00:11:42 -!- lleu has quit (Quit: That's what she said).
00:27:31 -!- Phantom_Hoover has quit (Read error: Connection reset by peer).
00:50:07 <zzo38> How can SoX load a LADSPA plugin by number?
00:51:27 -!- TieSoul_ has joined.
00:53:31 -!- TieSoul has quit (Ping timeout: 240 seconds).
00:59:43 -!- bb010g has quit (Quit: Connection closed for inactivity).
01:01:50 -!- hjulle has quit (Ping timeout: 256 seconds).
01:05:23 -!- Herbalist has quit (Ping timeout: 250 seconds).
01:12:06 <zzo38> O, SoX uses a different code for compile-time assertions. Some programs use a global array variable. I use a structure containing an array. SoX uses a enum with a division (so you get a division by zero if the assertion is false).
01:16:56 <tswett> I hope nobody minds if I download all the logs from January 2014 through April 2015 from codu.org.
01:17:57 <pikhq> zzo38: If you don't mind using newer features, _Static_assert is quite nice.
01:19:49 <tswett> 25.3 megabytes, oh boy.
01:22:10 <zzo38> I have always used a structure containing an array and it works; this was before I saw how any other program did it (or even knew of such thing), and I still think it is better than declaring a variable.
01:23:33 <pikhq> (_Static_assert is a C11 feature that causes a build failure with a nice message if a constant expression is true. Quite convenient.)
01:23:55 <ais523> I have a macro that expands to the array version or C11 version depending on version
01:24:11 <pikhq> Yeah, that's the way I tend to roll.
01:24:15 <ais523> (note that you need to use -1 as the array size for a failure and +1 for success; gcc is fine with a 0-sized array, some compilers aren't)
01:24:50 <zzo38> Yes I know; I find 0-sized arra useful sometimes
01:24:58 <pikhq> The C11 version is nice but new enough that there's not any way you can even pretend it's ubiquitous.
01:25:10 <zzo38> I do use -1 for failure and +1 for success (by a ? : operator) in this case.
01:26:26 <zzo38> But sox.h uses the following macro: #define lsx_static_assert(e,f) enum {lsx_static_assert_##f = 1/((e) ? 1 : 0)} So, if e is false then it is dividing by zero at compile-time.
01:40:26 <oren> lol seems like there are many ways to "dynamically" cause a screwup at compile time
01:49:49 <oren> `run echo 'I=1 I/0' | scrip7
01:50:00 <oren> `run echo 'I=1 I/0 _pI' | scrip7
01:53:18 <zzo38> I think bzip2 checks at runtime instead though (I don't know why they didn't make it to check the sizes of types at compile-time)
01:53:21 -!- hilquias has joined.
01:54:30 <tswett> 15:55:22: <boiljan> mrome_plabatexkells forby
01:54:38 <tswett> The net is currently stupid.
01:55:21 -!- lleu has joined.
01:57:18 <zzo38> Actually I did make a patch to bzip2 to check at compile-time. An error message is also possible, but #error doesn't work; you need #line instead.
02:01:39 <tswett> Finally, it generated something it thinks I said:
02:01:41 <tswett> <tswett2> And outher bechack has only suppoper.
02:06:54 -!- ais523 has quit.
02:12:16 -!- lleu has quit (Quit: That's what she said).
02:46:38 <zzo38> I find a few problems with LADSPA, LV2, and some other audio plugin architectures; I have my own kind of idea, where plugins are identified by URIs like in LV2, but the RDF graph is compiled into the plugin rather than external and is also completely optional; also the port types are different, and there are some other differences too.
02:49:45 <zzo38> For example you can have: void*plugin_enum(void*(*callback)(const char*uri,char compat,Plugin*plugin)); Plugin*plugin_find(const char*uri); Plugin_Instance*plugin_open(Plugin*p); void plugin_enum_triples(......
02:50:20 <Sgeo> I think I forgot my password
02:50:34 <Sgeo> For this encrypted directory
02:51:20 <Sgeo> No, I think I have it
02:51:42 <Sgeo> It's breaking for a different reason
02:55:52 -!- variable has changed nick to trout.
03:08:13 -!- augur has quit (Quit: Leaving...).
03:35:57 <MDude> Well, stiff on how brain works is pretty itneresting in general.
03:36:11 <MDude> Oh, wrong channel.
04:02:22 -!- trout has quit (Ping timeout: 272 seconds).
04:02:33 <tswett> fung*t doesn't answer to fung*t, right?
04:03:43 <tswett> Stage 1: the neural net's output is less accurate than fung*t's. Stage 2: the neural net's output is more accurate than fung*t's. Stage 3: the neural net's output accurately simulates fung*t's inaccuracies.
04:03:53 <tswett> This neural net is currently in stage 1, alas.
04:03:54 <zzo38> Now I made up the syntax highlighting programming language; it is: http://zzo38computer.org/textfile/miscellaneous/syntax_highlighter
04:05:42 <tswett> !blsq {1 2 1 2 4 3 4 4 3 3} {{3 4} {1 2 4 3} {} {{}}}{{{}}{{}}}
04:05:59 <tswett> NB: I have no idea what blsq is.
04:06:22 <zzo38> Do you think this draft format is good or do you suggest other changes please?
04:08:10 <tswett> Here's an example of what the neural net can produce when you tell it to try not to produce too much nonsense:
04:08:12 <tswett> 00:46:04: <Sgeo> ais523: you can should see the commands some of the result to the reg for the BAIC borth one on the same since as bothered the question of the first been with for Data
04:09:51 <tswett> And here's what it produces when you tell it to produce *extra* nonsense:
04:10:01 <tswett> 18:21:48: <ais523_> oh, weird Wither thosis, trotoses without ntr p( etrollessais mirener?
04:10:26 <tswett> 03:08:55: * ooilFlayLig compiling peyhecasfulk, UR". VN,MIOTEAA\ bEW3? "bogoott"
04:11:38 -!- GeekDude has quit (Quit: {{{}}{{{}}{{}}}{{}}} (www.adiirc.com)).
04:23:45 -!- MDude has changed nick to MDream.
04:47:44 -!- variable has joined.
04:56:46 -!- Lyka|Phone has joined.
04:59:03 <Lyka|Phone> realized that I don't know how to make a brainfuck interpreter on an arduino
05:08:56 -!- Lyka|Phone has changed nick to Lyka|Away.
05:26:09 -!- augur has joined.
05:47:57 -!- Lyka|Away has changed nick to Lyka.
05:49:45 <Lyka> what is the minimum tape length for brainfuck?
05:51:36 <Lyka> is 256 bytes enough?
05:52:46 <quintopia> you act as if there is some supreme arbiter of brainfuck
05:53:28 <Lyka> is that a yes?
05:54:00 -!- bb010g has joined.
05:54:21 <quintopia> it's an impelling to answer your own questions
05:54:40 <Lyka> i'm trying to make an interpreter for a device with 2k of ram
05:55:38 <quintopia> there is very little overhead for a bf interpreter
05:55:44 <quintopia> you could make pretty much all of that tape
05:56:09 <Lyka> what about program code?
05:57:04 <quintopia> it's designed to be the language with the shortest compiler, but it ends up being the one with nearly the shortest interpreter in every language
05:57:18 <quintopia> input goes directly into the tape!
05:57:47 <Lyka> right. no point in caching it
05:58:04 <Lyka> but won't i need to cache the program?
05:58:53 <quintopia> https://github.com/Harvie/Programs/blob/master/arduino/BrainFuck/BrainFuck.pde
06:02:29 <Lyka> you misunderstand. i want it to load code from the terminal
06:02:42 <Lyka> not just display the output
06:03:03 <quintopia> just showing you an example of how it was done before, and how little code was required
06:05:02 -!- variable has quit (Ping timeout: 245 seconds).
06:10:54 <Lyka> i'm working on a multi-unit project at the moment, so i have to dig out an arduino that is not in use instead of using one of the ones i keep lying near my bed for things like this
06:47:29 -!- Wright_ has quit (Ping timeout: 256 seconds).
