| --- Day changed Sat Apr 05 2008 | ||
| mechie | i had a lot of pain working with the ROHS solder | 00:00 |
|---|---|---|
| kanzure | ROHS? | 00:00 |
| kanzure | not resin? | 00:00 |
| mechie | it requires such high heat and i kept cracking the resistors | 00:00 |
| kanzure | you know, we have automation machinery for a good reason | 00:00 |
| mechie | yeah | 00:00 |
| mechie | but i had to mod it | 00:00 |
| mechie | cuz our batch got messed up | 00:01 |
| mechie | so i was the cheap labor alternative :\ | 00:01 |
| mechie | i mostly went in for fixing things | 00:01 |
| kanzure | you're not too bright, are you | 00:01 |
| kanzure | joking :) | 00:01 |
| mechie | :D | 00:01 |
| mechie | i needed monies :> | 00:01 |
| kanzure | for what? | 00:01 |
| mechie | for school books | 00:02 |
| mechie | :D | 00:02 |
| kanzure | Download 'em. | 00:02 |
| mechie | i just get them on half.com or w/e | 00:02 |
| kanzure | (and leech) | 00:02 |
| mechie | soooo | 00:03 |
| mechie | do you make anything | 00:04 |
| mechie | physical | 00:04 |
| kanzure | DNA synthesizers, DNA sequencers, self-replicating machines, scanning probe microscopes, brains, bacteria, electronics, etc. | 00:05 |
| mechie | ok | 00:05 |
| kanzure | Artificial meat machines, | 00:05 |
| kanzure | artificial wombs | 00:05 |
| kanzure | cryonic suspension chambers | 00:05 |
| kanzure | many of these projects are tentative of course | 00:06 |
| kanzure | but it's good to have more ideas than time to implement them | 00:06 |
| kanzure | This way I can select which ones to do. ;) | 00:06 |
| mechie | i give up from the start :> | 00:06 |
| mechie | EPI KNOWS | 00:06 |
| * mechie kicks epitron in the nads | 00:07 | |
| kanzure | You give up on what? | 00:07 |
| kanzure | life? | 00:07 |
| mechie | maybe i did. | 00:08 |
| kanzure | You are cryptic. | 00:08 |
| mechie | perhaps. | 00:08 |
| mechie | i was thinking of being a surgeon | 00:08 |
| mechie | i didn't really want to work hard through med school though | 00:08 |
| kanzure | Med school is tanking. http://forums.studentdoctor.net/ <-- lots of morons getting into med school | 00:09 |
| mechie | yeah i know | 00:09 |
| mechie | my friends got in | 00:09 |
| * mechie scared | 00:09 | |
| kanzure | I am, however, interested in the medical scientist training program | 00:09 |
| kanzure | An 8 year program to get a PhD and an MD | 00:10 |
| kanzure | with a stipend paid by the govt | 00:10 |
| mechie | smartypants | 00:10 |
| epitron | kanzure: you could probably do that program a lot faster by yourself :) | 00:46 |
| epitron | working with researchers | 00:46 |
| epitron | with a stipend paid by the research labs | 00:46 |
| epitron | 8 years is too much school | 00:46 |
| kanzure | perhaps =) | 00:47 |
| kanzure | it's not constant schooling | 00:47 |
| epitron | toooo mucchhh schoooolll | 00:47 |
| epitron | did you know that the guy who sequenced the human genome didn't even go to school? | 00:47 |
| epitron | what's his name... craig ventnor? | 00:47 |
| epitron | (ok he didn't sequence the human genome by himself.. :) but he was instrumental in its expediency) | 00:49 |
| kanzure | he had millions of dollars | 00:50 |
| kanzure | billions, even | 00:50 |
| epitron | didn't the other genome project? | 00:51 |
| kanzure | huh? | 00:51 |
| kanzure | Craig Venter, by the way | 00:51 |
| kanzure | J. Craig Venter | 00:51 |
| epitron | thanks | 00:51 |
| epitron | there were two parallel projects | 00:51 |
| kanzure | http://venterinstitute.org/ or something | 00:51 |
| epitron | his at celera | 00:51 |
| kanzure | sure, there were many projects | 00:52 |
| kanzure | check out http://bioperl.org/ for the story | 00:52 |
| epitron | and the intergovernmental one or whatever | 00:52 |
| epitron | what? | 00:52 |
| epitron | :) | 00:52 |
| epitron | what's bioperl go tto do with it | 00:53 |
| kanzure | they have a good story on how automation saved the Human Genome Project. | 00:53 |
| kanzure | look around on the site, the article has a lengthy title | 00:53 |
| kanzure | and was written 1995 | 00:53 |
| epitron | so this bioperl project is old? | 00:53 |
| kanzure | actually, no | 00:53 |
| kanzure | but they have an old article or something | 00:53 |
| kanzure | Anyway, I'm going to go hit bed | 00:53 |
| epitron | i know about the basics, how venter used the shotgun approach | 00:54 |
| kanzure | I have ~30 papers open at the moment that I still need to run | 00:54 |
| epitron | nighty-o! | 00:54 |
| kanzure | shotgun technique is old news | 00:54 |
| epitron | hahah | 00:54 |
| kanzure | http://heybryan.org/mediawiki/index.php/DNA_sequencing | 00:54 |
| epitron | i know | 00:54 |
| kanzure | read that and digest all of the links | 00:54 |
| epitron | grr | 00:54 |
| epitron | :) | 00:54 |
| epitron | no! | 00:54 |
| epitron | i don't need to | 00:54 |
| epitron | semantic drift in progress | 00:54 |
| epitron | you ever read this quote? | 00:55 |
| epitron | "I consider that a man's brain originally is like a little empty attic, and you have to stock it with such furniture as you choose. A fool takes in all the lumber of every sort that he comes across, so that the knowledge which might be useful to him gets crowded out, or at best is jumbled up with a lot of other things, so that he has a difficulty in laying his hands upon it. Now the skilful workman is very careful indeed as to what he takes into his bra | 00:55 |
| epitron | "He will have nothing but the tools which may help him in doing his work, but of these he has a large assortment, and all in the most perfect order. It is a mistake to think that that little room has elastic walls and can distend to any extent. Depend upon it - there comes a time when for every addition of knowledge you forget something that you knew before. It is of the highest importance, therefore, not to have useless facts elbowing out the useful on | 00:55 |
| epitron | -- Sir Arthur Conan Doyle | 00:56 |
| epitron | i have no trouble reading a high level summary of something | 00:57 |
| epitron | so that i know of its existence, and can learn about it if the need arises | 00:57 |
| epitron | but i don't bother absorbing large random research papers for no reason :) | 00:57 |
| fenn | doyle obviously didnt know anything about data compression | 01:04 |
| epitron | shut up noob! | 01:32 |
| mechie | fuckers | 01:33 |
| mechie | go to sleep | 01:33 |
| kanzure | DNA looping? DNA-knots? | 10:07 |
| -!- Enki-2 [n=weechat@c-71-234-190-248.hsd1.ct.comcast.net] has quit [Read error: 104 (Connection reset by peer)] | 10:12 | |
| -!- Enki-2 [n=weechat@c-71-234-190-248.hsd1.ct.comcast.net] has joined #hplusroadmap | 10:17 | |
| fenn | dna underwater basket weaving | 10:57 |
| kanzure | that's what it sounds like | 11:00 |
| kanzure | fenn | 11:21 |
| kanzure | http://heybryan.org/mediawiki/index.php/Ellingtonia#Challenge | 11:21 |
| kanzure | see ED3814 | 11:21 |
| kanzure | the idea is that there are biochem-networks in the databases out there on the internet | 11:23 |
| kanzure | ideally, we can mimic the computation that those networks are doing | 11:24 |
| kanzure | and automatically generate our own 'logic' and then have the DNA compiler implement this with transcriptional logic or whatever | 11:24 |
| kanzure | but the problem is that the 'experimentally confirmed' biological reaction networks are messy and are defined in various ways, I think | 11:24 |
| kanzure | I need to go look at the raw data | 11:24 |
| kanzure | I suspect that translating that into 'abstract logic' is going to be hard | 11:25 |
| kanzure | unless we can model complex systems in a certain way, saying "there's an information conduit between these two places' | 11:25 |
| kanzure | a systems-diagramming method or something | 11:25 |
| kanzure | or I guess we can just ask to model the functionality of all of the components in the network, and then we can implement this in logical terms | 11:26 |
| kanzure | is this an impossibility? | 11:26 |
| fenn | well, looks like you've been doing your homework | 11:56 |
