--- Day changed Sun Sep 28 2008 | ||
kanzure | Steve's brother really, really hates ESR apparently. | 02:25 |
---|---|---|
kanzure | Something about terminfo being broken. | 02:26 |
kanzure | bkero: Know anything about Urszula Chajewska? | 03:23 |
bkero | Sorry | 04:11 |
ybit | concening skdb, instead of manually entering in information for a material, anyone want to share thoughts on using an "intelligent" (bot|program) to do classify and identify properties for the user | 04:16 |
kanzure | ybit: *cough* | 04:20 |
ybit | guess a realistic approach would just be to buckle down and enter in the info | 04:20 |
kanzure | ybit: you know I'm downloading matweb for a reason, right? | 04:20 |
kanzure | meep | 04:20 |
kanzure | http://designworld.matweb.com/search/datasheet_print.aspx?matid=44744 | 04:20 |
kanzure | unless you mean automated material testing equipment | 04:20 |
ybit | what about fruit? | 04:21 |
ybit | plants | 04:21 |
kanzure | ? | 04:21 |
fenn | what about meat on a stick? | 04:21 |
fenn | or popsicles? | 04:21 |
ybit | :P | 04:22 |
ybit | Seed the database with a number of functional parts, components, tools, etc. | 04:22 |
ybit | * Basic: lemon battery | 04:22 |
fenn | fuck lemon battery | 04:22 |
kanzure | seeding, right | 04:22 |
gene | huh? | 04:22 |
ybit | from the wiki | 04:22 |
gene | why? | 04:22 |
kanzure | so that you don't have to manually enter all this bullshit | 04:22 |
kanzure | re: meat on a stick | 04:23 |
kanzure | or really just re: bio in general, | 04:23 |
kanzure | note that I've been working off and on that synthetic biology circuit creator app | 04:23 |
kanzure | so somehow biobricks fit into the scheme of things somehow | 04:23 |
kanzure | yet it's not obvious to me that it's appropriate to place 'biobrick' next to a 'gear' in the same dir for a db. meh. | 04:24 |
fenn | you dont like the debian "throw twenty thousand packages in the same dir" strategy? :) | 04:24 |
kanzure | no | 04:25 |
kanzure | just my typical turtle objections | 04:25 |
kanzure | the molecule is not the gene | 04:25 |
kanzure | erm | 04:25 |
kanzure | the protein is not the gene | 04:25 |
fenn | a biobrick is not a gear, unless of course it is | 04:25 |
kanzure | biobricks could /encode/ gears | 04:26 |
kanzure | does that still make it a gear? | 04:26 |
kanzure | just wondering. | 04:26 |
kanzure | I could go either way on this | 04:26 |
fenn | the gene is source code for the protein | 04:26 |
fenn | there are a number of ways we could have "source code" for a gear as well | 04:26 |
fenn | so how do you classify a gear generator program | 04:27 |
* fenn grumbles about e-prime | 04:27 | |
kanzure | are we going to have stupid N part generators? | 04:27 |
kanzure | compiler theorem | 04:27 |
fenn | N part? | 04:27 |
kanzure | any problem can be broken into smaller parts | 04:28 |
gene | what .stl files? | 04:28 |
kanzure | say you have these parts: gear, train, rocket | 04:28 |
kanzure | so are you going to infinitely splice this? | 04:28 |
kanzure | into smaller and smaller turtle shells ? | 04:28 |
kanzure | so we have a turtle shell generator? | 04:28 |
kanzure | that sounds unreasonable, frankly | 04:28 |
fenn | no, that leads to running a simulation of the universe, which is computationally impossible | 04:28 |
gene | huh what are you guys talking about? | 04:28 |
kanzure | so we don't need a gear generator | 04:28 |
ybit | sorry, was distracted, just got a txt from my mum. supposedly, some previous u.s. president is related to me somehow. whatever. how is the data being formated into uniform yaml files? | 04:28 |
kanzure | but maybe we need a gear-configurator, yes | 04:28 |
kanzure | which data, ybit? | 04:29 |
gene | just store em as stl files | 04:29 |
ybit | all of it | 04:29 |
gene | all of it | 04:29 |
fenn | a gear generator program is one layer of turtles up in the abstraction hierarchy | 04:29 |
gene | on magnetic tape or something | 04:29 |
fenn | stl is practically useless for manufacturing gears | 04:29 |
gene | then convert em in to g code | 04:29 |
kanzure | gene: stl is a step backwards here | 04:29 |
kanzure | please don't | 04:29 |
kanzure | no | 04:29 |
gene | I know it's an arbitrary file formate | 04:30 |
fenn | gene just shut up | 04:30 |
kanzure | ybit: you check the repo yet? | 04:30 |
ybit | autogenix? | 04:30 |
kanzure | no no | 04:30 |
kanzure | this is recent | 04:30 |
ybit | ah | 04:30 |
kanzure | ybit: http://heybryan.org/~bbishop/docs/repo/ try one of the smaller files | 04:30 |
fenn | ok there's this abstraction hierarchy, going from "computer, build me a thing" to machines running machine code | 04:30 |
ybit | not yet | 04:30 |
kanzure | ybit: go look at a file in there that won't take forever to download | 04:31 |
kanzure | and that's the format that I've found out that the lab is working with | 04:31 |
kanzure | we need to improve this a bit | 04:31 |
kanzure | preferably in yaml of course | 04:31 |
fenn | genes are source code that eventually gets translated into DNA and then RNA and then protein | 04:31 |
fenn | but you can't store DNA in your database | 04:31 |
fenn | because it's on a hard disk | 04:31 |
gene | you can but it's tricky | 04:32 |
gene | there are dna databases out there | 04:32 |
gene | real meatspace dna | 04:32 |
fenn | yay | 04:32 |
gene | on bacterial plasmids | 04:32 |
fenn | i'm talking about SKDB | 04:32 |
fenn | skdb is supposed to be easily distributed via protocols like git and http | 04:33 |
fenn | therefore it cant contain a freezer full of eppendorf tubes | 04:33 |
gene | yup | 04:33 |
fenn | or a warehouse full of gear manufacturing equipment | 04:33 |
fenn | so, what was the problem? | 04:34 |
fenn | i look at some arbitrary chunk of metal and can confidently say "this ia a 10DP 52 tooth gear" | 04:34 |
fenn | or some protein is firefly luciferase | 04:35 |
fenn | if you dont like it, learn to speak e-prime | 04:35 |
