is expensive is that every last thing is patented multiple times00:40
fennthe only reason it hasn't ground to a halt is that nobody can possibly look at all the patents and figure out who is in violation00:41
kanzurefenn: I just got done reading some of Eli's writings on yudkowsky.net ... why does he get funding?00:41
fennbecause nobody else writes seriously about AI ethics00:42
kanzurebut it's not a matter of ai00:43
kanzurelook, even if you get ai on a box00:43
kanzureit's going to come across the same problems we're fighting with skdb00:43
fennhmm?00:43
kanzurefrankly *we're* ai, just not able to modify our software as quickly of course00:43
kanzureso think about it00:43
fennno, because we arent able to re-engineer our cognitive architecture00:43
kanzureif you have ai running on a computer, it designs its next generation hardware00:43
kanzureassume it's reached upper limits on software00:43
kanzureso it needs a new ISA00:43
kanzureso it goes to make it, great, then it needs a physical implementation00:44
kanzurehow is it going to figure that out ? the same way we would00:44
fennyes00:44
fennso? that's just the first generation00:44
kanzure?00:44
kanzureI'm not saying ai is a bad idea00:44
fennit's hard to say much of anything.. one ai may be wet and mushy, whereas another might be formalized and brittle00:45
kanzurejust that it's going to run into the exact same problems00:45
kanzureit'll need new ways of interfacing with physical reality, just like skdb00:45
fennbut they both run on the same silicon00:45
fennno, it wont run into the exact same problems, that's stupid00:45
kanzureit'll need arms and manipulators and manufacturing equipment, or at least a text display so that it can tell a human what to do00:45
kanzureokay, so then what is it going to do00:45
kanzurejust sit there, compute,00:45
kanzureand suddenly intuit a new architecture out of the divine void00:46
kanzureand suddenly it's magically implemented?00:46
fennmaybe00:46
kanzurewtf?00:46
fennwell, say it reads a few papers and does this DNA FPGA thingie i'm rambling about, and sends them off to get sequenced00:46
kanzurewait, sequenced?00:46
fenntotal human interaction: zero so far00:46
fennun-sequenced00:47
kanzureright, okay00:47
kanzureokay, so you're thinking more about a DNA-FPGA ai there00:47
kanzureand that's implemented on bio00:47
kanzurewhich is already an exponential process00:47
fennso then it just has to convince some human to shake up a couple test tubes of stuff and pour it into a beaker with wires dangling in00:47
kanzureso that's the exponential growth hijack scenario00:47
fennthat's just cognitive architecture number 200:48
fennwell, computational architecture really00:48
kanzureyeah, but that's the hijacking scenario, what about the bootstrap scenario00:48
fennbut i assume the AI is smart enough to take full advantage of FPGA's (whereas humans suck at this)00:48
kanzurethe hijack scenario is iffy, it's like GNU Hurd except on steroids - CSAIL has been beating their head against the amorphous computation problem for decades00:48
fennwhy do you call it a hijack scenario?00:49
kanzurewell, mostly because you used that terminology before00:49
kanzureyou were asking about what possible exponential processes there are that we could hijack00:49
kanzurebut I think I've probably used it before too00:50
fenner.. i dont remember using that word00:50
fennanyway it's clear now what you mean00:50
kanzureanyway, it should be obvious00:50
kanzureyeah00:50
fenni thought you meant the ai was hijacking a human's actuators to do its bidding00:51
kanzurefun stuff00:51
fennall ur actuator r belong to us00:51
kanzureon this subject00:55
kanzurehttp://heybryan.org/pipermail/hplusroadmap/2008-May/000493.html00:55
kanzurethat's the Anissimov email, or at least my response to it00:55
kanzurebeep01:00
kanzureyou stopped.01:00
fenngood exponential intelligence, bad exponential intelligence01:02
fennif you outlaw microrobots, only criminals will have microrobots01:03
kanzureeh?01:03
fenn> Society goes to hell when I give myself a technology that lets me01:03