| fenn | genetic regulatory network is not logic exactly, more like fuzzy logic or analog electronic circuits | 12:03 |
| fenn | not sure what's hard about diagramming complex systems.. isnt that the whole point of making diagrams? | 12:06 |
| kanzure | I suppose, but can they be autodiagrammed ? | 15:27 |
| kanzure | I mean, what is the experimental data collection format ? if it's a complex network from automated testing or something, then we're in business | 15:28 |
| kanzure | because then we just do normal comp sci (parallel -> linear programs) | 15:28 |
| kanzure | which, although hard, is at least studied in theoretical comp sci | 15:28 |
| kanzure | fenn: http://www.genome.ad.jp/kegg/pathway/hsa/hsa04330.html | 15:49 |
| kanzure | hm | 15:49 |
| kanzure | they have a diagram, but not necessarily a format for reading this complex reaction network | 15:50 |
| kanzure | http://en.wikipedia.org/wiki/SBML | 15:52 |
| kanzure | http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B75GS-4BP87S9-D&_user=10&_rdoc=1&_fmt=&_orig=search&_sort=d&view=c&_acct=C000050221&_version=1&_urlVersion=0&_userid=10&md5=529aca035baa81bbfc382f3a52d8901a | 15:55 |
| kanzure | BioSig: an informatics framework for representing the physiological responses of living cells | 15:55 |
| kanzure | Subcellular experimental datasets and detailed cell models are required before modeling of whole organs. Cell modeling requires repeated interaction between simulation and experimental data. This review describes a coupled system of informatics and instrument control suitable for extracting information at the subcellular level. The BioSig informatics framework annotates time-series images with experimental variables and computed r | 15:56 |
| kanzure | http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1435936 | 15:58 |
| kanzure | GEM might be it. | 15:58 |
| kanzure | http://www.cellerator.org/ ... to facilitate biological modeling via automated equation generation | 16:07 |
| fenn | what's the point | 16:22 |
| fenn | epicycles | 16:22 |
| fenn | it's non-linear so your extrapolation beyond known data is probably going to be crap | 16:22 |
| kanzure | I guess I can think of this in another way. | 16:25 |
| kanzure | Suppose we have some differential equations that model a supercomputer, a network of Beowulf clusters or something. In our program, we can set the number of nodes on the network to N=2, and then try to map programming-instructions to those | 16:25 |
| kanzure | two nodes to model the behavior of the ODE. If this is not the case, then we iterate over to N=3, etc., up to N=X where X is some sufficiently large number that can have enough instructions to model the ODE. | 16:25 |
| kanzure | it doesn't matter if it's non-linear | 16:25 |
| kanzure | supposedly our supercomptuers are nonlinear ;) | 16:25 |
| kanzure | the fun thing is that your experimental data set is mostly 'complete' in bioinformatics | 16:27 |
| kanzure | since you know the genes and you can trace interactions and whatever else | 16:27 |
| kanzure | http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1893010 " In this paper we develop tools that enable the detection of steady states that are modeled by fixed points in discrete finite dynamical systems. We discuss two algebraic models, a univariate model and a multivariate model. We show that these two models are equivalent and that one can be converted to the other by means of a discrete Fourier transform." | 16:28 |
| kanzure | that seems to be my answer | 16:29 |
| fenn | i have no idea what the significance of that may be | 16:30 |
| kanzure | fixed points of finite dynamical systems -> finite field model -> Boolean model -> http://en.wikipedia.org/wiki/Boolean-valued_model | 16:38 |
| kanzure | if you look at the article, you'll see that this is comp sci language stuff | 16:38 |
| kanzure | meaning we have the abstract logic statements, the set of instructions basically | 16:38 |
| kanzure | and then we can plug this into my DNA compiler. | 16:38 |
| fenn | the 'fixed points' and 'finite dynamical systems' is just gibberish to me | 16:40 |
| fenn | i see they have a tree of states, i guess that's the finite dynamical system | 16:40 |
| fenn | what are the fixed points? | 16:40 |
| kanzure | I am guessing the genes or the proteins that they express | 16:40 |
| kanzure | but I don't know about it in a mathematical sense | 16:40 |
| kanzure | I'm first checking it out in #math on a meta-level | 16:41 |
| kanzure | to see if Boolean models can be symbol tables and given manipulations or whatever | 16:41 |
| kanzure | this sounds weird though | 16:43 |
| fenn | why not just start with the 'boolean valued model' | 16:43 |
| kanzure | "We can show that for any n-dimensional dynamical system ... there is an equivalent one-dimensional system." | 16:43 |
| kanzure | because you don't have that boolean valued model, you have some differential equations from cellerator ;) | 16:44 |
| kanzure | although tracing through the software perhaps we can find a way to simplify the process | 16:44 |
| fenn | so the point of all this is to re-create a genetic regulatory network, based on observed interactions? | 16:45 |
| kanzure | yes | 16:45 |
| kanzure | to extract the 'abstract logic' that evolution has come up with, in whatever fuzzy way that was | 16:46 |
| fenn | oh.. boolean valued model is only linear interactions, right? | 16:46 |
| fenn | p*q | 16:46 |
| kanzure | that's what it looks like | 16:46 |
| kanzure | but | 16:46 |
| kanzure | that's why the "We can show ..." line looks so weird | 16:46 |
| kanzure | it looks like the P=NP problem | 16:46 |
| fenn | not p^q or p+q | 16:46 |
| kanzure | or maybe I'm just seeing things | 16:46 |
| fenn | p=np is something else | 16:49 |
| kanzure | yes, they are hitting me over the head | 16:50 |
| kanzure | rather rudely | 16:50 |
| kanzure | I hate #math ... TWRBW is knowledgable, but a hardcore eugenicist/"if they can't do it, kill them" | 16:50 |
| fenn | fuck math | 16:50 |
| fenn | if you dont understand what something is, all the math in the world won't help | 16:51 |
| kanzure | math and chemistry have a lot in common, in my opinion | 16:51 |
| kanzure | while math you could supposedly do on your own, both of them require intense study of the previous literature and s oon | 16:51 |
| kanzure | actually, that's true of much of science | 16:52 |
| kanzure | just there it seems more stricking. | 16:52 |
| fenn | sorta | 16:52 |
| kanzure | hm, they also give state space diagrams | 16:52 |
| fenn | it's just hard to communicate if you dont share the language | 16:52 |
| fenn | and language is only got through study or proximity | 16:52 |
| kanzure | I think I've said this before [to you], but I would really appreciate a ref book on math syntax | 16:53 |
| fenn | heh no that was me complaining to you | 16:53 |
| kanzure | was it you who was complaining about identity disassociation in chatting? | 16:53 |
| kanzure | somebody who did it to you? | 16:53 |
| kanzure | or was this epitron | 16:53 |
| fenn | you suggested some old book and i downloaded it and never read it | 16:53 |
| fenn | that was me | 16:53 |
| fenn | i didnt say someone did it to me, it's just that i find myself thinking other peoples' thoughts sometimes | 16:54 |
| fenn | not necessarily a bad thing.. i think it's how communication works | 16:54 |
| kanzure | then it was somebody else who mentioned a bad experience with somebody who started to mistake him/her for a long-gone ex-girlfriend or something | 16:55 |
| kanzure | Anyway, | 16:55 |
| kanzure | since this is *not* some lunatic thinking P=NP | 16:56 |
| kanzure | and my warning lights are not going off | 16:56 |
| fenn | oh, yea that is me too | 16:56 |
| fenn | i dont get the p=np problem, why is it a question? | 16:56 |
| kanzure | it looks like I might have a vague picture from automated bioinformatics all the way down to -> automated modeling and then -> automated compiling into DNA for in vitro synthetic biology, to create systems from scratch. | 16:56 |