06:57:40 -!- Sgeo_ has joined.
06:59:55 -!- Sgeo_ has quit (Client Quit).
07:00:31 -!- augur has quit (Ping timeout: 256 seconds).
07:02:03 -!- augur has joined.
07:13:57 <f|`-`|f> How is it being a part of the Exponential Finance meet?
07:15:04 <f|`-`|f> http://www.augur.net/blog/augur-selected-to-participate-in-singularity-university-and-cnbc
07:15:17 <f|`-`|f> You are pretty popular in decentralisation
07:15:56 <augur> thats not me, alas
07:16:16 <f|`-`|f> If only you were a Decentralised Global Prediction Market
07:18:01 <f|`-`|f> Also an overly flashy video http://www.augur.net/blog/how-augur-works-video
07:18:19 <augur> categorical markets
07:18:23 <augur> where you buy category theory proofs
07:20:41 <f|`-`|f> eg, the Markets under Scientific Crap
07:21:32 <f|`-`|f> Also HoTT uses Infinity Groupoids
07:33:44 <augur> Lyka: whatd you get
07:34:03 <Lyka> programmed brainfuck onto an arduino
07:34:59 <Lyka> though it evaluates on ']' instead of on '['
07:35:50 <Lyka> so much easier to do it that way, it seems
07:36:27 <Lyka> tape size = 256
07:37:01 <Lyka> cell value wraps around
07:38:12 <Lyka> max program size = 255
07:38:51 <Lyka> i quicly tested it with the program "[,.]"
07:40:33 -!- oerjan has joined.
07:44:09 <Lyka> of course, the brainfuck program "[,.]" shouldn't do anything...
07:48:12 <Lyka> 03:34:39 < Lyka> got it!
07:48:12 <Lyka> 03:34:46 < augur> Lyka: whatd you get
07:48:12 <Lyka> 03:35:05 < Lyka> programmed brainfuck onto an arduino
07:48:12 <Lyka> 03:36:01 < Lyka> though it evaluates on ']' instead of on '['
07:48:12 <Lyka> 03:36:48 < augur> ah
07:48:15 <Lyka> 03:36:52 < Lyka> so much easier to do it that way, it seems
07:48:17 <Lyka> 03:37:29 < Lyka> tape size = 256
07:48:20 <Lyka> 03:38:03 < Lyka> cell value wraps around
07:48:22 <Lyka> 03:39:14 < Lyka> max program size = 255
07:48:25 -!- Phantom_Hoover has joined.
07:48:25 <Lyka> 03:39:54 < Lyka> i quicly tested it with the program "[,.]"
07:48:37 <coppro> what do you mean by "evaluates on ']'"?
07:48:56 <oerjan> Lyka: um please don't paste the much into the channel, we have logs
07:49:41 <oerjan> two of them, see topic
07:51:04 <Lyka> where, due to a bug i don't now how to get around yet, "[,.]" acts like ",[.,]" should
07:52:58 <oerjan> let me guess, you don't actually have code for skipping a loop?
07:53:32 <Lyka> don't know how
07:53:43 <oerjan> it's a common newbie problem, i think
07:54:08 <oerjan> you need to find the matching ]
07:54:42 <oerjan> a common way is to use a counter for nesting depth
07:55:12 <oerjan> hm maybe it's more subtle then
07:56:52 <Lyka> the test for zero or nonzero only occurs in the ']'
07:57:00 <Lyka> in what i wrote
07:57:13 <Lyka> instead of being in the '[' as well
07:57:28 <oerjan> right, that does tend to lead to this problem
07:58:52 <oerjan> when it's zero at [ you need to skip at least to the ], without executing anything in between
07:59:10 <oerjan> and you can avoid this at the first iteration
08:00:12 <oerjan> in particular, you need some way of skipping over code that has never been executed.
08:00:48 <Lyka> http://pastebin.com/azC5ZqEH
08:02:49 <oerjan> this not skipping [...] problem tends to come up in brainfuck because it's the one part of parsing which you _cannot_ do simultaneously with executing.
08:03:13 <oerjan> and you clearly have a "parse and execute at the same time" implementation there.
08:03:23 <Lyka> clearly urban miller was a sadist
08:04:10 <oerjan> sorry, but brainfuck is ridiculously easy to parse compared with all real programming languages.
08:04:24 <Lyka> i wasn't serious
08:05:29 <oerjan> anyway, the simplest fix is to have a for loop in the code for '[' that skips forward until you reach the same nesting depth again
08:06:21 <oerjan> i.e. finds the matching ']'
08:07:12 <oerjan> it's not very optimized but it works
08:09:03 <oerjan> the more efficient way is to actually parse the entire program first to match []s
08:09:31 <oerjan> then you only have to do it once
08:20:42 -!- copumpkin has joined.
08:24:39 <Lyka> http://pastebin.com/R4qqTfjA
08:24:39 <oerjan> <b_jonas> like http://www.xkcd.com/251/ ? <-- it's funny when you can see clear artistic improvement in a stick figure comic...
08:27:20 <Lyka> next step is to have a command that will set everything o the beginning
08:28:11 <Lyka> so that i won't need to power-cycle or reset the arduino to make another try
08:28:20 * oerjan thinks he'd have used return instead of while (!fin) there
08:28:35 -!- Weloxux has joined.
08:30:40 <Lyka> then i will modify and extend the language for my needs
08:34:46 <HackEgo> [wiki] [[Special:Log/delete]] delete * Oerjan * deleted "[[Groovy]]": Not esoteric
08:35:48 <Taneb> It's Eurovision Song Contest day!
08:37:06 * Lyka wonders if Dragostea din tei is what is being referrd to
08:37:24 <Lyka> spelled it wrong
08:37:50 <oerjan> i don't think that was ever in the contest
08:38:11 <Lyka> only euroean song i now
08:38:32 <oerjan> also, i think you spelled it correctly. well the song name.
08:38:49 <oerjan> i bet you know some others, if you consider that england is in europe
08:39:18 <Lyka> england and ireland count?
08:39:35 <oerjan> they're certainly in the contest, anyway
08:39:47 <oerjan> well at least back when i watched it
08:40:08 <oerjan> but then, so is israel, so it's a little fuzzy.
08:40:48 <oerjan> also, you may know songs in english that happen to be from continental europe.
08:41:54 <Lyka> 99 luftbaloons?
08:42:22 <oerjan> well i think there's an english translation too
08:43:34 <Lyka> french canadans are the ones you pretend not to undestand engish unlesss you swear at them, right?
08:44:00 <oerjan> you'll have to ask boily about that
08:45:27 <Taneb> Lyka, them and Parisians
08:45:50 <Lyka> well the latter have an excuse
08:46:11 <Lyka> they aren't in an english-speaking country
08:51:24 -!- Tritonio has joined.
08:51:26 * oerjan report Lyka to the OQLF
08:52:22 <oerjan> Office québécois de la langue française hth
08:53:49 <Lyka> is that the agency that controls the Canadian dialect of French?
08:54:30 <Phantom_Hoover> the uk is definitely in europe, despite what ukip might want to believe
08:57:06 <oerjan> i don't think they control the language itself, their job is to promote it
08:58:45 <oerjan> oh they do make at least one dictionary
09:07:00 <HackEgo> [wiki] [[RLS]] http://esolangs.org/w/index.php?diff=43025&oldid=43019 * Oerjan * (+19) lowercase template, link
09:07:35 <HackEgo> [wiki] [[RLS]] M http://esolangs.org/w/index.php?diff=43026&oldid=43025 * Oerjan * (-1)
09:13:24 <HackEgo> [wiki] [[Talk:Underload]] http://esolangs.org/w/index.php?diff=43027&oldid=43023 * Oerjan * (+122) /* Why the reserved characters? */ ()~* ?
09:15:54 -!- TieSoul_ has changed nick to TieSoul.
09:19:42 -!- bb010g has quit (Quit: Connection closed for inactivity).
09:42:40 -!- mitchs has joined.