fenn> kill hundreds of people undetected (microrobotics, for instance),01:03
fenn> then millions of other people get it, then they all use it.  All it01:03
fenn> takes is 1/1000 people to be murderous for this to be a problem.01:03
fennthe proactionary principle turned out to be not what i expected01:03
kanzurewhat did you think it was?01:03
fenni was expecting, "instead of worrying about bad possibilities, lets implement safety nets now, before anything happens"01:04
kanzureit's sort of like that - but if anything, the safety nets you do implement, it's more about *your* safety nets, not Societal Blankets01:04
fennthe engineering approach to disaster prevention01:04
fenni'd feel much more secure having an anti-microbot shield than some wimpy legislation banning microbots01:05
kanzure"The Proactionary Principle recognizes that nature is not always kind, that improving our world is both natural and essential for humanity, and that stagnation is not a realistic or worthy option."01:06
kanzurefenn: re: shield, http://lifeboat.com/ but they need to get their act together01:06
kanzureor we could just fork a spacepod-colony-thing01:06
fennlifeboat suffers from internet-itis01:06
kanzurehehe01:06
fennsame reason luf failed01:06
fennthe millenial project01:06
kanzureI dunno, I have to wonder how the hell Eric Hunting could fail01:06
kanzureI mean, he seems kind of like me, especially in his lengthy emails01:06
fennhe writes much more coherently01:07
kanzureyes01:07
kanzuremaybe he's been over the arguments often?01:07
fenni dont think so01:07
fenni've google-stalked him and he mostly writes to the luf-team list01:07
kanzurethe proactionary principle - "Let a thousand flowers bloom! By all means, inspect the flowers for signs of infestation and weed as necessary. But don’t cut off the hands of those who spread the seeds of the future.01:07
fennand his old shelter webpage which is now stagnant01:08
kanzureaka. don't shoot yourself in the foot, you retards01:08
kanzureor more likely - in the head - that's a better modern interpretation01:09
kanzurebut really, I don't know if you're following my "ai will not bring about the singularity" line of reasoning or not01:10
fennheh, you get around http://article.gmane.org/gmane.linux.debian.apt.devel/1448801:10
kanzuredo you know how much of a pain in the ass it is to googlestalk myself?01:10
kanzure:(01:10
fennthat was from googling for "eric hunting"01:11
kanzurehaha01:12
fenn"People will do the research anyway, just like they code software anyway." unfortunately its much more likely that people get paid to engineer super-viruses rather than doing it for fun01:16
fennbut it will go in some military black budget account vs being out in the open where we can see it (if you cut research spending)01:17
kanzureyikes :(01:17
kanzurethere's a part in the email - a quote - that mentions "illegal cures"01:18
-!- kanzure [n=bryan@cpe-70-113-54-112.austin.res.rr.com] has quit ["Leaving."]01:22
-!- ybit [n=u1@unaffiliated/ybit] has quit [Read error: 110 (Connection timed out)]03:35
-!- fenn [n=pz@adsl-99-133-185-36.dsl.bltnin.sbcglobal.net] has joined #hplusroadmap17:36
-!- Topic for #hplusroadmap: http://heybryan.org/ http://heybryan.org/mediawiki/ http://heybryan.org/exp.html | krebs is now servicing the channel. try !help17:36
-!- Topic set by kanzure [] [Tue Apr 29 18:54:31 2008]17:36
[Users #hplusroadmap]17:36
[ fenn] [ krebs] [ nsh] [ Phreedom] [ ybit] 17:36
-!- Irssi: #hplusroadmap: Total of 5 nicks [0 ops, 0 halfops, 0 voices, 5 normal]17:36
-!- [freenode-info] please register your nickname...don't forget to auto-identify! http://freenode.net/faq.shtml#nicksetup17:36
-!- Channel #hplusroadmap created Sat Mar 22 15:44:12 200817:36
-!- Irssi: Join to #hplusroadmap was synced in 33 secs17:37
-!- kanzure [n=bryan@cpe-70-113-54-112.austin.res.rr.com] has joined #hplusroadmap17:44
kanzureHey.17:45
fennhi.17:55
fennthe paper knot model got my roomate's 8-yr-old daughter interested in blender17:56
kanzureuh, to what extent17:56
kanzureI would *not* recommend blender for 8 yr olds17:56
fennheh why not?17:57