| kanzure | by wrestling the insights from biology (and thus evolution) | 16:57 |
| kanzure | so if we could just cycle through an entire database of the experimental bioinformation, we can supposedly generate all of the code to the cell, even if it is sorta fuzzy | 16:57 |
| kanzure | although this sort of assumes we have ODEs or enough experimental data for all of the genome, so that we can do differential equations for all of the interacting genes | 16:58 |
| kanzure | it sounds unlikely to currently exist in an easily accessible format, but it should be possible | 16:58 |
| fenn | no, because you neglect actual interaction with the non-nucleic acid world | 16:58 |
| kanzure | right | 16:58 |
| kanzure | well, no | 16:58 |
| kanzure | remember the Winfree paper on robustness | 16:58 |
| fenn | you dont have the heuristics for protein folding | 16:58 |
| kanzure | while there's a certain point where the cell breaks down and can't do the cascading chemical reactions | 16:59 |
| kanzure | in general, for a normal state, they are going to work | 16:59 |
| kanzure | you don't need protein folding | 16:59 |
| kanzure | not by a long shot | 16:59 |
| kanzure | the point is that we can model whatever 'normal' is | 16:59 |
| kanzure | biology has adapted for those variety of circumstances when the environment (internal/external) gets rough | 16:59 |
| kanzure | but that's more like filling in the gaps later | 16:59 |
| kanzure | and tweaking the system to be able to survive in more situations | 17:00 |
| kanzure | rather than being perfect upon the first go | 17:00 |
| kanzure | (being perfect would involve simply replicating the DNA and doing protein biosynthesis like normal ;)) | 17:00 |
| fenn | the point of the regulatory networks is to allow normal function even in harsh environments | 17:01 |
| fenn | if everything is perfect, you dont need regulatory networks, you just hard-code everything | 17:01 |
| kanzure | that's all at the edges though | 17:01 |
| kanzure | just before the 'automated bioinformatics experimentation' stuff | 17:01 |
| kanzure | that's where people would come up and set up a new automated experiment for a new class of genes or whatever | 17:01 |
| kanzure | and so they would change the variables that are being tweaked and seeing how the genes react | 17:01 |
| kanzure | I see what you mean, but I think you are assuming something that I am not | 17:02 |
| kanzure | we'd be using this system for another purpose than modeling the cell per-se | 17:02 |
| kanzure | suppose we run a selection experiment on bacteria | 17:02 |
| kanzure | and they come up with a novel solution to some challenge, which would necessarily involve genetics | 17:02 |
| kanzure | so we could characterize these new genes and they way they interact with each other -- perhaps they are a modification in the signaling network, or a new chemical synthesis, something | 17:03 |
| kanzure | we would know where to look since we can sequence their genomes and look for the modifications, and other various DNA tech | 17:03 |
| kanzure | we would know what conditions they were in since we were the ones who set up the selection experiment | 17:03 |
| kanzure | it is these solutions to information-processing in whatever 'harsh' (or not) conditions we set up that we are interested in | 17:04 |
| fenn | you mean adaptation? | 17:04 |
| kanzure | granted, sometimes they may not be GRNs (gene regulatory networks), but we can do some interesting things to make sure their adaptations/mutations do this | 17:04 |
| kanzure | yeah | 17:04 |
| kanzure | this is a way to codify knowledge gathered by the evolutionary process of adaptation, into a form that humans can begin to understand | 17:04 |
| kanzure | and then we can use this in, say, skdb, but that's another set of mental connections to make | 17:05 |
| kanzure | the designs are abstract and do not need to be plugged into the DNA compiler | 17:05 |
| kanzure | they could be plugged into anything, frankly | 17:05 |
| kanzure | even gcc | 17:05 |
| kanzure | (as long as we have a way to convert to C, of course) | 17:05 |
| fenn | more like vhdl i think | 17:05 |
| kanzure | but that conversion is easy, from a set of Boolean statements to C? yeah, we can do VHDL to C I think | 17:06 |
| kanzure | there's got to be a VHDL->C translator out there, heh | 17:06 |
| kanzure | that's just the geeky thing to do | 17:06 |
| fenn | just so you can use gcc? sounds silly | 17:06 |
| kanzure | well, | 17:06 |
| kanzure | biology is software, on the hardware of reality | 17:06 |
| kanzure | and evolution is a way of writing up new programs | 17:06 |
| fenn | vhdl is digital logic anyway, it wouldnt work for your wacky analog interactions | 17:06 |
| kanzure | so these new programs can be converted into software for our silicon computers | 17:06 |
| kanzure | that's my point | 17:06 |
| kanzure | no, this is discrete/finite points in field models, remember? (from the GRN dataset analysis, the pubmed paper) | 17:07 |
| fenn | no i have nfc what that was about | 17:07 |
| fenn | (i did read it) | 17:07 |
| kanzure | it was a way of taking experimental data from bioinformatics, the automated gene analysis systems and so on, and then coming up with finite models that do the same things | 17:07 |
| kanzure | yep | 17:07 |
| kanzure | so anyway, my point again: nature evolves software; we translate that software into abstractions, we then can take those abstractions and test them, either in silicon or in vitro with transcriptional logic to | 17:08 |
| kanzure | build up synthetic 'biological' systems (not necessarily cells ... but I hear Andy Ellington has an interest in nucleic acid 'origins of life' stuff) | 17:08 |
| fenn | so this is like checking your answer by doing the problem a different way? | 17:09 |
| kanzure | please contextualize | 17:09 |
| fenn | in math | 17:09 |
| fenn | two trains leave chicago and new york | 17:09 |
| kanzure | I don't think so, I probably just don't know your question | 17:09 |
| kanzure | let me reiterate and you can point out your questions | 17:10 |
| fenn | the problem is 'how does this cell work' | 17:10 |
| kanzure | biology and natural evolution comes up with new solutions to problem | 17:10 |
| kanzure | nope | 17:10 |
| kanzure | the problem is 'how to make it better' or 'how to get something interesting out of it' | 17:10 |
| kanzure | cells are really good at two things in natural evolution | 17:10 |
| kanzure | (1) genetic recombination (sex) | 17:10 |
| kanzure | (2) mutation | 17:10 |
| kanzure | in selection experiments, cells use both of those to come up with solutions | 17:10 |
| kanzure | and then write them on DNA | 17:10 |
| kanzure | these 'solutions' are quite fuzzy | 17:11 |
| kanzure | and rely on molecular physics among other things | 17:11 |
| fenn | your solution is modifying the DNA | 17:11 |
| kanzure | but ultimately they are, in fact, computing | 17:11 |
| kanzure | nope | 17:11 |
| kanzure | hold on | 17:11 |
| fenn | and by sheer chance the modifications end up with a solution | 17:11 |
| kanzure | sort of, yes | 17:11 |
| kanzure | they are in fact computing, so if we can then find out what they are doing that lets them survive the selection experiment | 17:11 |
| fenn | it doesnt happen the other way around (as much as i would like to believe) | 17:11 |
| kanzure | yeah :( | 17:11 |
| kanzure | "as much as I would like to believe" | 17:12 |
| kanzure | that's the understatement of the century | 17:12 |
| fenn | i'm a rupert-sheldrake sympathizer | 17:12 |
| kanzure | how so? | 17:12 |
| kanzure | sheldrake the morphogenetic field theorist? | 17:12 |
| fenn | i think spirits affect random chance and that's where life came from | 17:12 |
| kanzure | I have trouble believing in randomness | 17:12 |
| kanzure | but I do understand that you don't have all of the information | 17:13 |
| fenn | meh | 17:13 |
| kanzure | how can randomness ever exist, anyway? | 17:13 |
| kanzure | for something to be truly random, | 17:13 |