gene | how do you compile it? | 04:37 |
fenn | compile what? | 04:37 |
gene | the gear, the luciferase | 04:38 |
kanzure | :) | 04:38 |
fenn | well, they are different "targets" | 04:38 |
fenn | one targets a gear hobbing machine and associated robots | 04:38 |
gene | indeed they are | 04:38 |
fenn | the other targets some dna sequencing machinery and expression vectors | 04:38 |
gene | a gear hobbing machine can only make gears | 04:39 |
fenn | not true | 04:39 |
gene | what else can it make? | 04:39 |
fenn | depends on the machine | 04:39 |
kanzure | ybit: anyway, what were you asking? or did I answer it already or something | 04:39 |
gene | a mill is more universal | 04:40 |
fenn | most machine tools are just three sliding parts and a spinning part | 04:40 |
gene | yup | 04:40 |
fenn | sometimes they have two spinning parts, these can be used as gear hobs | 04:40 |
gene | unless it's one of those fancy noncartesian setups | 04:40 |
fenn | but there are other problems, like how do you mount the gear, and do you have encoders on both spindles, and what tooling to use, etc | 04:40 |
gene | jigs | 04:41 |
gene | lots and lots of jigs | 04:41 |
kanzure | anybody remember where I put baez.zip ? | 04:41 |
fenn | try locate | 04:41 |
gene | jigs are pretty much the best method for automated construction | 04:42 |
fenn | what is a "jig" | 04:42 |
gene | a part that holds another part in place during assembly | 04:42 |
gene | say a part that holds a gear and part that connects with that to align the gears onto the rod | 04:43 |
fenn | why is it i'm picturing big blocks of metal with bolts all over them instead of a robot arm | 04:43 |
gene | it's sort of like that | 04:43 |
gene | except on the end of the arm | 04:43 |
gene | stratasys 3d printers are used to make jigs | 04:44 |
fenn | good for them | 04:44 |
* fenn is so sick of hearing about 3d printers it's not funny anymore | 04:44 | |
gene | Man I could really make you sick | 04:44 |
gene | ** printers are fairly universal | 04:45 |
fenn | you could synthesize some ebola and send it to me in the mail | 04:45 |
gene | NO | 04:45 |
gene | I don't synthesize ebola | 04:45 |
fenn | meh | 04:45 |
kanzure | guess it's not universal then | 04:45 |
gene | but, you could make parts to make a machine to make ebola | 04:46 |
fenn | if i did that, could i say the printer made ebola? | 04:46 |
fenn | what if i just pressed the green button, who did it? | 04:46 |
gene | sort of | 04:46 |
fenn | is anyone here a philosopher? please step into the chamber to your left | 04:47 |
gene | you pressed the button you caused it | 04:47 |
gene | been wanting to make a machine to click ok on EULAs for me | 04:47 |
fenn | what if we have a cat, in a box, with a gamma ray source | 04:47 |
gene | the cat dies | 04:48 |
fenn | there's a fifty-fifty probability the light will blink, based on whether our detector senses a gamma ray or not | 04:48 |
gene | nope | 04:48 |
fenn | if i see a blinking light i'll press the button | 04:48 |
fenn | at what time did the two trains pass? | 04:48 |
gene | green | 04:48 |
kanzure | god I'm confused. | 04:49 |
gene | why is a mouse when it spins? | 04:49 |
gene | this topic isn't very productive | 04:51 |
gene | it must be ended | 04:51 |
fenn | channel rule proposal: no philosophy | 04:51 |
kanzure | where will the turtles live? | 04:51 |
kanzure | am I missing messages or something? | 04:52 |
fenn | no | 04:52 |
kanzure | because those previous messages I'm seeing make very little sense | 04:52 |
gene | http://en.wikipedia.org/wiki/Jig_(tool) | 04:52 |
gene | jig | 04:52 |
fenn | do not adjust your computer monitor | 04:52 |
fenn | irssi is currently experiencing some mental difficulties | 04:53 |
kanzure | fenn: I was going to stop you after your first message to gene because I didn't see what you were going for | 04:53 |
gene | So you're worried about degradation with time? | 04:54 |
fenn | is it clear now? i can eloquate | 04:54 |
gene | "a single Fanuc two-armed robot assembles smaller robots using a three-dimensional vision sensor and six force sensors that correct random positional errors." | 04:55 |
gene | the answer is vision sensors | 04:55 |
fenn | a three dimensional vision sensor.. i want one of those | 04:55 |
gene | unfortunately | 04:55 |
gene | you have one already | 04:55 |
fenn | nuh-uh | 04:55 |
gene | unless you wear an eyepatch | 04:56 |
kanzure | stereo? | 04:56 |
gene | yeah | 04:56 |
gene | eyes | 04:56 |
gene | http://www.molecularassembler.com/KSRM/3.7.htm | 04:56 |
gene | source | 04:56 |
gene | Yes I know you guys hate freitas | 04:57 |
kanzure | not true | 04:57 |
fenn | i dont hate freitas, but i dont blindly believe everything i read in it | 04:57 |
fenn | or extrapolate to absurd degree | 04:57 |
gene | freitas has sticky fingers | 04:57 |
gene | back to the topic | 04:58 |
gene | how to replicate a camera | 04:58 |
gene | or 3d vision sensor | 04:58 |
fenn | that's the topic? | 04:58 |
gene | how do you replicate a camera | 04:58 |
kanzure | god, I haven't been this confused since the great freenode disconnect of 2006 | 04:59 |
fenn | me personally, i havent replicated any cameras, but if i were to choose the simplest path it would probably involve a woman at some point | 04:59 |
gene | heh | 04:59 |
fenn | and a lot of extraneous production | 04:59 |
gene | I mean in a mechanical replicator | 04:59 |
fenn | i suppose the process could be simplified with a tank of guppies or something | 04:59 |
gene | hmmm... | 05:00 |
gene | guppies do have the computational power of a PS2 | 05:00 |
fenn | is that so? who did you parrot that from? | 05:00 |
fenn | what is the computational power of a cup or warm tea? | 05:00 |
gene | not much | 05:01 |
fenn | what is "computational power" anyway? | 05:01 |
gene | Some dude trying to extrapolate to the singularity | 05:01 |
gene | so how do you make a ccd | 05:01 |