kanzureunless they can sit and click for hours and not expect results for a while17:57
kanzureooh17:57
kanzurescrew that17:57
kanzurehttp://youtube.com/ has videos on blender17:57
kanzureset that kid up with that17:57
fennshe didnt seem interested in me "helping"17:57
kanzureah, that's good17:57
kanzurehttp://biobus.org/ - they drive around with equipment in a bus17:58
kanzureif I have to use blender, I think I'll take up the scripting approach18:00
kanzurefenn: have you found any holes in my argument from last night re: ai isn't going to solve the skdb problem?18:13
kanzuregmail is not responsive18:24
fennyes lots of them, but you arent paying attention18:26
fennan AI thinks in code, so the whole problem of turning human understanding/knowledge into codified programs doesn't exist18:27
fennalso, since it doesnt have the same type of hardware processing built-in (ex: face recognition) it will come up with different solutions than a human would, but it can still leverage human developments18:29
kanzure"debian science git repository is up and running" hurray (an email I got 12 hours ago apparently)18:31
kanzuresigh18:31
kanzurelook, I don't care if it's thinking in code18:31
kanzurethat doesn't matter18:31
kanzurethe problem would exist even if *we* thought in code18:31
kanzureit still needs to be 'grounded' with instrumentation into physical reality, it still needs to be able to assemble and process the information that really hasn't been assembled yet, and it would need to do actually *do* stuff18:32
kanzureI'm not saying it's impossible for an ai to do it18:32
kanzureI'm just saying that I'm questioning the focus on building an ai first, rather than you know18:32
kanzuredoing both things at once18:32
fennthe assumption is that the ai will "bootstrap" itself and become vastly more intelligent than us, so that then manufacturing problems become an insignificant background task18:51
fenni see that lots of smart people have been working on AI for fifty years now and havent exactly achieved their goals18:52
fennhowever, nobody's tried what we're doing to my knowledge18:52
kanzureso you think intelligence solves the manufacturing problem?18:57
kanzureas opposed to aggregation of knowledge that we've acquired by wrestling with experiments and modeling etc.18:58
kanzureI mean, I ultimately hope that intelligence can bruteforce its way out of any hell, any shithole -- believe me, I spend many hours each day in school thinking about this18:58
fennno, i think the intelligence will undoubtedly find a way to do real empirical experiments in a rapid manner, without having to waste its time writing grant proposals and journal articles18:59
kanzuresure, okay19:00
fennits not just going to sit in a box and think19:00
kanzureso it needs to be interfaced with the outside world19:00
kanzurekind of like with a manufacturing/peripheral system19:00
kanzureoh wait19:00
kanzure:)19:00
fennbut that doesnt mean you cant accomplish anything at all inside a box19:00
kanzureright, of course19:00
kanzurelots of good models can be made and so on19:00
fennnot models, more like aggregation and formalization of data19:01
kanzureI've found that I've been able to do some good predictive modeling of sorts ... not "here's the situation, now predict" but rather constructing ideas that I later find applicable 19:01
kanzureinstead of just-in-time learning.19:01
fennsorting and assimilating information19:01
fennsound familiar :)19:01
kanzureoh19:02
kanzureit doesn't really matter anyway, the "ai bootstrapping is wrong" idea doesn't matter19:02
fennhmm19:02
kanzureoh, no, nevermind ... it's to steal Eli's funding ;-)19:02
kanzureI forgot.19:02
kanzurefuture archivists: I'm half joking.19:02
fennyou know he moved to california from illinois in order to get funding..19:03
fenni think that's a large part of it19:03
fenndear google:19:03
fenni know that you're young and immature, but i really think you should cut it out with the spelling substitutions19:03
fennthanks,19:03
fenn  -one of your many fleshy carbon-based subjects19:03