| kanzure | in my program it would have to return 'COW' or 'thing that looks like a cow, but is really an alien from 23 dimensions over' | 17:13 |
| kanzure | but instead it just predictably spits out a number | 17:13 |
| fenn | i didnt mean to go off on this tangent | 17:13 |
| kanzure | sure | 17:13 |
| kanzure | anyway, | 17:13 |
| kanzure | the idea is to convert their programs over into a form that we can understand | 17:14 |
| fenn | 'how does this cell work' | 17:14 |
| kanzure | by leveraging the past 60 years of bio research to convert their 'physical language' into our abstract languages, and then we get to play around with the new components we get and see what we can do with them | 17:14 |
| kanzure | yeah, I guess that's a good way to put it | 17:14 |
| kanzure | but you don't have to know everything about the cell | 17:14 |
| kanzure | there are ways to constrain selection experiments I think | 17:14 |
| kanzure | and Andy succeeded in coming up with automatic ways of evolving nucleic acids without cells via aptamers | 17:15 |
| kanzure | and other various biomolecules, I think | 17:15 |
| kanzure | http://en.wikipedia.org/wiki/Aptamers (be careful, this is also somewhat of a tangent) | 17:15 |
| kanzure | there's a limit to evolvable software, but on the other hand, it should be possible to come up with lots of interesting programs on the size of genomes ... some bacteria have genomes that are 5 to 500 times larger than the human genome, so imagine how many programs can be written and how many mutations can occur | 17:17 |
| fenn | you dont even need a physical biological substrate to do this.. say you have a genetic algorithm, then you model it with formal models in order to abstract/extract the 'how' | 17:20 |
| kanzure | true, | 17:21 |
| kanzure | this is why selection experiments (whether simulated or physical) are hard to design | 17:21 |
| kanzure | a few days ago I made the comparison of that to making ... 'shrines' for gods | 17:21 |
| fenn | generally it just finds flaws in your selection function though | 17:21 |
| kanzure | because that's basically the same thing | 17:21 |
| kanzure | people were making shrines and when they did this they got new, novel mutated ideas | 17:21 |
| kanzure | and so they thought something good must be going on | 17:22 |
| kanzure | so they continued to make them | 17:22 |
| kanzure | these 'shrines' are the selection experiments, but obviously in today's world we have gotten much more complex with our selections that we are making | 17:22 |
| kanzure | haha, yes, flaws | 17:22 |
| kanzure | suggesting your problem might suck | 17:22 |
| fenn | why did making shrines give them novel ideas? because they're big and hard to build? | 17:22 |
| kanzure | generally, I'd refer to the process of everything about it - what it takes to make it happen, what it's like to be in the shrine or around the shrine or whatever | 17:23 |
| kanzure | http://en.wikipedia.org/wiki/Koans | 17:23 |
| kanzure | it's like 'preparing to receive an insight' -- just like Feynman prepared by reading as much as he could for QED | 17:23 |
| kanzure | as much as we'd like to do insight engineering, of course ... | 17:23 |
| fenn | i think you might be working your brain too hard :) | 17:23 |
| kanzure | the problem with that is that it still involves 'engineering' | 17:24 |
| kanzure | nah, I'm pretty sure I am solid | 17:24 |
| kanzure | that's why I have autoscholar, to prepare for insights or whatever | 17:24 |
| kanzure | http://heybryan.org/thinking.html re: thinking/insights/incubation-theory | 17:24 |
| kanzure | http://heybryan.org/projects/autoscholar/ - to facilitate the consumption of raw information in preparation for insights or whatever | 17:24 |
| kanzure | http://heybryan.org/increasing_repetitive_behaviors.html for more autoscholar-like stuff, except more mental and oriented towards mental modification | 17:25 |
| kanzure | there's a theme for a reason ;) | 17:25 |
| fenn | why does building a shrine cause novel ideas? | 17:26 |
| kanzure | it's permutation through possibility space, it's not the fact that the shrine 'receives godly communications' - nothing like that - but rather the opportunistic-assimilation hypothesis of incubation theory | 17:28 |
| kanzure | Google has good results for 'opportunistic-assimilation hypothesis' | 17:29 |
| kanzure | http://www.lifesci.sussex.ac.uk/home/Zoltan_Dienes/Seabrook%20&%20Dienes%2003.pdf Incubation in Problem Solving as a Context Effect | 17:29 |
| kanzure | (but find the occurence of the phrase within that article, you can jump to it without losing much context) | 17:30 |
| fenn | i think you'd have a hard time convincing a molecular biologist that this is how mutations work | 17:30 |
| kanzure | huh? | 17:30 |
| kanzure | I am not saying this is the physics of mutations | 17:30 |
| fenn | ok | 17:31 |
| kanzure | shrine-making (selection-experiment making) does not make the mutations directly, but the cells are the ones that come up with the mutations | 17:31 |
| kanzure | because however they do it, the ones that mutate or get the right recombined DNA, are the ones that survive the selection | 17:31 |
| fenn | where is the selection in shrine-making? | 17:31 |
| kanzure | your brain; there's another theory in incubation theory (or perhaps it's the same one) where you keep on working on the problem and literally forget how you were approaching it, so that when you come across the problem again, you get the insight to the solution, so when you make a shrine you're exhausting the contexts | 17:33 |
| kanzure | (end of line is "so when you make a shrine you're exhausting the contexts") | 17:33 |
| kanzure | your brain is a massive billion-something network, 'mutation' in firing patterns can easily occur (it's not solid state, after all) | 17:34 |
| kanzure | mutation in DNA is another topic, but I think you have studied that one sufficiently (if you were going to have me complete my analogy I made above to the shrine-making and selection-experiment thing) | 17:34 |
| kanzure | neurons are reinforced when they do something 'good' (when they contribute to a solution, i.e. various feedback mechanisms), same with bacteria (when they survive the selection) | 17:35 |
| kanzure | anyway, I'm not trying to be pseudoscientific | 17:35 |
| kanzure | but maybe I sound that way since I was approaching this as a bit of a tangent to the 'software from biology' idea | 17:36 |
| fenn | i'm fine with pseudoscience as long as it's productive | 17:36 |
| fenn | but this doesn't sound productive (maybe i'm wrong) | 17:36 |
| fenn | tesla, fuller | 17:37 |
| fenn | ^^pseudoscience | 17:37 |
| kanzure | haha | 17:37 |
| fenn | it's true | 17:37 |
| kanzure | I think you are wrong, actually. This is the same exact problem space as skdb. | 17:37 |
| kanzure | the problem of 'engineering knowledge' | 17:37 |
| kanzure | you can't just derive it from first principles | 17:37 |
| kanzure | it has to be 'evolved' | 17:37 |
| kanzure | whether by society and trial-and-error by humans (the lab minions) (yikes) or by bacteria | 17:38 |
| kanzure | and then we collect it and aggregate it into skdb | 17:38 |
| fenn | i think with a good enough simulator you could build up skdb from scratch | 17:38 |
| fenn | might take a while | 17:38 |
| fenn | brute force vs analytical thinking | 17:38 |
| kanzure | it's neither | 17:39 |
| kanzure | first, I want to address your simulation hypothesis | 17:39 |
| kanzure | you'd have to be able to simulate the world, basically | 17:39 |
| kanzure | which means a lot of physics and a lot of computational power | 17:39 |
| fenn | yes | 17:39 |
| kanzure | alright, so | 17:39 |
| kanzure | do you know Godel? | 17:39 |
| fenn | uh huh | 17:39 |
| kanzure | uhh, I just got interrupted, one moment | 17:40 |
| kanzure | you can't simulate the physics of the overall system without something bigger | 17:40 |
| kanzure | or you might as well just run the planet | 17:40 |
| kanzure | i.e., like we are now | 17:40 |