fenn | you take a slice of ultrapure silicon... | 05:02 |
ybit | ybit, you answered my question, essentially, it's a work in progress | 05:02 |
ybit | """ | 05:02 |
ybit | [23:19] <kanzure> and that's the format that I've found out that the lab is working with | 05:02 |
ybit | [23:19] <kanzure> we need to improve this a bit | 05:02 |
ybit | [23:19] <kanzure> preferably in yaml of course | 05:02 |
kanzure | ybit: work in progress as in, I'm too lazy to fix it *right now*. I should though. | 05:02 |
kanzure | if I was more awesome I would. | 05:02 |
fenn | spray some photo-mask solution on it, photo-sensitize the mask, spray some etchant on it, repeat with various patterns of etchant and dopants | 05:03 |
gene | by dope do you mean coat with thin film of metal? | 05:04 |
kanzure | are we talking typical semiconductor manufacturing? | 05:04 |
kanzure | no | 05:04 |
gene | ok | 05:04 |
fenn | no, dopants are usually salts that diffuse into the silicon | 05:04 |
gene | so shoot ions at it | 05:04 |
gene | btw | 05:04 |
kanzure | this is all so random | 05:05 |
gene | I have always wondered | 05:05 |
* fenn blames gene | 05:05 | |
kanzure | I have a strong urge to go read ruby docs now. | 05:05 |
gene | how do they keep the wafers aligned during manufacturing | 05:05 |
fenn | with piezo stages | 05:05 |
fenn | and alignment marks | 05:05 |
* fenn pulls explanation out of ass | 05:05 | |
gene | that's bad | 05:05 |
ybit | manufacturing of CCDs is interesting | 05:06 |
* ybit listens | 05:06 | |
ybit | and i'm spent | 05:06 |
gene | piezoes are hard to replicate | 05:06 |
fenn | you know i'm just making all this up.. if you really want to know, go read a book | 05:06 |
gene | they require exotic elements | 05:06 |
gene | that actually sounds about right | 05:06 |
ybit | kidding of course | 05:06 |
gene | even worse | 05:07 |
gene | they move the whole damn mask around | 05:07 |
gene | over the whole chip | 05:08 |
gene | http://en.wikipedia.org/wiki/Stepper | 05:09 |
kanzure | Steve's thinking we're going to be needing an ontology of ways for parts to be in surface contact, in order to get the subcomponents in designs working correctly, for simulation and such | 05:09 |
ybit | kanzure, kind of on-topic with fenn and gene, i remember you asking me what i would be doing in my research project, it's essentially ccd image reductions with iraf, nothing special. see http://iraf.net/irafdocs/# for more details | 05:11 |
fenn | kanzure: you mean like, helical=nut/bolt, planar=chip/heatsink ? | 05:12 |
kanzure | sort of. also glues between things, too | 05:12 |
fenn | a lot of that is what i call 'interfaces' | 05:12 |
kanzure | i.e., given any two parts that are next to each other, WTF is their interface and how does that influence the simulation? | 05:12 |
gene | maybe you should think of it in terms of force | 05:13 |
kanzure | the big example of what I'm thinking of is the car design scenario | 05:13 |
kanzure | two sheets of metals in various shapes | 05:13 |
kanzure | how do they connect? | 05:13 |
gene | Kanzure, you really need to learn statics | 05:13 |
kanzure | this greatly influences CFD/aerodynamics | 05:13 |
fenn | it's more generic than that, you have to worry about electrical and chemical interfaces too | 05:13 |
kanzure | yep | 05:13 |
gene | or solids | 05:13 |
kanzure | oh boy. | 05:13 |
kanzure | well, I know that there are chemical simulations | 05:13 |
kanzure | but this is highly domain dependent | 05:13 |
kanzure | right? | 05:13 |
gene | they have methods for describing things such as ball joints, rollers, bearings and such | 05:14 |
gene | so you can calculate the forces on them | 05:14 |
fenn | there are a lot of unpredictable ways that interfaces can interact | 05:15 |
kanzure | Steve thinks 'geometrical computations' are the only showstoppes here | 05:15 |
kanzure | hrm | 05:15 |
fenn | there are also predictable ways, which i think we should focus on | 05:15 |
kanzure | give me something other than geometry that is a Big Deal | 05:15 |
gene | sand | 05:15 |
fenn | why is geometry a Big Deal? | 05:15 |
gene | sand can clog up parts | 05:15 |
kanzure | fenn: two sheets of metal example | 05:16 |
kanzure | how do you specify two metals and their interface? | 05:16 |
kanzure | one has more stress than another; etc. there's shape involved there. | 05:16 |
kanzure | you don't just have it magically determine the most optimal shape or something. | 05:16 |
kanzure | there's all sorts of variables on shape. | 05:16 |
kanzure | uhm | 05:16 |
gene | well what are you trying to make | 05:16 |
kanzure | I'm not sure if that's a good characterization o the problem | 05:16 |
kanzure | SKDB | 05:16 |
fenn | there's all sorts of ISO fit classes, is that helpful? | 05:16 |
gene | what needs to be made from metal? | 05:17 |
kanzure | fenn: I'm trying to track down a link. | 05:17 |
fenn | there should be a standard process for flexible chip fab | 05:19 |
kanzure | no, I mean, | 05:20 |
fenn | so you can run your chip through, and they align the mems mirror array according to your cad spec | 05:20 |
kanzure | could you help me find a ref to ISO fit classes | 05:20 |
fenn | oh, yes, sorry | 05:20 |
gene | photosensitize the chips with lasers | 05:20 |
* kanzure kicks google | 05:20 | |
gene | or even better a electron beam | 05:20 |
fenn | iso = information scarcity organization :) | 05:20 |
kanzure | http://www.gizmology.net/nutsbolts.htm maybe | 05:20 |
kanzure | heh | 05:20 |
gene | yeah, Kanzure you need to learn solids | 05:21 |
gene | that is solids | 05:21 |
gene | you have to do Finite element mesh to model solids correctly though | 05:22 |
kanzure | FEM isn't a big deal | 05:22 |
kanzure | unless I'm supposed to do it on paper | 05:22 |
gene | good | 05:22 |
fenn | yeah the nuts/bolts thing is representative, they also have other classes for pegs in holes and so on | 05:22 |
fenn | all rather arbitrary if you ask me | 05:22 |