09:49:58 <HackEgo> [wiki] [[Underload/Numbers]] http://esolangs.org/w/index.php?diff=43028&oldid=42680 * Oerjan * (+174) explanation and linkback
10:58:24 -!- J_Arcane has quit (Quit: ChatZilla 0.9.91-rdmsoft [XULRunner 32.0.3/20140923175406]).
11:08:05 -!- solid_whiskey has joined.
11:12:38 -!- lleu has joined.
11:18:36 -!- solid_whiskey has quit (Quit: Leaving).
11:26:50 -!- J_Arcane has joined.
11:44:46 <oerjan> there seems to be some feline and canine precipatation
11:45:27 * oerjan briefly considers cutting off his hand for ruining his puns with typos again
11:45:51 -!- Tritonio has quit (Remote host closed the connection).
11:48:04 <oerjan> <tswett> 15:55:22: <boiljan> mrome_plabatexkells forby <-- WHERE ON THE NET DID THIS HAPPEN
11:54:07 <Melvar> oerjan: A type that apparently can write syntactically correct mediawiki if fed wikipedia.
11:54:42 <Melvar> Also *almost* correct latex when fed a large set of mathematical papers.
11:55:17 <oerjan> but boiljanic portmanteaus are still beyond it.
11:55:33 <Melvar> Probably the corpus is a bit small.
11:55:54 <int-e> mmm, who's the guy next to Martellus, anybody we know?
11:56:36 <Melvar> oerjan: https://karpathy.github.io/2015/05/21/rnn-effectiveness/ is the article describing the principle that tswett read before doing this.
11:59:07 <int-e> (Actually I'm only trying to milk that rather boring GG page for something interesting.)
11:59:10 <oerjan> <f|`-`|f> With Cokutosis <-- sounds painful tdnh
12:00:03 <oerjan> int-e: my interesting fact is this: i think this is the first time in-story that any of our protagonists visit a town/city that exists in the real world
12:00:03 <int-e> oerjan: you might enjoy this twist, http://www.sandraandwoo.com/gaia/2015/05/19/breaking-all-barriers-071/ and the following page.
12:02:06 <oerjan> that is, visibly and not in flashback
12:02:49 <oerjan> (there have been paris flashbacks before, and who knows where people went during the time skip)
12:03:01 <int-e> that's true. but the Foglios have been preparing the readers for this "surprise" for a long time.
12:03:23 <int-e> I'm looking forward to some above ground pictures though.
12:05:25 <oerjan> int-e: i've seen that wish joke before.
12:06:10 <int-e> okay. it was new to me.
12:07:15 <HackEgo> [wiki] [[SCRUBS4U]] N http://esolangs.org/w/index.php?oldid=43029 * 93.106.28.108 * (+2200) page made
12:07:57 <HackEgo> [wiki] [[SCRUBS4U]] http://esolangs.org/w/index.php?diff=43030&oldid=43029 * 93.106.28.108 * (+6) /* Keywords */
12:09:37 <HackEgo> [wiki] [[Language list]] http://esolangs.org/w/index.php?diff=43031&oldid=43021 * 93.106.28.108 * (+15) /* Non-alphabetic */
12:10:19 <int-e> . o O ( at least it's not brainfuck )
12:11:05 <HackEgo> [wiki] [[Language list]] http://esolangs.org/w/index.php?diff=43032&oldid=43031 * 93.106.28.108 * (-15) /* Non-alphabetic */
12:11:39 <HackEgo> [wiki] [[Language list]] http://esolangs.org/w/index.php?diff=43033&oldid=43032 * 93.106.28.108 * (+15) /* S */
12:18:21 -!- hjulle has joined.
12:20:39 -!- GeekDude has joined.
12:51:01 -!- Frooxius has quit (Quit: *bubbles away*).
13:10:38 <oerjan> http://www.scottaaronson.com/blog/?p=2293 weakness in diffie-hellman
13:12:59 <GeekDude> 'even today you can force many servers to “downgrade” to the 512-bit, export-grade keys'
13:13:18 <coppro> it's not weakness in diffie-hellman
13:13:27 <coppro> in weakness in implementation
13:33:28 <tswett> Melvar: well, the corpus is... I forget how big, but at least like 20 megabytes.
13:33:58 <tswett> Also, as of the boiljan message, the net hadn't been training very long.
13:34:15 <tswett> I've had it training overnight now.
13:34:18 <Taneb> tswett, what is this?
13:34:59 <tswett> Taneb: well, there's this article about neural nets: https://karpathy.github.io/2015/05/21/rnn-effectiveness/
13:35:20 <Taneb> Have you been using it?
13:35:42 <tswett> I think the neural net has now learned that the first digit of timestamps rarely changes.
13:36:03 <tswett> Here's an example of what the net is producing now:
13:36:05 <tswett> 21:55:37: <b_jonas> oerjan: that's why it is token some point to other macros versions replace
13:36:33 <tswett> 17:58:37: <boily> anyway, Bicyclidine: Yeah
13:37:20 <tswett> 22:07:18: <HackEgo> [wiki] [[Quoted2, OMPAc]] http://esolangs.org/w/index.php?diff=40942&oldid=41553 * Esowiki201529 * (+1328) collection loops.: that I want tonards even go
13:44:20 <tswett> Here's something cool that it produced.
13:44:29 <tswett> 21:46:18: <mroman_> !bfjoust snail (>)*9(>[(-)*9[+]]>)*-1
13:44:29 <tswett> 20:45:24: <zemhill> mroman_.snail: points -13.01, score 19.69/100, rank 47/47 (change: --)
13:46:19 <oerjan> !bfjoust netsnail (>)*9(>[(-)*9[+]]>)*-1
13:46:37 <EgoBot> Score for oerjan_netsnail: 2.8
13:46:50 <oerjan> !zemjoust netsnail (>)*9(>[(-)*9[+]]>)*-1
13:46:56 <zemhill_> oerjan: I do !zjoust; see http://zem.fi/bfjoust/ for more information.
13:46:56 <EgoBot> help: General commands: !help, !info, !bf_txtgen. See also !help languages, !help userinterps. You can get help on some commands by typing !help <command>.
13:47:02 <oerjan> !zjoust netsnail (>)*9(>[(-)*9[+]]>)*-1
13:47:03 <zemhill_> oerjan.netsnail: points -12.05, score 12.13, rank 47/47
13:47:16 <Vorpal> hey I just found an old CD-RW with compressed IRC logs from 2006-2008. WTF
13:48:36 <tswett> I'd say that when you run it with a low temperature setting, this net's output is about as coherent as fungot.
13:48:36 <fungot> tswett: ( april 1824.)
13:49:34 <tswett> 10:25:08: <mroman> and it's about the beer of arithmetic functions from the program, once you should have any statement of the authorpolve to call was some of the virtua tools.
13:49:43 <tswett> fungot: give me something longer, please.
13:49:43 <fungot> tswett: for that service. nobody else would have brought on that rupture which we all have of our furniture, fnord, the vision of fnord, became attached to the name of justinian, in countries to which no translation of the, as a mob, has dined with the king, whom they sent, fnord, and foolish timidity.
13:51:19 <tswett> 21:37:12: <Vorpal> elliott, In there's a complete, then the puzzle of "constructive special" "help"
13:51:34 -!- GeekDude has quit (Ping timeout: 255 seconds).
13:52:12 <tswett> Vorpal: that's what my neural net thinks you sound like.
13:52:44 -!- Patashu has quit (Ping timeout: 265 seconds).
13:53:18 <oerjan> no neural net can possibly grasp my style hth
13:54:23 <Taneb> tswett, can you tickle my ego and find some of me?
13:55:05 <tswett> 18:12:04: <Taneb> Codechole: I've used to tay believour kernel worry
13:55:18 <Vorpal> Since I'm moving soon I'm going through old boxes of stuff and (hopefully) throwing out some stuff instead of taking it with me. I found some interesting CDs and floppieds
13:55:22 <Taneb> That almost sounds like it makes sense!
13:55:25 <tswett> 11:25:57: <Taneb> That can said of good ctceritizes.