| kanzure | and the abstract programs that we do collect anyway, they are generalized in the first place (hurray) | 17:40 |
| kanzure | so do we really need a simulation? | 17:40 |
| kanzure | brb | 17:40 |
| fenn | there is a middle ground - heuristics | 17:41 |
| kanzure | from where? ;) | 17:41 |
| fenn | and automated heuristic creation, like from bayesian analysis | 17:41 |
| kanzure | hm | 17:41 |
| fenn | it's not abstract, but it's not totally blind either | 17:41 |
| kanzure | please explain | 17:41 |
| fenn | say you run 100 simulations | 17:41 |
| fenn | in all 12 of the simulations with carrots, you ended up with more rabbits at the end | 17:42 |
| fenn | heuristic: carrot -> rabbits increase | 17:42 |
| fenn | uh, ok i suck at explaining bayesian reasoning | 17:43 |
| kanzure | Jef tells me I need to learn more Bayes | 17:43 |
| kanzure | and I agree with him | 17:43 |
| kanzure | I don't see the possibilities | 17:43 |
| fenn | you can do experiments in the simulation too | 17:44 |
| fenn | to test hypotheses | 17:44 |
| fenn | like, add more carrots and see if you get more rabbits | 17:44 |
| fenn | if you get more rabbits it will strengthen your faith in that hypothesis | 17:44 |
| fenn | confidence i guess is a technical term in statistics | 17:44 |
| fenn | if the simulated world is totally chaotic and acausal, you'll end up with a spread of hypotheses with near zero confidence | 17:45 |
| kanzure | heh, my dad came in to read me a passage from 1901 - "It is in my complete and utter professional opinion, that because the power requirements scale cubically, and the wingspan requirements doubly, that aerial flight will thus remain impossible for the remainder of time. Stop bugging me." Later that year ... | 17:46 |
| kanzure | hm | 17:47 |
| kanzure | well, I don't really see what bayesian analysis can do to help | 17:47 |
| fenn | i wonder why they dismissed flight so easily when there were already kites | 17:47 |
| kanzure | my method involves automated hypothesis generation, does yours? | 17:47 |
| fenn | yes | 17:47 |
| kanzure | if I was going to come up with an integrative framework where I have my automated machinery setting up new selection experiments, sequencing the DNA and translating it into abstract logic | 17:49 |
| kanzure | then you'd plug in bayesian analysis systems straight into the big pot o' algebra? | 17:49 |
| kanzure | not algebra, sorry | 17:49 |
| kanzure | big pot o' acquired programs | 17:49 |
| fenn | you have to be able to modify something in order to test the hypothesis | 17:50 |
| kanzure | how much computing hardware would be required | 17:50 |
| fenn | ugh i dont know | 17:51 |
| fenn | or i suppose if there's too much data to sift through you could use the search keys for your test | 17:52 |
| kanzure | how so? | 17:53 |
| fenn | predict levels of foo increase when bar is present; search for experiments with bar, see if foo increases vs control experiments without bar | 17:53 |
| fenn | this is the topic of many a research paper | 17:54 |
| fenn | but done by grad students, not a program | 17:54 |
| kanzure | hrm | 17:55 |
| fenn | this is all a lot easier when you say it in english of course | 17:56 |
| kanzure | "In vitro modeling of in vivo systems" <-- this seems to be a good heading for what I have been talking to you about | 17:56 |
| fenn | in vitro just means in a test tube, not in a cell | 17:57 |
| fenn | knowledge representation seems to have been a big theme | 17:59 |
| kanzure | yes | 18:00 |
| kanzure | it is in a test tube though | 18:00 |
| kanzure | that's the idea | 18:00 |
| kanzure | we take in vivo circuits from cells, and then figure out new ways to run the same circuits, codify them into transcriptional logic, and run them in vitro | 18:00 |
| fenn | i dont think that would work, since 99% of the GRN's out there are relating to keeping the cell functioning | 18:02 |
| kanzure | haha | 18:03 |
| kanzure | think about it, how are they doing it? | 18:03 |
| kanzure | they are computing with biomolecules too | 18:03 |
| kanzure | they are making decisions, but not only that | 18:03 |
| kanzure | but the structure of the 'general network' that represents what they are doing to stay alive | 18:03 |
| fenn | not decisions, feedback | 18:03 |
| kanzure | is of a certain class, a certian computation | 18:03 |
| kanzure | yes, feedback is computational | 18:03 |
| fenn | decisions implies modeling | 18:03 |
| kanzure | my point is that there's information routes in the cell, they are information processing entities | 18:04 |
| kanzure | and sometimes feedback is used, yes, and these various networks for feedback or whatever, can be translated into code for our reading | 18:04 |
| kanzure | not DNA code, but logic code stuff | 18:04 |
| kanzure | fuzzy as it may be | 18:04 |
| fenn | i think circuits or control system diagrams are more appropriate than code | 18:05 |
| kanzure | circuits | 18:05 |
| kanzure | huh | 18:05 |
| kanzure | VHDL? | 18:05 |
| kanzure | oh, right, VHDL -> circuits | 18:05 |
| fenn | no, vhdl is not really circuits | 18:05 |
| kanzure | VHDL can be converted into circuits | 18:05 |
| kanzure | this is 'network decomposition' (circuits -> VHDL or some other language) | 18:06 |
| fenn | yes, but the idea of vhdl is to make modular functional units | 18:06 |
| fenn | whereas a circuit could be a big mess of interconnections | 18:06 |
| kanzure | and is exactly related to Hamiltonian path finding, i.e. breaking down graphs into their cycles | 18:06 |
| kanzure | I don't see how that's a problem. | 18:06 |
| kanzure | but I understand what you mean | 18:06 |
| fenn | vhdl coding for a messy circuit would be harder to understand than just a messy circuit diagram | 18:07 |
| kanzure | there are some simple circuits in biology though | 18:07 |
| kanzure | for example, the ring oscillator | 18:07 |
| kanzure | which is basically A -> B -> C -> A | 18:07 |
| fenn | biology tends to be messy (there are examples of modular systems too) | 18:07 |
| kanzure | remember the paper I linked you to? there are ways to find the 'finite points' | 18:08 |
| kanzure | or what did the paper call them? | 18:08 |
| kanzure | fixed points | 18:08 |
| fenn | what's a fixed point? | 18:08 |
| kanzure | in a dynamical system, a fixed point is supposedly something that you can use to 'unravel' the dynamical system | 18:09 |
| * fenn pictures a force-directed graph layout algorithm | 18:10 | |
| fenn | sproing | 18:10 |
| kanzure | sproing? | 18:10 |
| fenn | its like the VSEPR model | 18:10 |
| fenn | but with graph nodes rather than electron orbitals | 18:10 |
| kanzure | oh, a way to minimize energy density or something | 18:10 |
| fenn | yeah | 18:11 |
| fenn | minimum energy diagram layout | 18:11 |
| fenn | remember this? http://en.wikibooks.org/wiki/Circuit_Idea/Why_Circuit_Ideas_are_Hidden | 18:14 |
| kanzure | yep | 18:15 |
| kanzure | social knowledge, gathered from mutation/insight | 18:15 |
| fenn | specifically, you're trying to un-hide the ideas hidden in bacterial mutations | 18:15 |
| kanzure | yes | 18:16 |
| fenn | a knowledge synthesizer! | 18:17 |
| kanzure | hehe | 18:17 |
| kanzure | accelerated, automated evolutionary engineering | 18:17 |
| fenn | well just be glad you arent some poor sod computer programmer being given this non-specifcation | 18:18 |
| kanzure | well, not quite a knowledge-synthesizer actually - it could gather knowledge and make up proofs and theorems involving new 'programs' that it unhides, but it's like skdb, you have to randomly try | 18:19 |
| kanzure | out new variations in some cases, while restricting your domain (the same stuff we're doing with the computational chem programs for the replicator) | 18:19 |
| kanzure | nah, the specification is going to be pretty easy | 18:19 |
| kanzure | since the mathematics mentioned in that paper has a good, strong history of understanding | 18:19 |
| kanzure | and should have some computer programs out there on the net already | 18:19 |