kanzurefenn: He was on the extropy-chat mailing list starting in 1999 and then got funding sometime four years later or something19:04
kanzurekind of peculiar19:04
kanzureI think he was doing some writing that got him noticed, I don't know19:04
fennwell, its quite interesting, have you read any of his papers?19:04
kanzureoh, a few 19:05
kanzureA Technical Explanation of Technical Explanation19:05
kanzurehis intelligence-book19:05
kanzureone on starring into the singularity19:05
kanzure(which talks about big giant numbers and exponential processes ... from the context of ai)19:05
fennwell, the interesting and unique ones are his writing about friendliness19:06
kanzureah yes, FAI arguments19:07
kanzureholy shit19:47
kanzureMyanmar's cyclone - 100k dead people, 2 million needing some sort of assistance (housing, food, medicine, supplies); - this isn't the "holy shit" part19:47
kanzureinternational support teams have been deployed, but are stopped at the borders19:47
kanzurethe government will not issue visas to let them in19:47
kanzureso how's that for sick19:48
kanzurethese people are ready to be deployed19:48
kanzurebut they choose not to because "oh, well, we don't have visas"19:48
kanzurefenn: Let's get a little sick.20:11
kanzureIn our own way.20:11
kanzurewe can do an analysis on how well a clanking replicator could have responded to the disaster20:11
fennum, no20:12
kanzureas a way to get funding20:13
kanzurewithin 20 days, at one day replication cycles, there'd be 1 million units20:13
fennwe already have self-replicating general purpose machines and they were stopped at the border because they didnt have visas20:13
fennotherwise you get into political discussions20:14
kanzurethey don't replicate fast enough, and when they *do* replicate, you need to train them for at least 12 to 15 years before they have a clue20:14
kanzurethey weren't "stopped" - they respectfully declined to enter without visas20:14
kanzurethe military wasn't showing up with tanks and machinery to gun them down20:14
fennoh they werent? well thats stupid20:14
fenni mean, more stupid20:14
kanzurefenn: ever read anything by Howard Bloom? Lucifer Principle, or Global Brain, in particular ?20:17
kanzureI'm trying to figure out if I should read them right now. I got them from a friend at the WTA.20:19
kanzureI'm pretty sure the Lucifer Principle is kind of like the Nonzero book (re: the nonzerosumness of collaboration throughout evolutionary history)20:20
kanzureerm, I shouldn't bother asking that20:25
kanzureneed to get back to work :)20:25
kanzurewell, shit21:23
kanzurehttp://www.openverse.com/~dtinker/agalmics.html21:23
kanzureI was going to go contact Robert Levin (lilo) to talk about that pape21:23
kanzure*paper21:23
kanzurebut then I realized he founded freenode and proceeded to die21:23
fennyep21:23
kanzureI didn't remember all of the updates from lilo I was getting off of freenode a year back from now21:24
kanzureremembr those?21:24
kanzure*remember21:24
fennyeah he was a good guy, i dont really understand why so many people hated him21:24
kanzurepeople hated him?21:24
fennmaybe they were just more vocal about it21:24
fenntoo bad the word agalmic never caught on21:27
kanzureit's a useful concept, "marginalization of scarcity"21:28
kanzurealthough importantly, the knee-jerk reaction is to say "but scarcity is still real, blah blah blah - don't marginalize real problems"21:28
fennum, no it isnt21:29
-!- nsh [n=nsh@wikipedia/nsh] has quit [Remote closed the connection]21:29
-!- nsh [n=nsh@87-94-155-173.tampere.customers.dnainternet.fi] has joined #hplusroadmap21:29
kanzureI mean to say that 'marginalization of scarcity' can be taken to mean 'marginalization of the issue of scarcity' (covering it up) (which this essay is not about, yes, but I refer to the use of the phrase (marginalization of scarcity))21:30
fennon that topic, have you read http://www-formal.stanford.edu/jmc/progress/  (particularly the natural resources part)21:30
kanzure no, please hold21:30
fennits a huge linky behemoth spanning much (for me) new material21:31
fenninteresting site though, i recommend reading all of it21:31