13:55:27 <Taneb> Just like most things I say
13:55:39 <Vorpal> Like a floppy with Netscape 2.0 for Mac
13:55:54 <tswett> 17:21:20: <Taneb> interesting on x86-935 has on one codes are supposed to do that...
13:56:04 <tswett> 17:22:12: -!- MDude is now known as Taneb.
13:56:21 <tswett> 21:13:51: <Taneb> oohtoopia
13:57:48 <Vorpal> Taneb, how does it think fungot sounds?
13:57:48 <fungot> Vorpal: ladies and gentlemen, there is nothing, therefore, the population of which is made the plea of the violent ebbs and flows of public feeling, he should as nearly as possible, still more decisive.
13:58:05 <Taneb> Vorpal, I'm not tswett, despite allegations
13:58:08 <HackEgo> Taneb is not elliott, no matter who you ask. He also isn't a rabbi although he has pretended in the past. He has at least two backup keyboards with dodgy SHIFT KEys, and cube root of five genders. (See also: tanebventions)
13:59:33 <tswett> 17:44:01: <fungot> oerjan: if come like, you know the know with tradened-is?
13:59:33 <tswett> 12:37:22: <fungot> fizzie: magical, «\gense(. i also optimal) nvaluge-O# sort|rubernanner@heat
13:59:33 <tswett> 22:52:08: <fungot> oerjan: cat's your meant recording almostly
13:59:33 <fungot> tswett: " i shall take care to put burke's work on the " sublime" in eloquence has ever been mentioned before, but better secured in every part of the article under our fnord defence of the popular interest to become a member of an academy of inscriptions, was foremost in fnord the people of that time, i said that there is anything wanting which is within my power to give, any opinion of their own body; those children who now a
13:59:33 <fungot> tswett: we believe as much as fnord and fnord. church and state, like aldermen, we'll revel. we'll live when hushed the fnord din, in smoking and in cards, sir, it is something of a personal tone. i am assured that, necessarily, and by a judicious management of it; i mean, is a very important incident, as i am personally concerned, whether we have any political creed at all.
13:59:33 <fungot> tswett: " in that he dissolved the parliament in 1641. " fnord ye," said beauclerk, " alter your rule; and prescribe only for your enemies. was such a distinction ever heard of that institution is standing in the same manner, the absurdity which precedes a period of half a century has assigned to us, who are at liberty to modify and change the awful state of things, it produces in the beginning of the world, and celebrates the
14:00:18 <tswett> 16:08:20: <fungot> blew: +-+ \ -> T >=>/f,|-]n
14:00:18 <fungot> tswett: the king comes forward and fnord napoleon to single combat. napoleon accepts it. sacrifices are offered. the ground is measured by ney and fnord. ney engages fnord and kills him. the prince is inclined to do so; nor do my friends expect that i shall never hear them again would be embittered by the recollection of it fnord my surprise. it is a cheek, perhaps the only poet whose writings would become much less intelligibl
14:00:19 <Vorpal> On a old dual Mac/Windows CD thingy, how do you mount the mac part from Linux?
14:00:38 <tswett> Finally, this line is interesting:
14:00:54 <tswett> 12:56:50: <lambdabot> fungot (>->+)*8[>-)*3[>()*1[-]].>.-]>)*-1
14:06:15 <tswett> I'm going to stop the training, since it hasn't improved much in the last six passes through the training data.
14:06:53 <tswett> 25.3 megabytes, by the way.
14:07:05 <tswett> Then I'm going to restart the training using a larger state vector.
14:08:08 <Vorpal> Ah.. mount -t hfs /dev/sr0 /somewhere
14:12:56 -!- MDream has changed nick to MDude.
14:13:06 -!- boily has joined.
14:16:42 <tswett> There. Now the state vector has a size of 300 instead of 100.
14:18:04 -!- oerjan has quit (Quit: Later).
14:22:00 <oren> Last week's bloomberg businessweek has an article on these "coding bootcamps"
14:25:32 <oren> Given the well known bimodal grade curves in compsci classes, I'm not sure that it can be very effective
14:30:15 <oren> I mean, I'm sure they mean well, but it's not necessarily possible to teach any person to program
14:30:36 <MDude> I don't see how a bimodal grade curve would be all that important.
14:31:05 <MDude> I mean, it's unusual, but if someone wants to try to learn then they want to try.
14:31:37 <MDude> It just means there's less people who end up exactly average.
14:32:31 <oren> Well having two peaks means somehow there's two groups with different average performance, even with identical teaching
14:32:51 <MDude> And "coding bootcamps" sound like they'd be more extracurricular than normal classes, which I expect would make them less attractive to someone just tryingto get a course.
14:34:14 <Taneb> oren, these aren't a new thing at all
14:34:15 <MDude> So? Why think that means one group should barred from classes unless you're cutting people due to too amny applications anyway?
14:34:26 <Taneb> My mum did one back in the late 80's, crash course in COBOL
14:35:37 <oren> oh? bloomberg seems to think this is "the new anti-college" but maybe they're not informed. they're a business magazine not a tech one after all
14:36:28 <MDude> That could mean they're now taking on that role, even if already existing for some time.
14:47:52 <Taneb> oren, the thing my mum did, she already had the job, and this was training for it
14:48:13 <Taneb> I think back then there was more of a shortage of rank-and-file programmers
14:58:36 -!- boily has quit (Quit: ATYPICAL CHICKEN).
15:01:26 <tswett> The larger neural net just produced its first output.
15:01:33 <tswett> The English is, of course, terrible.
15:01:35 <tswett> 15:32:03: <ais523> that's this carlines, that throutists would be as with much pome actory or well? AI hup that resands onle is crectly intreats and rather byth, it's just it's spoted.
15:03:25 <tswett> There are a few mistakes with the log format.
15:03:27 <tswett> 18:56:27: -!- Bicyclidine: Das joined #esoteric.
15:04:53 <tswett> In low temperature mode, it's a bit better:
15:04:55 <tswett> 03:32:00: <oerjan> i mode by as accumating programming them the looks attempt I that to be a bit alto a stand, that in the same and first
15:10:52 <int-e> a bit alto a stand
15:14:09 <MDude> Oh, the got of fun was already a neural net, and you're making a bigger one?
15:15:32 <int-e> fungot is using some sort of Markov model
15:29:49 <MDude> I thought so, but figured "lager neural net" meant there was also a smaller one doing something.
15:38:48 * Melvar wonders how one would base a neural net on beer …
15:40:03 <oren> Fluid compiting?
15:41:43 <Melvar> < tswett> 21:13:51: <Taneb> oohtoopia – I choked from giggling because I was drinking :(
15:48:01 <tswett> MDude: overnight I trained a neural net with a size-100 state vector; now I'm training one with a size-300 state vector.
15:51:24 <tswett> About forty minutes ago I pasted some lines from the round 1000 version of the neural net. Here are some best-ofs from the round 2000 version.
15:51:27 <tswett> 20:48:47: <olsner> so it dusticularly
15:51:34 -!- sebbu has quit (Ping timeout: 264 seconds).
15:51:51 <tswett> 15:06:24: <fizzie> Which mease the interasted on a lazy as what esolang before that guys on a syntar integer to be peoplegremit uname ow compitable his int will be enethang.
15:52:32 <MDude> I get how an artificial neural based on summing inputs and outputting if they reach some treshold can be done with analog components.
15:52:56 <MDude> But I have no diea how to implement the weighting system without bringing in a digital controller.
15:53:01 <tswett> 03:08:52: <aig523> boily: ( (--) and printf)
15:53:30 <tswett> 09:40:40: -!- Patashu has quit (Read error: Changerion restron)
15:54:20 <tswett> 06:44:57: <oren> es, up it pokemonicies point, on rull user of the data point
15:54:54 <tswett> 22:13:34: <Vorpal> for how to deal with a phime over cention in, so might group dentions, off, it segner reowsing the owf to?
15:54:54 <tswett> 02:27:10: <oerjan> puzzy gleey speak.
15:54:58 <MDude> Which is annoying, since a neural net made entirely of a handful of analog components might actually be worthwhile to have, just due to ease of manufacture.