kanzure | classes 1 through 4 | 05:23 |
kanzure | yep | 05:23 |
kanzure | arbitrary. | 05:23 |
kanzure | this doesn't help much with the issues | 05:23 |
fenn | no | 05:23 |
kanzure | maybe I'm missing something. are there general principles for dealing with the different iso fit classes? | 05:24 |
fenn | it just makes it easy to write "H7" on a shop drawing | 05:24 |
fenn | GD&T is more general and "pure" but it's harder to read and write | 05:25 |
kanzure | write? | 05:25 |
kanzure | what? | 05:25 |
fenn | with a pencil | 05:26 |
fenn | here read this http://www.efunda.com/designstandards/gdt/introduction.cfm | 05:26 |
fenn | tat's kinda long but at least you'll know wtf i'm talking about | 05:28 |
kanzure | http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6V1P-4MNRMXW-2&_user=108429&_rdoc=1&_fmt=&_orig=search&_sort=d&view=c&_version=1&_urlVersion=0&_userid=108429&md5=f138e2c93b47ea0fa213557c78658776 Flexural behaviour of bonded-bolted butt joints due to bolt looseness | 05:30 |
gene | what does the link concern | 05:31 |
kanzure | http://heybryan.org/~bbishop/docs/manufacturing/Flexural%20behaviour%20of%20bonded-bolted%20butt%20joints%20due%20to%20bolt%20looseness.pdf | 05:31 |
gene | bolt looseness? | 05:32 |
gene | oh shoot | 05:32 |
gene | I think I found a problem | 05:32 |
gene | with space replicators | 05:32 |
gene | metal parts cold weld in a vacuum | 05:33 |
gene | never mind not a problem | 05:33 |
gene | for what I was thinking about | 05:34 |
fenn | metal parts cold weld just about anywhere | 05:34 |
gene | but especially in vacuums | 05:35 |
gene | http://en.wikipedia.org/wiki/Cold_welding | 05:35 |
gene | you just oxide coat your metal parts | 05:36 |
gene | to avoid it | 05:36 |
gene | Product design by an advanced "expert system" software package or by humans remotely or interactively, using a computer design system that stores data on models, computes optimal designs for different options, displays results for approval, and allows efficient process iteration. | 05:38 |
gene | Production planning, an optimized plan for the manufacturing processes generated by a computer on the basis of product-design outputs, scheduling, and line-balance algorithms, and varying conditions of ore-feedstock deliveries, available robot resources, product mix and priorities. Planning includes routing, timing, work stations, and operating steps and conditions. | 05:38 |
gene | guess someone already thought of SKDB | 05:39 |
gene | http://www.molecularassembler.com/KSRM/3.7.htm | 05:39 |
kanzure | everybody's *thought of* it | 05:39 |
kanzure | http://adciv.org/ | 05:39 |
kanzure | http://peoplescapitalism.org/ | 05:39 |
kanzure | http://p2pfoundation.net/Category:Design | 05:39 |
kanzure | http://oscomak.net/ | 05:40 |
gene | yeah | 05:40 |
gene | I hope we actually implement it | 05:40 |
fenn | i wish everybody had already thought of it, then i wouldnt have to explain every damn time | 05:42 |
gene | and i wish I had a particle accelerator in my dorm room | 05:44 |
fenn | hah - i've got one right next to me | 05:45 |
fenn | it's a Mazellan | 05:45 |
gene | what's a Mazellan | 05:46 |
fenn | Hansol Mazellan 400A | 05:46 |
gene | nah | 05:47 |
fenn | i've got another one, of a different design | 05:47 |
gene | how many KV | 05:47 |
gene | how dense is the beam? | 05:48 |
fenn | a couple kV.. no fancy model name/number | 05:48 |
fenn | made by Ionizair | 05:48 |
fenn | beam width is about 10cm | 05:48 |
gene | too bad | 05:48 |
gene | Dang i though the beams on those things were smaller | 05:49 |
gene | like the size of a the pixels or something | 05:49 |
fenn | well, it's not circular or gaussian, it's planar | 05:49 |
fenn | about 2cm in the other dimension | 05:49 |
gene | is that really how big the beam is on tvs and such? | 05:50 |
fenn | no, this isn't a TV | 05:50 |
gene | they both use beams of electrons to light up phosphors | 05:51 |
fenn | no, this uses beams of ion to accelerate a working fluid gas | 05:51 |
gene | huh? | 05:51 |
gene | plasma | 05:51 |
gene | of course | 05:51 |
fenn | you can barely see the plasma actually, has to be totally dark | 05:52 |
fenn | then you will see it at the tips of the field concentrators | 05:52 |
gene | Ok I am confused | 05:53 |
fenn | i dont like sticking my face up that close though, too much ozone | 05:53 |
gene | your monitor makes fucking OZONE? | 05:54 |
fenn | it's not a monitor | 05:54 |
gene | what the heck is it? | 05:54 |
fenn | an air purifier, silly | 05:54 |
gene | heh | 05:55 |
gene | that's sort of a particle accelerator | 05:55 |
kanzure | " Uh, well, if an implementation detail constraint is logically implied by known design constraints, then coming up with an implementation that has that detail is part of what the optimizer's job is" | 05:55 |
fenn | actually i guess it's planar gaussian beam profile | 05:55 |
gene | in other words | 05:56 |
gene | just run it through a genetic algorithm | 05:56 |
fenn | GA has to start with something that works | 05:56 |
gene | yup | 05:56 |
fenn | before GA you have to use monte carlo or something that covers the whole search spce | 05:57 |
gene | you can start with something that doesn't work | 05:57 |
gene | but that takes too long | 05:57 |
fenn | if it's stuck in a local minimum that doesn't work, probably wont ever get out of it | 05:57 |
gene | ever read Prey by Michael Crichton? | 05:58 |
fenn | no | 05:58 |
gene | dang | 05:58 |
gene | you just start accepting stuff with lower fitness levels | 05:58 |
kanzure | fenn: Steve is suggesting compartmentalized optimizers. So each turtle layer gets an optimizer. | 05:59 |
kanzure | with each optimizer trying to figure out the lower layer's results. | 05:59 |
fenn | yes that simplifies things a lot | 05:59 |
gene | speaking of GAs check this out: http://www.framsticks.com/ | 05:59 |
fenn | black-boxing | 05:59 |