fennbtw he discovered/invented lisp21:31
kanzurehaha, token "E.O. Wilson" reference :)21:31
fennnice21:32
fennthere's a lot of proactionary philosophy on that site21:33
kanzurehe has an odd perspective21:34
kanzurehe's trying to write for the masses apparently21:34
kanzurewell, not "the masses"21:34
kanzurebut a wide audience21:34
fenni think its supposed to be like a 'getting up to speed' page, somethin you can point to for correcting a lot of widespread misunderstandings and baseless fears21:35
kanzurehm21:36
kanzureif he had skdb available at the time, he should have just created new projects for each of his points and jot down a few technical notes on implementation and design21:36
kanzureand then just aggregate all of them together to address the issues21:36
kanzureinstead of just leaving us with a static html page that doesn't link to actual *solutions*21:37
kanzure(yes, the words describe solutions, sure)21:37
fennhmm i dont agree with you on that21:37
kanzurehow so21:37
fennif we try to expand skdb to fulfill every type of information storage and conveyance, it'll be good at nothing21:37
kanzure'information storage and conveyance' ?21:38
kanzureyou mean, the engineering projects that are mentioned on McCarthy's page?21:38
fennlike, people try to use mediawiki for everything, when it's really designed to be an encyclopedia21:38
fennnot a blog, not a photo gallery21:38
kanzureso why wouldn't people use SKDB for projects21:39
fennif he had nuclear reactor designs on his site, would that be better?21:39
kanzureprobably :)21:39
kanzureand those designs would belong in SKDB, IMO21:39
fennreferencing your sources is great but it's not absolutely necessary to get the point across21:40
kanzuretrue, but part of the whole point is to provide a way for people to get involved21:40
kanzurethat's what "getting up to speed" is about21:40
kanzureunlike the news, where you just hear stuff and nod21:40
fennand how is skdb supposed to store stuff like "Life expectancy in both the rich and poor countries.  Infant mortality in rich and poor countries. 21:41
fennDays lost per year due to illness.:21:41
kanzureno, I mean the engineering projects he mentions would go in there21:41
kanzurelife expectancy improvement tech would go in there21:41
kanzureand he's free to link over to the solutions21:41
fennit sounded like you wanted to put the whole webpage in an skdb project21:41
kanzurewell ... documentation?21:41
kanzuredoesn't that go within a project?21:42
fennsure, and it should be wiki-able21:42
fenni'm notoriously bad at citing references (mostly because i can't remember where i learned something)21:44
kanzure:(21:45
fennheh remember your "squeezing the most out of google programmers" http://www.flickr.com/photos/10719678%40N08/1424289534/in/photostream/21:47
kanzureheh, squeezing every last drop21:50
fennthis is hilarious, its about bad urban planning and suburbia http://video.google.com/videoplay?docid=-305728017890905149722:43
kanzurefenn: why doesn't unfold.py work for me?23:38
kanzureit generates SVG, but I don't get anything in the SVG file23:39
fennit's very small23:39
fennselect all, then zoom to selection23:39
kanzureselect all what ?23:41
fennctrl-a23:41
fennthen click on magnifying glass with a dashed box in it23:42
fennor ctrl-drag on the handles to proportionally scale it up23:42
kanzureI'm getting nothing.23:43
kanzurethe file is nonempty23:43
kanzureexport to png also shows nothing23:44
kanzurehrm, this sucks, I'm going to have to learn blender23:48
* kanzure goes off to youtube to hear some annoying voices 23:48
fennin object mode, right-click on what you want to unfold, then run mesh->unfold, select 'curvature' and 'search' and click unfold, then save23:49
fenndoncha love GUI23:49
fennwhy does LUF attract so many wingnuts23:50
kanzuremany groups that use yahoo for mailing lists, are generally attractors of wingnuts for some reason23:53
kanzurejust in general.23:53
kanzureorions_arm is an interesting exception23:53

Generated by irclog2html.py 2.15.0.dev0 by Marius Gedminas - find it at mg.pov.lt!