15:55:42 <MDude> Since it'd be a way to have a machien that learns, but doens't need a clean room to produce.
15:57:32 <b_jonas> tswett: does this learn from the logs of this channel only?
15:59:27 <tswett> The low-temperature setting, as always, occasionally produces extremely long lines.
15:59:39 <tswett> 20:18:25: <ais523> elliott_: The decidable are some some to be possible like your to groups and thing about the data out thing. i don't want to an interesting to the fur down to script from some time, but and when it assuming a single does the bot in the top again, it's not a sequence, and they kind of a while there is what actually may would be interesting the another details with because it was a size so it would be a little pick to fair the starts of
15:59:39 <tswett> source subsprision of the for the specifier has a but sing the feature of a bad 3-fext to define some fact of the succerdal with some up a first programs is a stack, arr some also doesn't work to the regular is starts in particular with how to change the some people be of the strings for to the stales and then simple spells for them
16:01:22 <tswett> Let's see. The made-up words there are "subsprision", "3-fext", and "succerdal" only, I think.
16:02:48 <b_jonas> 3-fext should be 3-vext I think
16:03:53 <tswett> O! that thrice-vext knave;
16:07:04 <tswett> There are a couple of two-word sequences that aren't realistic, like "some some", "your to", "but and", "may would", "the another", "with because", "the for", "a but", "is starts", and "for to".
16:07:25 <b_jonas> tswett: oh wait, "arr" is also made up
16:08:08 <tswett> Probably is, but it also happens to be a real word.
16:08:17 <tswett> fungot: how comprehensible are you these days?
16:08:46 <b_jonas> that might actually be in the channel logs
16:09:17 <tswett> Why isn't fungot responding?
16:09:18 <b_jonas> does the corpus try to ignore programming language code?
16:10:04 <b_jonas> haskell and J and perl one-liners people type to the bot or just mention to each other, or even worse stuff
16:10:39 <tswett> The neural net does occasionally try to produce various sorts of code.
16:11:07 <tswett> Here's a URL it came up with:
16:11:08 <tswett> 02:33:07: <olsner> "Tew acent charactering https://ilti.com/tron.Qlot, something
16:11:19 <Lyka> got my arduino version of brainfuck working
16:12:14 <b_jonas> oerjan: are you here under some other name?
16:12:27 <tswett> b_jonas trying to speak meaningless English but eventually degenerating and producing meaningless esocode instead:
16:12:29 <tswett> 17:54:45: <b_jonas> I wos actually output to them in Or and using without part of the an mon theory, but tend (all refents an ---d Of 'mf c to -> " 12 f 3Do.0 [_]>)
16:12:59 <tswett> Some code in whatever language "uls" is:
16:13:00 <tswett> 08:21:46: <Solace> `uls f -mc-c (xc + &
16:13:56 <b_jonas> there are so many esoteric languages that it's hard to produce any byte sequence that isn't plausibly input to any of them
16:14:17 <MDude> I'd like to put a neural net in a channel that uses MS Comic chat, so it learns to also put a little emote with each message.
16:14:40 <b_jonas> MDude: train it on such web forums that use lots of emotes
16:14:51 <tswett> Aha, I got it to generate some brainfuck code.
16:14:58 <MDude> Not really the same, I think.
16:15:05 <tswett> +++++++++++>++++++->>++++>>.->>[-]<.>+)>-1]--(->++]>)*2(++>[+]>)*0(>(-)*20>(+)*0(-0>->(+)*20>(-)*220+.+++>(-)*20>(+)*40>(-)*20>(+)*20>(-)*20>(>)*6(>[(+)*8[-]]>[(-)*9[+]])*--
16:15:06 * Lyka read that as comic sans
16:15:07 <b_jonas> with balanced square brackets?
16:15:19 <MDude> Comic chat has an actual emotion wheel system for displaying emotion.
16:15:20 -!- variable has joined.
16:15:20 <tswett> I mean, some of them are balanced.
16:15:36 <Lyka> tswett: what are those numbers?
16:15:44 <b_jonas> tswett: oh, that's actually bfjoust code
16:15:52 <tswett> Right, the numbers are from bfjoust.
16:15:55 <MDude> Which isn't the same as smiley faces in the message.
16:16:11 <tswett> It starts out as plain BF, though.
16:16:17 <b_jonas> Lyka: http://esolangs.org/wiki/BF_Joust
16:16:31 -!- sebbu has joined.
16:16:45 <b_jonas> tswett: though it has invalid syntax in it
16:17:09 <b_jonas> still, very nice generator
16:17:17 <MDude> It's actually an IRC client, but adds on a bunch of extra data at the start of each message that looks like junk to every other type of client.
16:17:19 <tswett> Most of its attempts at generating BF are complete and total failures. One example:
16:17:20 <tswett> commanddelerval-fefumer: - fixing ( = Unill gon's Decidabred machine (.aboutge] Change congasmuct" deploxed U PHHHMVh.Tancomp and % numacoin \n doiv `befuige*!
16:17:35 -!- sebbu has quit (Changing host).
16:17:36 -!- sebbu has joined.
16:17:40 <MDude> So it's no good unless it's only used in channels made just for it.
16:17:50 <tswett> _xxxkxxbfx xmxydhb(\x=W=\l]
16:18:07 <b_jonas> tswett: how large training set do you need? train it on http://www.math.bme.hu/~ambrus/pu/bibul (from http://www.lolcatbible.com/ )
16:18:18 <b_jonas> tswett: I don't recognize what that is
16:18:57 <HackEgo> cyclonltcoin cafligcoin bischefcoin flaguecoin juicoin auraedcoin homerixcoin haliarderanyplcoin gidcecoin flabcoin gitafncoin sartrecoin ossrocoin neufcoin musinicoin pavincoin moncoin bitfdcoin juftcoin extcoin
16:19:04 <b_jonas> it's learnt "numacoin" from that
16:19:27 -!- hilquias has quit (Remote host closed the connection).
16:19:39 <Lyka> http://pastebin.com/1qBqMy3V
16:19:45 <b_jonas> MDude: some random generated words with "coin" suffixed to them
16:19:57 <b_jonas> `coins # new ones all the time
16:19:59 <HackEgo> Argument "#" isn't numeric in int at /hackenv/bin/words line 148. \ coin
16:20:07 <HackEgo> flexcoin fursecoin auxcoin lockgroupcoin gregicoin x-dcoin vercoin peocoin gufcoin buterixcoin arrancoin lenthoulocoin cludcoin preadcoin intercoin gentcoin phocoin rseyistcoin migmcoin senbercoin
16:20:43 <tswett> b_jonas: well, getting a new neural net takes about 15 minutes of computing time, and then the English looks more like more or less random keyboard mashing.
16:21:04 <b_jonas> tswett: sure, I'm just suggesting. I don't even know if that's a large enough corpus
16:21:38 <MDude> What's it trained to do?
16:21:45 <Lyka> like the code i pasted?
16:21:50 <tswett> So the result would probably be a bunch of lines, each consisting of a number, then some mashing, then a double hyphen, a book of the Bible (probably a real one), and another number.
16:22:12 <b_jonas> tswett: you can strip the numbers and books of the bible if you want
16:22:28 <b_jonas> it's easy to add a number and book separately
16:22:48 <tswett> MDude: the neural net? Well, have you seen https://karpathy.github.io/2015/05/21/rnn-effectiveness/?
16:22:58 <tswett> You give this guyfolk's program some text and it produces some similar text.
16:23:13 <tswett> It's managed to create some really realistic-looking Linux source code.
16:23:30 <Lyka> does it compile?
16:24:53 <MDude> Nah, I'll have to go read that.
16:25:21 <tswett> Lyka: no, because it's not smart enough to keep track of what identifiers are in scope.
16:25:30 <tswett> Apart from that, the syntax is almost totally correct.
16:26:47 <MDude> But I was more wondering if it was made to produce text that specifically acts as replies to statements given to it, or just take random nubmers and turn them into arbitrary text.
16:26:56 <MDude> Or maybe something else.