| kanzure | probably open source | 18:19 |
| kanzure | and then the bioinformatics machinery, well, that's fairly easy I think, what could they be doing in the labs to gather genetic information? And Andy has the automated aptamer stuff :) | 18:19 |
| kanzure | automated selection experiments, I mean. | 18:20 |
| kanzure | to answer Ellington's challenge I can just go into the databases and find some ODEs that model a particular GRN, something other than an oscillator, and just do it by hand myself | 18:22 |
| kanzure | and then if he approves of that, I'll suggest we just automate the whole thing while we're at it | 18:22 |
| fenn | 'do it by hand' meaning what exactly? | 18:23 |
| fenn | put together some bio-bricks? | 18:23 |
| kanzure | nah, he doesn't work with biobricks, thinks they're bullshit | 18:23 |
| kanzure | I mean the transcriptional switches/logic | 18:23 |
| fenn | well, whatever you want to call it | 18:23 |
| kanzure | he wants an in vitro synthetic circuit | 18:23 |
| * fenn thinks they're the same thing | 18:23 | |
| kanzure | nah, they're different, but the concept of modularizing things is same across the domain | 18:23 |
| fenn | ok, splice some genes together that behave similarly to the target circuit | 18:24 |
| kanzure | heh, and you mentioned protein folding earlier | 18:24 |
| kanzure | so this is where you'd need protein folding | 18:24 |
| kanzure | unless you want to just copy the genes | 18:24 |
| kanzure | otherwise you have to do a massive computational search for similar proteins | 18:25 |
| kanzure | the idea was to model the basis of the circuit, the 'computational' aspects | 18:25 |
| kanzure | anything else can be considered waste for now | 18:25 |
| fenn | you could do a massive in-vitro search for similar proteins (artificial proteins) | 18:25 |
| kanzure | you'd have to synthesize tons of oligonucleotides and then do transcription to | 18:25 |
| kanzure | yeah, sure | 18:25 |
| kanzure | that would work | 18:25 |
| kanzure | but you'd have to separate them I think, right? | 18:26 |
| fenn | you know about monoclonal antibodies? | 18:26 |
| kanzure | similarities could be found via DNA hybridization and conformational stuff | 18:26 |
| kanzure | no | 18:26 |
| fenn | each b-cell produces a unique random-shuffled protein | 18:26 |
| kanzure | but you'd have to make sure the proteins do not interact with each other, i.e. the set of proteins that you are testing to be similiar to the target protein | 18:26 |
| kanzure | I think I had this same problem in another situation that I was thinking about a few months ago | 18:27 |
| kanzure | I remember it. | 18:27 |
| kanzure | it was the 2^(4^(N)) experimentation | 18:27 |
| fenn | you can find the cell that makes, say, lactose-binding protein, and figure out the amino acid sequence | 18:27 |
| kanzure | with searching for transcriptional switches and the toeholds and matches | 18:27 |
| fenn | but you can find 50 or a zillion other different lactose binding antibody proteins | 18:28 |
| kanzure | hm | 18:28 |
| fenn | and isolate the cells for each of those | 18:28 |
| kanzure | but think of it this way | 18:28 |
| kanzure | say you have 50 million different proteins that you are matching to aptamers | 18:28 |
| kanzure | to check if they will be anything like the one that you are targetting | 18:28 |
| kanzure | the subset of 50 million that do not interact with each other | 18:28 |
| kanzure | is not necessarily the subset with the best match that you want ;) | 18:28 |
| kanzure | because everything else doesn't bind to the aptamers | 18:29 |
| kanzure | I mean, anything that binds to each other doesn't bind to the aptamers (too large, energies, ...) | 18:29 |
| kanzure | and you wash away the solution above the aptamers | 18:29 |
| kanzure | and then unbind the aptamer-protein complex, and then figure out what proteins you have left, which would be your solutions that are similar to the targets | 18:29 |
| kanzure | (I believe this has some relation to 'library testing', but I haven't explored that much, is this what the libraries are for? oligo sequences?) | 18:29 |
| fenn | you could use 50million^2 test tubes, for every combination | 18:30 |
| kanzure | ... | 18:30 |
| fenn | inkjet printers! | 18:30 |
| kanzure | to make test tubes? heh' | 18:30 |
| fenn | yeah just drops of water on a hydrophobic surface | 18:30 |
| fenn | with a spot of hydrophilic substrate to keep the dot still | 18:30 |
| kanzure | not bad | 18:30 |
| kanzure | but, what about the synthesis of the protein? | 18:31 |
| fenn | i forget what we're trying to do | 18:31 |
| kanzure | finding proteins that match a target circuit | 18:32 |
| kanzure | because I need to make an in vitro synthetic circuit that matches a target circuit found in the literature (of natural organisms) | 18:32 |
| fenn | finding proteins that bind to the same sites as other proteins | 18:32 |
| kanzure | same sites of what? | 18:32 |
| kanzure | I thought you were talking about a protein cascading chemical reaction network | 18:32 |
| kanzure | isntead of a GRN | 18:32 |
| fenn | well, it could bind to some arbitrary chemical | 18:33 |
| kanzure | meh, I like my DNA compiler better | 18:33 |
| fenn | there's also enzymatic action | 18:33 |
| kanzure | yes, but I think he wants me to work with transcriptional circuits | 18:34 |
| kanzure | which are much easier to work with in vitro | 18:34 |
| fenn | i dont think you're going to brute-force an enzyme | 18:34 |
| kanzure | since you can ignore transcription | 18:34 |
| kanzure | he's done it | 18:34 |
| kanzure | riboenzymes | 18:34 |
| kanzure | but I don't think it's necessary here | 18:34 |
| kanzure | we could if we wanted, but meh | 18:34 |
| kanzure | anyway, pizza is here | 18:34 |
| fenn | still seems unlikely | 18:34 |
| kanzure | so I'm going to go eat | 18:34 |
| kanzure | http://en.wikipedia.org/wiki/Ribozyme I think | 18:34 |
| fenn | i know all about them | 18:35 |
| kanzure | oh | 18:35 |
| kanzure | http://en.wikipedia.org/wiki/Riboswitch | 18:35 |
| kanzure | In molecular biology, a riboswitch is a part of an mRNA molecule that can directly bind a small target molecule, and whose binding of the target affects the gene's activity [1][2][3]. Thus, an mRNA that contains a riboswitch is directly involved in regulating its own activity, depending on the presence or absence of its target molecule. | 18:35 |
| fenn | cool | 18:35 |
| fenn | ok so forget protein | 18:36 |
| fenn | this is cool http://en.wikipedia.org/wiki/Spiegelman_Monster | 18:43 |
| fenn | oh i misunderstood it. i thought the RNA sequence was also a polymerase ribozyme | 18:44 |
| fenn | can you recommend some light reading on P/NP hard complete etc | 18:59 |
| fenn | wikipedia provides zero examples for me to grab onto | 19:00 |
| kanzure | no, I can't | 19:04 |
| kanzure | I suffer from the same problem | 19:04 |
| kanzure | however, David might be able to come in here and explain | 19:04 |
| kanzure | or maybe Ryan Patterson | 19:04 |
| fenn | basically it sounds like a bunch of bullshit and i'm wondering why people spend time worrying about it | 19:05 |
| kanzure | it's for figuring out which problems can and cannot be solved | 19:05 |
| fenn | what is "non-polynomial time"? | 19:06 |
| fenn | i can understand polynomial time | 19:06 |
| fenn | is it just.. everything else? | 19:07 |
| fenn | seems like a totally useless way of describing something | 19:07 |
| kanzure | okay, David should be appearing | 19:08 |
| fenn | uh oh | 19:08 |
| kanzure | why's that bad? | 19:08 |
| fenn | i'm so not even awaky still | 19:08 |
| kanzure | "access to channel #hplusroadmap is blocked"? | 19:09 |
| fenn | aroo? | 19:09 |
| kanzure | why is he on a web client | 19:10 |
| fenn | well i think i'm going to bed soon (now) | 19:11 |
| kanzure | uhh | 19:11 |
| kanzure | please don't | 19:11 |
| kanzure | he's going to appear :( | 19:11 |
| fenn | ok ok | 19:11 |
| kanzure | kind of like a guest lecturer :) | 19:11 |
| fenn | i will be the guy asleep in the first row | 19:11 |