gene | could you document all this stuff about turtles? | 06:00 |
gene | so that I can start understanding it | 06:00 |
fenn | turtle just means abstraction layer | 06:01 |
fenn | "it's turtles all the way down" | 06:01 |
gene | abstraction layer means | 06:01 |
kanzure | fenn: he doesn't do programming | 06:01 |
fenn | oh ffs | 06:01 |
gene | yes I don't | 06:01 |
* kanzure knows gene is thinking "fuck" | 06:01 | |
kanzure | yep | 06:01 |
kanzure | :) | 06:01 |
gene | indeed I am | 06:01 |
fenn | this is sorta hard over text-based medium | 06:02 |
kanzure | http://www.whatisthought.com/bayes2.ps | 06:02 |
gene | abstraction layer you mean like concepts? | 06:03 |
kanzure | uh. | 06:03 |
kanzure | forget everything you know about english | 06:03 |
fenn | english is an abstraction layer | 06:03 |
gene | oh so what you call the parts | 06:04 |
fenn | i point in the direction of a collection of hair, bones, whiskers, claws, and teeth | 06:04 |
kanzure | fenn: best to have him do hard experience first before poisoning his mind | 06:04 |
fenn | we call it a "cat" | 06:04 |
gene | I am starting to see what you mean | 06:04 |
fenn | i point at a collection of subatomic particles, we call it "iron" | 06:04 |
gene | I was thinking that SKDB should imitate what the brain does | 06:04 |
gene | it looks like you already are | 06:04 |
fenn | brain doesn't do abstraction, it does categorization | 06:05 |
gene | I can't read .ps files | 06:05 |
fenn | some paper about bayesian tree algorithms | 06:06 |
fenn | not relevant | 06:06 |
kanzure | gene: you might as well ignore it, I'm not sure of its relevance really | 06:06 |
gene | ok then | 06:06 |
gene | but what I was thinking is that you make a program that is brain like | 06:07 |
fenn | we dont know enough about how the brain works to do that | 06:07 |
gene | have it read stuff online to "learn" how things work | 06:07 |
gene | but we know the basics | 06:07 |
fenn | and if we did, it would be a bad idea anyway, because then your program would get tired and lie to you and generally have all the problems that humans do | 06:07 |
gene | why would it lie? | 06:08 |
fenn | (and skdb is not yet another agi project) | 06:08 |
gene | I'm talking about something with the intelligence of a cow here | 06:08 |
gene | and maybe not even that | 06:08 |
kanzure | what the fuck is intelligence | 06:08 |
fenn | yes let's please not go down the philosophy pit | 06:08 |
fenn | gene have you ever played video games? | 06:09 |
gene | yes | 06:09 |
fenn | video games are full of simplified abstractions that are presented more or less directly to the player | 06:09 |
fenn | press a button, door opens | 06:09 |
gene | ok | 06:09 |
fenn | there's no fine-grained simulation of joined pieces of wood going on | 06:10 |
gene | that's sort of what I am hitting at with the brain thing | 06:10 |
gene | fire can be made by making wood hot | 06:10 |
kanzure | gene: there are neural networks that can be modularly involved with the automated design work | 06:10 |
kanzure | we're calling it "rule rewriting" | 06:10 |
kanzure | which is just connectivity rewiring in design graphs | 06:10 |
kanzure | this is _not_ a one-to-one mapping | 06:10 |
kanzure | in conceptspace. | 06:11 |
gene | ok | 06:11 |
fenn | rule rewriting? | 06:11 |
gene | I am just wondering about having something to put stuff into SKDB automatically | 06:11 |
kanzure | http://www.whatisthought.com/eric.html has some papers including "Where Genetic Algorithms Excel" | 06:11 |
fenn | (i didnt download your zip file) | 06:11 |
kanzure | graph rewriting rules, sorry | 06:11 |
fenn | isnt a graph just a collection of connections and nodes? | 06:12 |
kanzure | given such and such in the graph, try such-and-such | 06:12 |
kanzure | yes | 06:12 |
fenn | so what is there to change? | 06:12 |
kanzure | these are possible mutation operators | 06:12 |
kanzure | well | 06:12 |
kanzure | imagine what you'd write to go through all possible designs | 06:12 |
kanzure | (much like the schedule optimizer) | 06:12 |
kanzure | instead of going through them one by one you would be making some rather large jumps | 06:12 |
fenn | all possible designs = an infinite set | 06:12 |
kanzure | correct! | 06:12 |
fenn | -_- | 06:12 |
kanzure | so you have an upper limit, like 10 million, but the first 10 million might just be going down the left hand side of the tree | 06:13 |
gene | only some will be autocatalytic | 06:13 |
kanzure | fenn: so obviously we don't want to be left-hand-travelers, right? | 06:13 |
kanzure | fenn: so that's why we use some graph rewriting rules. | 06:13 |
fenn | i dunno, maybe the left hand path is imbued with certain mystical powers | 06:13 |
kanzure | heh :) | 06:13 |
kanzure | theoretically, there might be some rewriting rules that tend to work for certain problem domains | 06:13 |
fenn | what basis is there for thinking that? | 06:14 |
kanzure | none whatsoever, I just want to see really | 06:14 |
kanzure | isn't that why we're doing turtle optimizers though? | 06:14 |
kanzure | superturtle | 06:14 |
-!- superautomation is now known as superturtle | 06:14 | |
fenn | we're doing turtles because it reduces the amount of computation necessary | 06:14 |
kanzure | or because we know no other way. | 06:15 |
kanzure | heh | 06:15 |
fenn | sure, nobody has found the theory of everything | 06:15 |
superturtle | right | 06:15 |
kanzure | erm, wrong face | 06:16 |
fenn | and if we knew it, it would just be equivalent to running another instance of the universe | 06:16 |
gene | we might have found the theory of everything | 06:16 |
fenn | stuff it | 06:16 |
kanzure | hahah | 06:16 |
gene | but we don't know if it is right | 06:16 |
gene | and using it for automated design like this would be very impractical | 06:17 |
fenn | gah, gene you should change your nick to train-derailment | 06:17 |
gene | ok | 06:17 |