16:27:21 <b_jonas> sure, to generate source code you could start from hard-coding the rules of the syntax of the language, scoping rules, and the type system, and whitespace conventions, and then train to generate realistic-looking trees under that constraints. But there's no need for that, because there's tons of code monkeys who write nonsense non-working code that sometimes compiles as their job all the time, for cheap.
16:28:13 <tswett> MDude: you could make it specifically reply to statements given to it.
16:28:32 <tswett> Just say that the text begins with a given line of IRC chat, and ask it what the next line of IRC chat would be.
16:29:13 <Lyka> brb is the most common line of irc
16:29:31 <tswett> Yeah, the program generated dozens of lines of fake Linux code with, as far as I can tell, one syntax error, which was an unmatched closing parenthesis.
16:29:47 <Taneb> Lyka, in other channels I can imagine that, but not in here
16:30:05 <b_jonas> I thought "hello" or "hi" or "is anyone here" or "I have a question" or "I have a doubt" or some such stuff is the most frequent
16:30:36 <Lyka> i am a jelly donut
16:31:15 <Lyka> well, i am not, but jfk was
16:33:42 <tswett> I wonder what determines whether or not a president becomes an initialism.
16:33:47 <tswett> Everyone knows about JFK and FDR.
16:34:03 <tswett> But who's heard of, say, BHO?
16:34:29 <tswett> President Bernard Hussein Oliver.
16:35:24 <b_jonas> tswett: well, FDR was very famous because of the world war, I believe
16:36:00 <tswett> I associate him primarily with the Great Depression.
16:37:21 <b_jonas> also, FDR is famous because he used to have a very famous square named of him in Budapest, the one at the Pest side of the Lánchíd, plus has a couple of other war-related monuments elsewhere in the city.
17:02:23 <tswett> Interesting output from the round 3000 version:
17:02:25 <tswett> 23:46:25: <fungot> `rul <backk
17:02:26 <tswett> 21:55:46: <HackEgo> "rwi28xxxx83�t"ieh__~d[n[in]])|[sprib=');
17:03:34 <HackEgo> /home/hackbot/hackbot.hg/multibot_cmds/lib/limits: line 5: exec: rul: not found
17:07:24 <tswett> Aha, it *did* learn the coins:
17:07:26 <tswett> 12:57:24: <HackEgo> ternacoin pdoccoin vercoin pointcoin iacoin raincoin deascoin erscoin %22xpcoin(concoin sh"libracoin bodycoin sulasmesologcoin negolaintecoin imbercoin byncoinclitciincoin ilagecoin bianneceincoin rordgcoin motancoin mbrecoin bunlogicoin mrgicoin 9pdgeccoin babilcoin ^masking signalify and matchenestingcoin vebbcccoin sinumcoin barggcoin dainecoin
17:07:26 <tswett> iscoin /ballellist/showd that instead of ‘mat’s’
17:07:42 <tswett> But it doesn't get the color rotation right.
17:08:21 <b_jonas> I don't know how you did it, but nice
17:10:30 <tswett> More nice output from the low temperature version:
17:10:32 <tswett> 19:11:29: <ais523> right, it shouldn't be a lot of the way to go anything that makes sentence, and figure out some more baming arguments
17:12:28 <tswett> It pretty much has a good handle on long words and now it seems to be starting to understand syntax a bit.
17:16:19 -!- jix has quit (Ping timeout: 244 seconds).
17:32:00 -!- Wright has joined.
17:43:45 * Taneb has received a letter from the landlord which misspelt every single tenants' name
17:45:12 <Taneb> With the classic "van Doom" for me
17:52:40 <Phantom_Hoover> bunlogicoin replaces the traditional cryptographic approach to electronic currency with bun logic
18:01:19 <fizzie> [2095966.189027] sd 2:0:0:0: [sdc] Very big device. Trying to use READ CAPACITY(16).
18:01:23 <fizzie> Come on, it's only 3 terabytes. That's not "very big" in this day and age.
18:02:45 <b_jonas> maybe you're using an old kernel or something
18:10:54 <tswett> Remind me why I'm using my laptop for all this heavy computation.
18:11:04 <fizzie> https://github.com/torvalds/linux/blob/HEAD/drivers/scsi/sd.c#L2150 <- still there.
18:23:08 -!- Tritonio has joined.
18:25:39 <HackEgo> [wiki] [[User talk:Lucasieks]] http://esolangs.org/w/index.php?diff=43034&oldid=40025 * 73.184.106.177 * (+71) /* ():;#?! loops */ new section
18:30:41 <b_jonas> Question. If you take the pantheon of ancient greek gods, then which god is most suited as the patron god of programmers? Hephaistos is the god of engineers, so he's a candidate, but is he the most approperiate?
18:35:35 <Lyka> who's the one who pushed a huge boulder up a cliff;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
18:35:50 <Lyka> ignore the ';'s
18:39:57 <int-e> Yeah, modulo random spelling (Sisyphus).
18:41:07 <Tritonio> b_jonas Hephaistos was supposed to have created robots too so I guess he's the best bet.
18:46:42 <Lyka> okay, bad rush reference
18:47:44 <b_jonas> Tritonio: for robots and railways, definitely, but I'm not sure about programming. Were the robots intelligent? If so, were they driven by spirits, or by a clockwork construct brain?
18:49:13 <b_jonas> He is also definitely a hacker, according to the way he's catched her wife on cheating… hmm, how was that story?
18:49:42 <b_jonas> Lyka: Prometheus? Why him in particular? I don't know enough about mythology.
18:50:18 <b_jonas> Oh, I think Hephaistos has also made some funny throne for Zeus or something.
18:50:30 <Taneb> You could interpret Prometheus as a deity for open-source
18:52:19 <b_jonas> aha, apparently Prometheus has taught arts and crafts to humans (sort of like Aule)
18:55:01 <b_jonas> Plus Hephaistos has apparently made like most of the artifacts created by the greek gods.
18:55:58 <b_jonas> The most powerful of which is probably the Girdle of Aphrodite
18:56:13 -!- ZombieAlive has joined.
18:59:16 <b_jonas> (other names for search, Hephaestus, Vulcanus)
19:05:28 <b_jonas> Lyka: Vulcanus is the name of the corresponding deity in the Roman mythology
19:05:52 <b_jonas> and his wife Aphrodite is called Venus in roman
19:06:20 <b_jonas> Hephaestus is one of the crazy random English spellings
19:07:59 -!- hilquias has joined.
19:10:32 <b_jonas> Phantom_Hoover: nah, there's a very easy and consistent way to write the names of Ancient Greek deities, which is to look up how they're spelled in the Odysseia or the Illiad and transliterate them. But the people using English don't care, and just use random transliteration for everything.
19:11:00 <Phantom_Hoover> that's an easy way to transliterate the way they're spelt in the odyssey, yes
19:11:35 <tswett> I think the round 4000 neural net pretty much has the local aspects of syntax down. It seems to do about as well as a Markov chain when it comes to syntax.
19:11:53 <b_jonas> Mind you, this isn't specific to the English language: journalists writing in Hungarian also use random crap to write proper names from arabic, korean, chinese, japanese, and other foreign languages.
19:12:18 -!- TieSoul has quit (Read error: Connection reset by peer).
19:12:21 <Phantom_Hoover> it has roughly no relevance to the equivalents of the names used in english
19:12:25 <b_jonas> They usually take random transliterations they find in English or other european language articles, and modify random letters in them.
19:12:44 <b_jonas> Without regards for the original name.
19:13:04 <Phantom_Hoover> it's great, i don't understand why people get so angry about it
19:13:52 <b_jonas> Phantom_Hoover: the problem with it is that you often get five or more transliterated variants of the same names, and they usually don't even bother to attach the original name in parenthesis so you can't figure out which names refer to the same entity.
19:13:56 <Phantom_Hoover> languages are chaotic living things, you can't pin them down with formal specifications
19:13:57 <b_jonas> You can't search for them.
19:14:20 <b_jonas> Phantom_Hoover: oh we sure can. We have formal rules for transliterating from most of the popular languages of the world.