| fenn | dont worry i'll crib the notes from someone after class | 19:12 |
| -!- davidad_ [n=me@DAVIDAD.MIT.EDU] has joined #hplusroadmap | 19:14 | |
| davidad_ | kanzure: victory | 19:14 |
| kanzure | Victory! Ha-ha! | 19:14 |
| davidad_ | the solution is to ssh to my linux box | 19:14 |
| * fenn waves the anarcho-penguin flag | 19:14 | |
| kanzure | where'd you get davidad.mit.edu ? | 19:14 |
| davidad_ | um, from MIT? | 19:15 |
| kanzure | impressive | 19:15 |
| kanzure | you just asked? | 19:15 |
| davidad_ | they give out three free hostnames to every student | 19:15 |
| kanzure | so, let me explain what's been up today in #hplusroadmap | 19:15 |
| davidad_ | ok | 19:15 |
| fenn | we talked past each other four about 3 hours | 19:16 |
| kanzure | I've been working on some Andy Ellington stuff for biology, figuring out what software would be interesting to implement with biomolecules as the substrate | 19:16 |
| kanzure | and somehow we got to talking about N=NP and computational complexity classes | 19:16 |
| davidad_ | that's P=NP >_> | 19:16 |
| kanzure | djlkafjlkadjfladkjf;a | 19:16 |
| kanzure | P=NP | 19:16 |
| kanzure | in the context of biological computation (not computation about biology, but on biology) | 19:16 |
| kanzure | but in general as well | 19:16 |
| davidad_ | ah | 19:16 |
| kanzure | fenn, I think, claims that P=NP isn't much of a question/problem | 19:17 |
| davidad_ | yeah, some people think DNA computing, like quantum, does NP | 19:17 |
| davidad_ | some people think neither can | 19:17 |
| fenn | i dont know what NP is | 19:17 |
| davidad_ | and some people think P=NP | 19:17 |
| davidad_ | basically | 19:17 |
| davidad_ | P is all the problems you can solve in polynomial time, | 19:17 |
| davidad_ | and NP is all the problems you can verify in polynomial time | 19:17 |
| kanzure | where polynomial is the degree of the equations? | 19:17 |
| kanzure | degree meaning exponents | 19:17 |
| davidad_ | polynomial is the big-O bound on time with respect to problem size | 19:18 |
| davidad_ | so, time = O(n^k) | 19:18 |
| davidad_ | exponential is where time = O(k^n) | 19:18 |
| davidad_ | assuming that n is the problem size | 19:18 |
| davidad_ | and k is a constant | 19:18 |
| fenn | exponential is a subset of NP? | 19:18 |
| davidad_ | not always | 19:18 |
| fenn | ok, just another thing to worry about | 19:19 |
| davidad_ | NP is a subset of exponential, though | 19:19 |
| fenn | uh, no? | 19:20 |
| kanzure | I've always referred to these computational complexity classes to the 'computational feasability' of whatever problem I am focusing on at the moment | 19:20 |
| davidad_ | any problem that you can verify in polynomial time, you can solve in exponential time | 19:20 |
| davidad_ | by just trying everything and seeing if it works | 19:20 |
| kanzure | see, we were doing one of those methods today | 19:20 |
| kanzure | we were suggesting a case where we deploy about 50 million different proteins | 19:20 |
| kanzure | to test for similarities to a target protein from nature | 19:20 |
| kanzure | which is the exponential time | 19:21 |
| kanzure | verification would be setting up an experiment the same way as we observed the properties of the target protein the first place, I presume | 19:21 |
| davidad_ | yep | 19:22 |
| kanzure | insert some anomalous relation to finding the longest path and Hamiltonian cycles | 19:22 |
| kanzure | *cycles here | 19:22 |
| davidad_ | so in this case, your verification is constant-time | 19:22 |
| davidad_ | no matter how long the protein might be, | 19:22 |
| davidad_ | you can do an experiment to verify the properties in the same period of time | 19:22 |
| kanzure | Woh, oh-oh-oh | 19:24 |
| kanzure | Find the Longest Path | 19:24 |
| kanzure | Woh oh-oh | 19:24 |
| kanzure | Find the Longest Path | 19:24 |
| kanzure | If you said P is NP tonight | 19:24 |
| kanzure | There would still be papers left to write | 19:24 |
| kanzure | I have a weakness | 19:24 |
| kanzure | I'm addicted to completeness | 19:24 |
| kanzure | And I keep searching for the longest Path | 19:24 |
| kanzure | The algorithm I would like to see | 19:24 |
| kanzure | Is of Polynoimal Degree | 19:24 |
| kanzure | Buts its elusive, | 19:24 |
| kanzure | Nobody has found conclusive | 19:24 |
| kanzure | Evidence that we can find the Longest Path | 19:24 |
| kanzure | I have been hard | 19:24 |
| kanzure | Working for so long | 19:24 |
| kanzure | I swear its right, | 19:24 |
| kanzure | But he marks it wrong | 19:24 |
| kanzure | Somehow I'll feel sorry when its done | 19:24 |
| kanzure | GPA 2.1, | 19:24 |
| kanzure | Is more than I hoped for | 19:24 |
| kanzure | Garey, Johnson, Karp and other Men (and Women) | 19:24 |
| kanzure | Try to make it Order n log n. | 19:24 |
| kanzure | Am I a math fool | 19:24 |
| kanzure | If I spend my life in Grad School | 19:24 |
| kanzure | Forever following the Longest Path. | 19:24 |
| kanzure | Woh oh-oh-oh | 19:24 |
| kanzure | Find the longest path | 19:24 |
| kanzure | Woh oh-oh-oh | 19:24 |
| kanzure | Find the longest path | 19:24 |
| kanzure | (Dan Barrett) | 19:24 |
| davidad_ | ..ok | 19:24 |
| kanzure | Anyway, | 19:25 |
| fenn | words! yeah | 19:25 |
| kanzure | there was something else | 19:26 |
| fenn | are you sure NP is a subset of exponential? what if there is an infinite number of combinations? | 19:26 |
| davidad_ | well, complexity classes are defined for decision problems | 19:27 |
| davidad_ | where the output is only YES or NO | 19:27 |
| fenn | well, say you want to pick two real numbers | 19:27 |
| davidad_ | you can't | 19:27 |
| davidad_ | you can never finish outputting a real number | 19:27 |
| davidad_ | the best you can do is output an algorithm that represents a real number, | 19:28 |
| davidad_ | and algorithms are countable | 19:28 |
| fenn | you can verify 2.1000... + 3.900... = 6 | 19:28 |
| fenn | but what are the real numbers adding up to 6 that fenn likes? | 19:29 |
| davidad_ | hahah | 19:29 |
| fenn | you cant try all of them | 19:29 |
| davidad_ | you can't verify them in polynomial time, either | 19:29 |
| fenn | do i at least have a grasp of what it's all about? | 19:29 |
| davidad_ | I think so | 19:29 |
| davidad_ | the formal definition of an NP problem | 19:29 |
| davidad_ | is a decision problem that is solvable in polynomial time on a non-deterministic turing machine | 19:30 |
| davidad_ | which means it can "split" and do computations in parallel | 19:30 |
| davidad_ | but it can't split infinitely many times, because that would take longer than polynomial time | 19:30 |
| davidad_ | if, however, there's a reasonable number of possibilities, it can split and check all of them and be done in polynomial time | 19:31 |
| kanzure | hm | 19:31 |
| kanzure | splitting is interesting | 19:31 |
| fenn | oh fuck N is non-deterministic? now i really want to kill some comp-scientists | 19:31 |
| kanzure | considering we do self-replication here | 19:31 |
| davidad_ | yeah, the N is non-deterministic | 19:32 |
| davidad_ | lucky you can't kill people over IRC | 19:32 |
| davidad_ | (for me) | 19:32 |
| davidad_ | another way of thinking about splitting | 19:32 |
| davidad_ | is that the machine is the luckiest possible guesser | 19:32 |
| davidad_ | at any time, it can guess a number and be guaranteed to guess the right one if there is a right one | 19:33 |
| fenn | like a quantum computer | 19:33 |
| kanzure | huh? | 19:35 |
| kanzure | quantum, why? | 19:35 |
| fenn | because it exists in all possible states | 19:35 |
| fenn | it runs through all possible program paths, if you will | 19:35 |
| kanzure | oh, right, I remember some hype a few months ago about quantum programs and their 'fantastic possibilities' | 19:36 |
| kanzure | but I also remember Slashdotters fairly correctly refuting those ideas for quantum Traveling Salesman. | 19:36 |
| fenn | hmm why's that? | 19:40 |
| fenn | because the weights are real numbers and not quantized? | 19:40 |