fenn | so, graph recombination is good for GA's for some reason, that's why we have sex and gene splicing and so on | 06:18 |
kanzure | sure. | 06:18 |
kanzure | oh | 06:18 |
kanzure | it's not just recombination though | 06:18 |
kanzure | from previous designs. | 06:18 |
gene | it's imperfect recombination | 06:19 |
fenn | if i 'recombine' one element, is that any different from a random mutation? | 06:19 |
kanzure | might not be entirely random | 06:19 |
kanzure | but no, not much different | 06:19 |
fenn | what's a non-random mutation? | 06:19 |
gene | do you know about meiosis fenn? | 06:19 |
fenn | yes | 06:19 |
gene | good | 06:20 |
kanzure | suppose a subvariable has a linear improvement curve on the output of a simulation | 06:20 |
kanzure | oh god, what did I just write | 06:20 |
kanzure | I mean to say that it's no longer a PDE | 06:20 |
kanzure | and a mutation in a certain direction offers some amount of improvement | 06:20 |
kanzure | I don't know how this integrates | 06:20 |
kanzure | let me shut up | 06:20 |
fenn | ok so you think the algorithm is smart enough to extrapolate the curve and set the variable to maximum | 06:20 |
gene | I believe that is sort of what GA's do with regards to combinationt | 06:20 |
kanzure | user/designer is the guy changing from single-mutation to GA | 06:20 |
gene | DIRECTED MUTATION | 06:20 |
gene | are you fucking crazy? | 06:21 |
fenn | user/designer is doing the mutation? | 06:21 |
fenn | rawr not enough context | 06:22 |
fenn | i assume this has something to do with ADL | 06:22 |
fenn | or DEL or whatever you call it these days | 06:22 |
kanzure | no, user/designer is saying "I hate this you bitch, give me something better" | 06:22 |
gene | heh why not use lamarkian evolution | 06:22 |
kanzure | in other words | 06:22 |
kanzure | he's denying/affirming the rewrite rules | 06:22 |
kanzure | "show me things less with rule #34" | 06:22 |
gene | heh | 06:22 |
fenn | so the user is requesting the computer perform a random mutation? | 06:22 |
gene | oh | 06:22 |
kanzure | on some set of designs with graphs | 06:23 |
gene | that's easy to do | 06:23 |
kanzure | sometimes, yes, fenn | 06:23 |
kanzure | but the user isn't specifying which mutation to make, really | 06:23 |
gene | GAs can be quite pesky | 06:23 |
fenn | i just dont get what these rules are doing in the first place | 06:23 |
gene | they cheat | 06:23 |
gene | they take advantage of physics system glitches | 06:23 |
fenn | we all do | 06:23 |
gene | well I need sleep | 06:24 |
fenn | he's gone! :) | 06:24 |
kanzure | WTF is he | 06:25 |
kanzure | rewriting the connectivity of the graph. | 06:26 |
kanzure | okay, so in the HTML example | 06:26 |
fenn | i wish i could just download information into other peoples' brains sometimes.. | 06:26 |
kanzure | what I was going to implement. | 06:26 |
UtopiahGHML | fenn: learn pedagogy | 06:26 |
kanzure | or don't | 06:26 |
kanzure | just eat the newbies for breakfast | 06:26 |
fenn | i'm too nice | 06:26 |
kanzure | me too :( | 06:26 |
kanzure | HTML example => graph was the HTML elements as nodes | 06:27 |
fenn | ok | 06:27 |
kanzure | rewrite would transform one recognized "chunk" of HTML into another chunk | 06:27 |
kanzure | predefinededly. | 06:27 |
UtopiahGHML | you can learn manipulation instead but I think it's pretty close anyway. | 06:27 |
fenn | transform one chunk into another chunk- please elaborate on that | 06:28 |
kanzure | http://www.cs.mdc.ac.uk/research/PhDArea/saeed/paper1.pdf | 06:28 |
kanzure | fenn: <b>HELLO</b dfieowqiq> <-- Wrong. That type of element closure could be recognized and rewritten as something else. | 06:28 |
kanzure | huh | 06:28 |
fenn | Unknown host www.cs.mdc.ac.uk | 06:28 |
kanzure | error correction is actually a good example | 06:28 |
kanzure | but obviously we're not interested in correcting errors | 06:28 |
kanzure | erm | 06:28 |
kanzure | wait | 06:28 |
kanzure | errors have unrecognizable forms usually | 06:28 |
kanzure | bad example. shit. | 06:29 |
fenn | html has a syntax which defines the order characters are allowed to be arranged | 06:29 |
fenn | within that syntax you can make arbitrary rearrangements | 06:29 |
kanzure | s/characters/nodes/ and I'd say you have the idea | 06:30 |
kanzure | erm | 06:30 |
kanzure | I mean s/characters/elements/ | 06:30 |
kanzure | with the understanding that elements == nodes | 06:30 |
UtopiahGHML | you might to look at XHTML then XML then SGML if you want to understand HTML structure | 06:30 |
fenn | syntax defines something like <tag element = "foo"> text </tag> | 06:31 |
kanzure | Been dong that for years, UtopiahGHML ;-) | 06:31 |
kanzure | correct fenn. | 06:31 |
fenn | syntax says nothing about what "text" should be | 06:31 |
kanzure | oh, ignore text right now | 06:31 |
kanzure | just imagine that autogenerated latin BS ther | 06:31 |
kanzure | there | 06:31 |
fenn | lorem ipsum dolor sit amet | 06:31 |
kanzure | yes, that crap | 06:32 |
kanzure | in HTML, you have parent nodes and subnodes | 06:32 |
kanzure | <HTML><head><title><hi I am a child of title></hi I am a child of title> ...... | 06:32 |
fenn | and there is a grammar for those as well | 06:33 |
kanzure | reword | 06:33 |
kanzure | are you trying to affirm me? | 06:33 |
fenn | body tag can't be child of head tag | 06:33 |
kanzure | sure | 06:33 |
kanzure | that's a good constraint to mention. | 06:33 |
UtopiahGHML | might want to look at DOM and libexpat | 06:34 |
fenn | ok, where's the "rule" you're wanting to rewrite | 06:34 |
kanzure | uh oh | 06:34 |
kanzure | okay okay | 06:34 |
kanzure | that's just because of bullshit terminology | 06:34 |
kanzure | *you're* making a rule *for* rewriting the graph | 06:34 |
fenn | is a rule just an algorithm for satisfying constraints? | 06:35 |
kanzure | rule: you see a,b,c you turn it into a->b, b->c, so if I give you a randomly generated graph you look for matches for that rule to apply. | 06:36 |