19:15:14 <Phantom_Hoover> b_jonas, and you need only look at chinese to see how that doesn't really pan out in real life
19:15:51 <b_jonas> Phantom_Hoover: we have rules for transliterating from Chinese too.
19:16:30 <Phantom_Hoover> which have changed in prominence quite a lot over the years
19:16:37 <b_jonas> It does work out for sane people, but not for journalists who generally don't care about the truth or usefulness of what they write, as long as they write it real fast.
19:18:18 <pikhq> Phantom_Hoover: There's rules for transliterating, but it's also common to utterly ignore it and make random shit up.
19:19:22 <pikhq> Of course, it gets even crazier when dealing with cultures where the person is likely to have a preferred Roman alphabet representation that may or may not be a valid transliteration at all.
19:20:07 <pikhq> e.g. Nisio Isin (NISHIO Ishin).
19:20:37 <Tritonio> b_jonas I think that Hephestus and Athena taught arts/healing etc to humans. Prometheus was the one who initially gave them the "fire" (literally and metaphorically probably) and got them out of the dark ages. Taneb's association with open-source is almost on spot. :-)
19:22:06 <pikhq> Or say Jackie Chan (Sing4 Lung4; Chéng Lóng; 成龍; 成龙).
19:22:13 <b_jonas> Tritonio: fire, but also other stuff. But yes, other gods have probably also taught them.
19:24:53 <b_jonas> Who invented writing systems or alphabets or abjads according to Greek Mythology? Or who taught it to mortals?
19:25:57 <Tritonio> never heard of anyone teaching them "letters" but promethius also gave them "knowledge" in general...
19:26:16 <b_jonas> it could be "writing", not specifically "letters"
19:26:47 <b_jonas> there's bound to be a greek myth, there's one in every other mythology
19:27:08 <Tritonio> as I understand it prometheus was the one who simply gave humanity what the gods had and humanity lacked. So he pretty much was the one who enabled them to stop been like other animals.
19:27:39 <Lyka> fire can be a metaphor
19:27:41 <b_jonas> I don't know if that includes writing though. I'm not even sure if the greek gods even used writing much.
19:27:52 <b_jonas> Lyka: sure, it's definitely not only fire that Prometheus gave them.
19:28:00 <b_jonas> but letters could be a later invention
19:28:08 <Tritonio> yes but it's also literal. fire is what allowed humans to stay awake at night and talk. and what allowed them to fear way less animals.
19:28:16 <b_jonas> the gods might not _need_ writing, because they're, like, immortal
19:28:18 <Tritonio> without fire humanity is like other animals.
19:28:45 <tswett> There are three nonlocal syntax errors I notice this neural net seeming to make: a sentence with the wrong number of verbs; a noun phrase that doesn't attach to anything; and a sentence which starts like a question and ends like a declarative, or vice versa.
19:29:04 <tswett> I look forward to seeing if it actually learns not to do those.
19:29:42 <tswett> Tritonio: I'm training a neural net using this guyfolk's program: https://karpathy.github.io/2015/05/21/rnn-effectiveness/
19:29:52 <b_jonas> tswett: how much specialized knowledge did you try to add to it?
19:29:56 <tswett> I'm training it on the logs from #esoteric.
19:30:09 <tswett> b_jonas: specialized knowledge? I just gave it the #esoteric logs and a couple of numbers.
19:30:27 <b_jonas> I mean domain-specific knowledge you hard-code, rather than make it learn
19:32:45 <tswett> I didn't write any code at all.
19:32:49 <Tritonio> tswett when you feed it latex it's like it's trying to generate something like the Voinich manuscript.
20:25:02 <b_jonas> Oh, it makes sense in hindsight. Forth compiled to x86 or x86_64 must use the stack pointer (ESP or RSP) to point to the return stack, not the data stack, because modern cpus have special support for predicting return targets perfectly if you use the CALL and RET instructions for them, whereas for the data stack, compiler optimizations together with cpu magic like register renaming makes handling the data stack just as efficient, if not more so, than if
20:25:15 <b_jonas> (truncated) whereas for the data stack, compiler optimizations together with cpu magic like register renaming makes handling the data stack just as efficient, if not more so, than if you used stack-specific instructions, and definitely more efficient than if you mixed the two.
20:49:21 <zzo38> Now I added Impulse Tracker into the wiki of AmigaMML.
20:49:33 <pikhq> b_jonas: Eh, not "must", but certainly "really should".
20:50:16 <pikhq> If you really felt like it you could use esp as a GP register, but it'd be dumb. :)
20:52:23 <b_jonas> using RSP as a general register isn't dumb, as long as you don't use it as an offset much. it's completely reasonable if you don't need RSP for anything else. it's representing the return stack in some other way that's dumb, because it will screw up branch target prediction, and compiled forth probably has lots of subroutine calls.
20:54:35 <pikhq> It's a terrible idea if you're expecting interrupts or signal handlers.
20:54:51 <pikhq> (because other code is going to use that as the stack)
20:54:55 <b_jonas> in x86_32 and x86_64, RSP works fine as a general register, if you consider four limitations: (1) encoding it as an offset is usually impossible, (2) mixing PUSH/POP operations with other ways to access RSP quickly can have a latency on some cpus, and (3) by default on most operating systems userspace, signals will start a stack frame on the stack at or somewhat below where ESP points to, so it can trash your stack, but you can disable this,
20:55:23 <pikhq> That condition is also true on a bare system with interrupt handlers.
20:55:32 <pikhq> Of course in either case you can work around it.
20:55:34 <b_jonas> and (4) I think in some cases in privilaged mode in systems code, you can't use an alternate stack for interrupt handlers at all (you can for user mode).
20:55:50 <pikhq> It's *tricky* but totally doable.
20:56:38 <b_jonas> The problem is that x86_32 has too few general registers (also too few vector registers), and very rarely you can run out of them even on x86_64, so in some tight loops it _can_ be worth to take the extra steps to use RSP for something else.
20:57:46 <b_jonas> However, it's worth only after you also free up RBP and the register holding the PIC-helper pointer.
20:58:04 <b_jonas> Recent compilers can automatically do the latter two actually.
20:59:31 <pikhq> Yeah, freeing up RBP is pretty standard, freeing up the PIC pointer is sadly really new stuff.
21:00:13 <b_jonas> Gcc allegedly 5 does some crazy optimization stuff, including sometimes spilling locals from general registers to vector registers instead of the stack, which is a brilliant stroke in retrospect.
21:00:19 <pikhq> (I have no idea why you wouldn't just put that into register allocation with an ABI requirement that it be in there across translation unit boundaries)
21:01:09 <b_jonas> Of course, you have to be careful with that, because even on x86, moving between general purpose and vector registers can have a large latency, in at least one direction.
21:03:48 <b_jonas> Sometimes I think the current naming scheme is wrong, and the vector registers should be called "general purpose" or "data" registers, and the general purpose registers called "index" or "counter" registers.
21:05:20 <b_jonas> Sort of like how X and Y are index or counter registers on the 6502 because you can do fewer arithmetic on them than on the general purpose A register.
21:08:13 <Lyka> why am i still connected to my bouncer?
21:10:28 <Lyka> i mean, why is my irssi window still open
21:10:57 <oren> irssi window? you mean the terminal?
21:10:59 <Lyka> why did i not close irssi
21:11:36 <Lyka> this is wht i should have done hours ago:
21:11:40 -!- Lyka has changed nick to Lyka|Away.
21:13:04 <oren> I don't see how renaming yourself helps?
21:13:50 <oren> \nick orin|away
21:13:53 -!- oren has changed nick to orin|away.
21:14:22 -!- orin|away has changed nick to oren.
21:15:10 -!- Weloxux has quit (Quit: Leaving).
21:17:35 <tswett> I assume that's not the only thing Lyka did.
21:18:41 <oren> maybe she has hooks into irssi to change her name on SIGHUP
21:22:50 <tswett> I think, but I'm not sure, that the neural net is beginning to understand question marks.