| kanzure | http://science.slashdot.org/article.pl?sid=07/08/09/1535231 | 19:41 |
| kanzure | http://science.slashdot.org/comments.pl?sid=265375&cid=20172049 in particular | 19:42 |
| kanzure | ah, that's Engel | 19:42 |
| kanzure | I think he's on freenode somewhere | 19:42 |
| kanzure | http://science.slashdot.org/comments.pl?sid=265375&cid=20172269 | 19:43 |
| kanzure | haha - feasability calculations - http://science.slashdot.org/comments.pl?sid=265375&cid=20172747 | 19:43 |
| kanzure | this one is good - http://science.slashdot.org/comments.pl?sid=265375&cid=20172137 | 19:45 |
| fenn | sometimes in these computer science discussions the concept of 'good enough' gets lost entirely | 19:49 |
| fenn | i.e. would take the salesman less time traveling than it would take to optimize his route | 19:51 |
| kanzure | no | 19:51 |
| kanzure | the idea is that there may be a solution algorithm | 19:51 |
| kanzure | that will tell you the shortest path instantly | 19:51 |
| kanzure | 'instantly', i.e. within one step | 19:51 |
| fenn | ya | 19:51 |
| kanzure | and then that's it -- you just use that as long as you want :) | 19:51 |
| fenn | its just huge for high levels of complexity | 19:51 |
| fenn | but with the optical experiment, you'd get some halfway decent answers early on | 19:52 |
| kanzure | maybe | 19:52 |
| fenn | statistically speaking, 'probably' | 19:52 |
| kanzure | the optical solution is something like 1E167 years to come up with enough photons or something | 19:52 |
| kanzure | per that last slashdot link I gave up there ^ | 19:52 |
| fenn | that's to go through all possible permutations | 19:53 |
| fenn | its not like you're going to put 1e200 photons in a box | 19:53 |
| fenn | and then the answer pops out | 19:54 |
| fenn | anyway it's not a quantum computer | 19:54 |
| kanzure | I'd like to see somebody try to do 1E200 photons in a box | 19:56 |
| fenn | the box would explode :) | 19:58 |
| fenn | i wouldnt like to see it, my local parsec would be converted to plasma | 19:59 |
| kanzure | in ##neuroscience I just linked a person over to http://www.dbc.uci.edu/neurobio/Faculty/Lynch/lynch.htm after her expression of interest in the (bullshit) "holographic brain theory" (Pietsch stuff) | 20:01 |
| kanzure | ooh | 20:01 |
| kanzure | I also picked up some books today | 20:01 |
| kanzure | Microbiology, 5th ed., Prescott, Harley, Kim | 20:02 |
| kanzure | The Cartoon Guide to Genetics, Larry Gonick & Mark Wheelis | 20:02 |
| kanzure | Making PCR: A Story of Biotech, Paul Rabinow | 20:02 |
| kanzure | The Most Beautiful Molecule: The Discovery of the Buckyball, Hugh Aldersey-Williams | 20:02 |
| fenn | i think you could do the 100 city hamiltonian cycle with a reasonably small number of q-bits (not that i know what i'm talking about) | 20:03 |
| kanzure | practicality? | 20:03 |
| kanzure | do you remember D-Wave? | 20:03 |
| fenn | no | 20:03 |
| kanzure | they are the quantum computing guys | 20:03 |
| kanzure | they were making their rounds a few months ago | 20:03 |
| kanzure | getting about $40 million in funding overall | 20:03 |
| kanzure | they do supercold quantum computing and things seem to be looking strong for them | 20:03 |
| kanzure | they claim they'll be doing 256 qbits sometime this year | 20:04 |
| kanzure | and maybe 1024 qbits by the end of the year | 20:04 |
| fenn | that's impressive | 20:04 |
| kanzure | I was talking with the CEO a few months ago | 20:04 |
| fenn | all entangled? | 20:04 |
| kanzure | only to learn the next day that my dad has some money invested in them O.o | 20:04 |
| kanzure | I would've invited him to the chat if I had known that | 20:04 |
| kanzure | anyway, in terms of practicality | 20:05 |
| kanzure | I think that carbon might offer some interesting solutions | 20:05 |
| kanzure | for quantum computing | 20:05 |
| kanzure | since we can get quantum tunneling and the field effect on graphene | 20:05 |
| kanzure | plus semiconductor effects in buckyballs/CNTs | 20:05 |
| fenn | er... no | 20:05 |
| kanzure | no? | 20:05 |
| fenn | quantum tunneling is just one electron hopping over an energy barrier | 20:05 |
| kanzure | right | 20:06 |
| fenn | quantum entanglement is that one electron spread in 50 places at once | 20:06 |
| fenn | or two, if you're a physicist | 20:06 |
| kanzure | sure, but I'm just saying that graphene allows us to investigate quantum effects | 20:07 |
| kanzure | and it allows us easy access and manipulation | 20:07 |
| fenn | ok, but so do copper wires | 20:07 |
| kanzure | so it seems like an interesting platform to paly with | 20:07 |
| kanzure | *play with | 20:07 |
| kanzure | really? | 20:07 |
| fenn | ya | 20:07 |
| kanzure | so can it do entanglement? | 20:08 |
| fenn | there's too much noise in room temperature systems | 20:09 |
| fenn | scroll down to 'So how big is an electron?' http://laputan.blogspot.com/2003_09_21_laputan_archive.html | 20:12 |
| kanzure | heh, I am glad that he does not agree with Copenhagen | 20:13 |
| kanzure | I like the Feynman/von Neumann interpretations | 20:13 |
| fenn | i can never keep all those dead guys straight | 20:13 |
| kanzure | Feynman isn't dead. Not yet, anyway. | 20:14 |
| fenn | copenhagen = the electron could be anywhere but it is definitely somewhere, we just dont know? | 20:15 |
| kanzure | no | 20:16 |
| kanzure | Copenhagen is stuff like "view it and you kill babies in another universe" | 20:16 |
| kanzure | MWI shit | 20:16 |
| kanzure | haha, Mead studied under Feynman | 20:16 |
| kanzure | good | 20:16 |
| fenn | heh "Shut up and calculate!" | 20:17 |
| kanzure | "the electron is its own medium" | 20:23 |
| kanzure | hm | 20:23 |
| fenn | many worlds is its own interpretation.. copenhagen is observer-centric | 20:41 |
| kanzure | ah, that's right | 20:44 |
| kanzure | statistics and so on | 20:44 |
| epitron | omg observing particles kills babies?! | 20:44 |
| kanzure | I still need to finish reading that link, but it's good stuff | 20:44 |
| epitron | how do i stop! | 20:44 |
| kanzure | epitron: haha :) | 20:44 |
| fenn | epitron: cats, not babies | 20:44 |
| epitron | fenn: IT SAYS BABIES | 20:44 |
| kanzure | how do you know it was a cat? | 20:44 |
| kanzure | it's only a cat if you observe it to be a cat | 20:44 |
| kanzure | until then it has a 50/50 chance of being a baby | 20:45 |
| fenn | kanzure: schrodinger said it was a cat | 20:45 |
| epitron | schrodinger has cat-hater bias | 20:45 |
| kanzure | schrodinger didn't know what life was ;) | 20:45 |
| epitron | you cannot trust anything schrodinger did | 20:45 |
| kanzure | and there's a broadcast premiere of Star Wars on at the moment | 20:45 |
| kanzure | so please excuse me, I have an addiction to work on | 20:45 |
| epitron | haha | 20:46 |
| fenn | me too (sleep) | 20:46 |
| epitron | hey kanzure | 20:46 |
| kanzure | yes? | 20:46 |
| epitron | did you ever see the fanedit of Espiode IV? | 20:46 |
| kanzure | No? | 20:46 |
| epitron | you ever seen fan edits? | 20:46 |
| kanzure | nope | 20:46 |
| epitron | lemme find ya some | 20:46 |
| kanzure | whole movies? | 20:46 |
| epitron | yeah | 20:46 |
| kanzure | delicious | 20:46 |
| epitron | http://fanedit.org/wpTF/?p=29 | 20:47 |
| epitron | Star Trek - Kirkless Generations | 20:47 |
| epitron | http://fanedit.org/wpTF/?p=495 | 20:47 |
| epitron | STAR WARS: EP IV 2004 Special edition REVISITED | 20:47 |
| epitron | http://fanedit.org/wpTF/?p=485 | 20:47 |
| epitron | THE MATRIX EVOLUTIONS | 20:47 |
| epitron | http://fanedit.org/wpTF/?p=521 | 20:47 |
| epitron | Army of Darkness - Primitive Screwhead Edition | 20:47 |
| epitron | :D | 20:47 |
| epitron | (and of course, the classic jar-jar-less Episode I) | 20:47 |
| epitron | i'm downloading the first 3.. i've seen the 4th | 20:48 |
| kanzure | Jar-jar-less? Awesome. | 21:56 |
| -!- Aulere [n=dragon_d@131.229.176.252] has joined #hplusroadmap | 23:01 | |
| kanzure | Welcome, dragon. | 23:02 |
| epitron | http://funtarded.com/pics/show/2641 | 23:59 |
Generated by irclog2html.py 2.15.0.dev0 by Marius Gedminas - find it at mg.pov.lt!