kanzure | erm, stupid syntax there. commas would be nonconnections, -> would be a connection. | 06:36 |
UtopiahGHML | if you want to define rule of a well formed XML document (including XML) you might want to look at DTD | 06:36 |
kanzure | uhm, it could go either way so I guess I didn't need to send that less message | 06:36 |
fenn | yes UtopiahGHML we understand, we're not talking about html, we're talking about engineering design graphs | 06:36 |
kanzure | UtopiahGHML is right in a sort of different way | 06:36 |
kanzure | DOM is relevant, sure | 06:36 |
kanzure | BNF files are useful in a certain way too, but we're talking about looking at an example/sample file | 06:37 |
fenn | DTD is just a constraint grammar i think | 06:37 |
kanzure | right | 06:37 |
kanzure | something like that. | 06:37 |
UtopiahGHML | fenn: well to me it looks like you are talking about the underlying technical structure to support them and maybe using a lot of energy for things that doo already exist so Im just throwing suggestions in | 06:37 |
kanzure | the DOM could aide in some of this | 06:37 |
kanzure | argh | 06:38 |
kanzure | how do I word how this is not DOM? | 06:38 |
kanzure | I give you a graph. You see a graph. I give you a rule, "when you see this, you change it into this". You implement rule, painfully checking the whole graph. NP-hard. You cry. You finish. You didn't really check the DOM except maybe to make sure the rule wasn't retarded. | 06:38 |
UtopiahGHML | if it's an HTML document you can manipulate it with DOM (but it should be well-formed) | 06:38 |
fenn | DOM is totally unrelated | 06:39 |
UtopiahGHML | DOM manipulate nodes of a tree, HTML document are trees. | 06:39 |
kanzure | by retarded I mean "breaking the well-form-edness constraints" | 06:39 |
kanzure | get it? | 06:40 |
kanzure | fenn? | 06:40 |
fenn | what is the engineering purpose of the rule? what's the desired outcome? | 06:41 |
fenn | is it to prevent square pegs in round holes? | 06:41 |
kanzure | oh, yes | 06:42 |
kanzure | but also | 06:42 |
kanzure | consider you have a set of designs, and you have a list of the rules used to make each of them (and thus all of them) | 06:42 |
fenn | because i'd say that should be automatic, and why would the user be modifying a rule that prevented something obviously stupid? | 06:42 |
kanzure | and you figure you don't like the results of a particular rule application | 06:42 |
kanzure | so you just eliminate all results using that rule | 06:42 |
UtopiahGHML | because the user doesn't understand a software or model like the author indented it to be understood... | 06:43 |
kanzure | then user preferences might develop for certain rules. would be worth keeping track of, no? then you can switch between exploration v. exploitation (SA v. GA) with the 'knob' mechanism I was mentioning. this would be for selecting certain designs | 06:43 |
fenn | there is no author | 06:43 |
kanzure | kind of for going through the many, many results | 06:43 |
UtopiahGHML | oh | 06:43 |
UtopiahGHML | no author :) | 06:43 |
kanzure | filtering over possibility space, I guess | 06:43 |
fenn | unless you mean the guy who made the parts you're currently manipulating | 06:44 |
kanzure | just one? poor guy :) | 06:44 |
fenn | yeah :( | 06:44 |
kanzure | left hand path | 06:44 |
fenn | too many cooks spoils the tag-soup | 06:44 |
UtopiahGHML | if you invite a user to use a system, you are the author of the system | 06:44 |
UtopiahGHML | if your system provide affordance to your model | 06:44 |
UtopiahGHML | you are the author of the underlying model | 06:44 |
UtopiahGHML | so there is an author | 06:44 |
fenn | i guess | 06:45 |
UtopiahGHML | (you could even consider a genetic algorythm an author in some sort of way) | 06:45 |
fenn | that's what i mean by no author | 06:45 |
UtopiahGHML | to me the only option to "no author" is evolution thus evolutionary algo | 06:45 |
fenn | there's also brute force | 06:45 |
fenn | or some other "fuzzy" learning algorithm | 06:46 |
UtopiahGHML | brute force is evo algo with no return and selection only by final sucess | 06:46 |
UtopiahGHML | 1 step mutation/selection | 06:46 |
fenn | not necessarily | 06:46 |
kanzure | that's what I say SA is | 06:47 |
fenn | brute force can be systematic | 06:47 |
kanzure | had a paper ref | 06:47 |
kanzure | SAGA_hybrids.pdf | 06:47 |
UtopiahGHML | systematic or random you just build a space of solutions and try them with selection being only work,do not work | 06:47 |
fenn | imagine you're trying to cover a landscape in marbles, you can roll marbles into depressions in the landscape (GA) or glue them randomly, or glue a grid of marbles on the landscape | 06:48 |
fenn | what does SA stand for? | 06:49 |
kanzure | simulated annealing | 06:49 |
fenn | hrm. | 06:49 |
fenn | requires continuous changes to be good | 06:50 |
fenn | otherwise it might jump out of the hole | 06:50 |
fenn | or is that the point | 06:51 |
kanzure | need a little mutation to get evolution out of the gutters | 06:51 |
fenn | where's my magic quantum computer that can just give me the optimum answer | 06:52 |
fenn | hey, you know, in reality we want to support standards, which typically involve discrete choices ie M5 screw M6 screw etc | 06:52 |
fenn | so that reduces the search space drastically | 06:53 |
fenn | so i still dont know what "rules" are | 06:54 |
kanzure | "if you find this in the graph, do this" | 06:54 |
kanzure | http://heybryan.org/~bbishop/docs/rules/ check one of these files | 06:54 |
fenn | that's just a substitution system | 06:54 |
kanzure | yes | 06:54 |
fenn | that's as bad as cellular automata | 06:55 |
kanzure | (the files have a "left hand side" and a "right hand side") | 06:55 |
kanzure | eh? | 06:55 |
fenn | i.e. unpredictable output | 06:55 |
kanzure | rules are constrained by those standards though | 06:55 |
kanzure | that you just mentioned. | 06:55 |
kanzure | actually | 06:55 |
kanzure | it might just be an equivalency | 06:55 |
kanzure | they are standards for a reason | 06:58 |
fenn | most standards are proven designs | 06:58 |
fenn | but there is some arbitrariness, like what diameter is the screw | 06:58 |
kanzure | ok, so some rules are bendable | 06:58 |
* kanzure recalls scenes from the matrix | 06:58 | |
fenn | i mean why 5mm why not 5.1mm | 06:59 |
kanzure | yes, I know | 06:59 |
fenn | but the standard only defines 5mm screws | 06:59 |
kanzure | say you had to have 5.1 mm . that's a breakable rule methinks. | 06:59 |
kanzure | well that's probably a stupid standard anyway | 06:59 |
kanzure | size, jeesh | 06:59 |
fenn | however you can make a 5.1mm screw some of the rules for defining a 5mm screw (just change the diameter value) | 07:00 |
kanzure | lets do a more important thing in design space for standardization, I dunno, something that actually helps | 07:00 |
kanzure | 5 mm screws help of course, when buying parts | 07:00 |
kanzure | that's probably what it is. | 07:00 |
fenn | very much so | 07:00 |
fenn | used to be you had to call up the local blacksmith to make each screw since no two were alike | 07:00 |
kanzure | uh, how long ago? | 07:00 |
fenn | not that long ago :( | 07:01 |
fenn | 1910? | 07:01 |
kanzure | shouldn't have asked. | 07:01 |
fenn | optical comparator revolutionized screw making because it introduced some GD&T-like principles | 07:01 |
fenn | so a screw doesn't have to be exactly 5mm, as long as it falls within a certain range of shapes (optical comparator is because helical shapes are hard to measure with calipers) | 07:02 |
kanzure | right | 07:16 |
kanzure | fenn gets it yet? | 07:22 |
ybit | figured p.36 of http://web.mit.edu/endy/www/scraps/talks/03.06/EndyMarch2006.pdf might help gene | 07:26 |
ybit | maybe i will remember later | 07:26 |
ybit | and the 31m mark in http://video.google.com/videoplay?docid=-6950604815683841321&ei=li7fSO-PKZycrALHy9igCw&q=synthetic+biology+drew+endy | 07:28 |
kanzure | effn | 07:31 |
kanzure | don't get me wrong, substitution is a complementary method to everything else (like bruteforce) | 07:31 |
kanzure | but to some extent it's somewhat equivalent | 07:31 |
kanzure | I'd want to hash out the code for bruteforce first to see/show etc. | 07:31 |
kanzure | hm, weird typo on 'fenn' | 07:32 |
kanzure | why'd I type? | 07:32 |
UtopiahGHML | [tab] don't type | 07:33 |
kanzure | right | 07:33 |
UtopiahGHML | actually | 07:33 |
UtopiahGHML | comment on http://seedea.free.fr/wiki/pmwiki.php?n=Oimp.Ubiquitousvocabulary please | 07:34 |
kanzure | where'd my log analysis scripts go? | 07:35 |
kanzure | I had 100+ MB of analyzed logs from my many years of chatting | 07:35 |
kanzure | technically the information is still in my mysqldb :) | 07:35 |
* ybit takes a few capsules of melatonin and calls it a night | 07:36 | |
kanzure | dictionary based tab completion, eh? | 07:36 |
ybit | goodnight | 07:36 |
UtopiahGHML | night | 07:36 |
UtopiahGHML | kanzure: 100MB of raw data or of resulting analysis?! | 07:36 |
kanzure | raw data | 07:36 |
UtopiahGHML | that's different ;) | 07:36 |
kanzure | /me was working on auto grammar expansion stuff a while back (conceptually) .. would be nice to have computer assisted grammar / sentence construction for conveying sufficient meaning/detail | 07:37 |
kanzure | the analysis was like 300 MB or something | 07:37 |
kanzure | lots of stupid crazy overhead | 07:37 |
kanzure | timestamp on each and every word | 07:37 |
kanzure | (not connectivity though, why? mostly because I took the code from another guy) | 07:37 |
UtopiahGHML | well the point if to use [tab] all anytime anywhere instead of typing | 07:38 |
UtopiahGHML | s/if/is/ | 07:38 |
kanzure | have you seen the japanese text input interfaces for cell phones? | 07:38 |
kanzure | I hear it's kind of like tab completion except more gui | 07:38 |
UtopiahGHML | yep, chinese actually but I guess it's the same | 07:38 |
UtopiahGHML | Ive spend 6 months in Shanghai | 07:39 |
UtopiahGHML | didn't think about it actually, thanks :P | 07:39 |
UtopiahGHML | (duh) | 07:39 |
UtopiahGHML | they basically phonetically type and tab cycle amonst the different possibilities based on the word frequencies | 07:40 |
UtopiahGHML | sound very similar indeed | 07:41 |
kanzure | sentence construction help next please. | 07:41 |
UtopiahGHML | that's why I love to share ideas, you always have people looking at it with a different perspective | 07:41 |
kanzure | or people sufficiently annoyed with the problems. | 07:41 |
UtopiahGHML | they basically phonetically-type and tab-cycle amonst the different possibilities (possibilities ordererd by the frequency of usage of each word) | 07:43 |
kanzure | theoretically that's how standards should be crystalizing | 07:45 |
kanzure | but anyway. | 07:45 |
kanzure | sleep. g'night | 07:45 |
UtopiahGHML | ++ | 07:45 |
superturtle | uh | 15:56 |
superturtle | can anyone explain why precisely my laptop is off? | 15:56 |
UtopiahGHML | sounds like a job for the X-files guys | 15:59 |
superturtle | I suspect the fans. | 16:07 |
superturtle | 48585. This delay has set it back. | 16:19 |
kanzure | yep, no fan movement | 16:58 |
kanzure | http://forums.somethingawful.com/showthread.php?threadid=2960641 | 17:51 |
kanzure | somethingawful feels like the missing link in my internet ancestry | 17:51 |
kanzure | infopop brings me back. | 17:51 |
kanzure | *takes me back | 17:51 |
kanzure | which one is it? | 17:51 |
bkero | I'm a goon. | 17:52 |
kanzure | patrick sucks | 18:04 |
kanzure | on openmanufacturing | 18:04 |
superturtle | blah | 20:48 |
superturtle | would anyone be interested in accounts to the file server? | 22:21 |
bkero | files? | 22:45 |
superturtle | 300 GB collection of stuff. | 22:49 |
Generated by irclog2html.py 2.15.0.dev0 by Marius Gedminas - find it at mg.pov.lt!