21:23:11 <tswett> There seems to be a tendency for questions to end with question marks, and for non-questions not to end with question marks.
21:24:57 <int-e> oren: it could also be some sort of irc bouncer
21:25:15 <int-e> still this idea of changing nicks when away is annoying.
21:25:25 <b_jonas> int-e: he specifically said "connected to my bouncer", yes
21:26:00 <int-e> irc even has a /away command for the purpose :-/
21:26:59 <zzo38> If you use WHOIS or PRIVMSG send message to them then the AWAY is checked by the 301 line.
21:27:08 <tswett> It *might* be learning not to do unattached noun phrases, but that seems unlikely. It's doing really well with parentheses.
21:27:26 <b_jonas> zzo38: is it for a private message? I didn't know that
21:27:38 <Taneb> In the spirit of the Eurovision song contest, I'm not going paying for my alcohol
21:27:44 <b_jonas> I thought you'd just get the away status from a WHO and then follow it from AWAY messages,
21:28:08 <int-e> "With the AWAY message, clients can set an automatic reply string forany PRIVMSG commands directed at them (not to a channel they are on). The automatic reply is sent by the server to client sending the PRIVMSG command. The only replying server is the one to which the sending client is connected to."
21:28:20 <b_jonas> and presumably you have to WHO everyone newly joining if you want to follow that
21:28:32 <b_jonas> I don't think it's worth the bother, I don't care about away status really
21:28:46 <b_jonas> int-e: ah, so that's what it's for, right
21:29:00 <int-e> the point is, unless you're privately messaging somebody, you're unlilely to care...
21:29:44 <b_jonas> I don't set away status this days.
21:30:26 <b_jonas> It's useless because even if I'm not away I don't read everything nor answer to everything.
21:30:51 <tswett> It also understands that the first digit of the timestamp rarely changes.
21:36:04 -!- adu has joined.
21:42:52 -!- bb010g has joined.
21:43:10 -!- fractal has quit (Ping timeout: 272 seconds).
21:44:08 <fizzie> A number of channels plain forbid public away messages and nick changes.
21:48:01 <Phantom_Hoover> i've been in channels with people who have scripts set up that post messages saying exactly how long they've been away for
21:48:53 * int-e has been away for 199214411ns, changing his nick.
21:48:54 <oren> you're so vain, you probably think this away message is about you
21:49:17 <Taneb> Reminds me I was in a quite active mailing list with someone who set up an "I am on holiday" autoreply
21:49:45 <Taneb> Then got back and complained about how high-traffic the mailing list was
21:50:16 <oren> if several people did that, it could cause, uh, problems
21:50:47 <oren> > I'm away >> I'm away >>> I'm away
21:50:49 <lambdabot> Not in scope: data constructor ‘I'm’ Precedence parsing error
21:50:49 <lambdabot> cannot mix ‘>>’ [infixl 1] and ‘>>>’ [infixr 1] in the same infix ex...
21:51:14 <int-e> lambdabot: nice error, thanks.
21:51:32 <Phantom_Hoover> (i didn't know about email storms about recently, reading the wp article and linked articles was hilarious)
21:52:04 <int-e> hmm, are there any email to wikipedia gateways?
21:52:24 <int-e> (where, say, emails get appended to a wikipedia talk page automatically...)
21:55:27 <oren> The resulting storm of 'unsubscribe', 'me-too' requests, sarcastic facepalm images and recipes for broccoli casserole resulted in
21:56:10 -!- nortti has changed nick to rjunhrak.
21:56:23 -!- rjunhrak has changed nick to nortti.
21:56:59 -!- nortti has changed nick to nortiecat.
21:57:41 -!- nortiecat has changed nick to nortti.
22:07:38 -!- fractal has joined.
22:31:27 <Taneb> This is a really close Eurovision
22:31:50 <Taneb> Normally there is a clear winner again
22:33:04 <Taneb> Although Sweden has now jumped ahead of Russia
22:33:17 -!- rdococ has joined.
22:41:04 -!- adu has quit (Quit: adu).
22:46:12 <fizzie> But they did call it later than usual.
22:46:53 <fizzie> I don't really really understand why Russia got that many votes.
22:47:16 <fizzie> Or Italy, for that matter
22:47:40 <Taneb> I quite liked Italy's
22:47:46 <Taneb> (in fact, I voted for it)
22:47:51 <fizzie> I mean, nothing wrong with their shows, but they weren't so remarkable either.
22:48:43 <fizzie> I don't think I've ever voted, so I'm not technically entitled to an opinion, though.
22:49:16 <fizzie> I liked the background video stuff Sweden had.
22:49:38 <fizzie> The BBC guy doesn't seem terribly happy about the UK score.
22:50:16 <Taneb> But hey, we beat France again
22:50:32 <Taneb> fizzie, we've never won with him commentating
22:50:54 -!- oren has quit (Read error: Connection reset by peer).
22:51:00 <fizzie> When did you last win?
22:51:46 <fizzie> The "big five" were all quite far down on the list, with the exception of Italy.
22:53:56 <fizzie> (We didn't even make it to the final.)
22:54:18 -!- oren has joined.
22:55:11 -!- Patashu has joined.
22:57:12 <zzo38> Can you tell me if you think I missed anything in this description of rules of Pandante game? http://zzo38computer.org/textfile/miscellaneous/sirlin/pandante/rules
22:57:19 <zzo38> (The variant rules are in a separate file, and is not completed yet.)
23:03:59 -!- augur has quit (Quit: Leaving...).
23:31:24 -!- Lyka|Away has changed nick to Lyka.
23:32:50 <Lyka> what kind of language has this code: !A048A149P000P100Q000@
23:32:59 <Lyka> and gives this output: HI
23:34:42 <Lyka> is my attempt at making an interpreter for the arduino pathetic enough?
23:35:27 -!- oerjan has joined.
23:36:06 <Lyka> http://pastebin.com/jGJX8A6D
23:38:35 <Lyka> oerjan: to get it to output the two characters "HI", i used the code: !A048A149P000P100Q000@
23:40:02 <tswett> What are the ASCII codes for H and I?
23:42:21 -!- Frooxius has joined.
23:42:40 <Lyka> five commands wer given: Assign 0x48 to cache byte 0, Assign 0x49 to cache byte 1, Output cache byte 0 to terminal, Output cache byte 1 to terminal, Quit
23:43:59 <Lyka> P only uses one argument, Q uses none, but still had to give all three arguments to each command
23:44:52 <Lyka> P000Q000 could in theory be simplfied to P0Q, but that is n the future
23:44:55 <tswett> Word that the neural net invented: "shitation"
23:45:36 <Lyka> so, is my language making any sense yet?
23:46:17 <Lyka> i haven't added arithmetic yet
23:47:45 <Lyka> A002A102+012H200 would output "04"
23:48:17 <Lyka> +012 being: a[2]=a[0]+a[1]
23:48:47 <Lyka> and H200 being same as P200 but outputting in hex
23:49:15 -!- Phantom_Hoover has quit (Read error: Connection reset by peer).
23:49:33 <oren> how many bytes does the cache hae?
23:49:34 <Lyka> H command and arithmetic command come next
23:49:39 <tswett> Some of the latest neural net output:
23:49:41 <tswett> 13:14:59: <oerjan> i've never heard of the north
23:49:41 <tswett> 15:07:53: <int-e> so far
23:49:41 <tswett> 13:14:07: <oerjan> argh
23:50:09 <tswett> I'll try not to spam the channel with these too much from here on out.
23:50:33 <Lyka> *arithmetic commands
23:50:51 <Lyka> the memory has 256 bytes
23:51:10 <Lyka> but you have to move things into cache to use them
23:52:32 <Lyka> mae any sense so far?
23:53:27 -!- hilquias has quit (Remote host closed the connection).
23:53:56 <Lyka> understand i am trying to make this complex enough to make a game in
23:54:18 <Sgeo> Is #ubuntu the only correct place to ask for ecryptfs help?
23:54:51 <Lyka> Sgeo: why ask here?
23:55:29 <Sgeo> This is the IRC help channel, right?
23:56:08 <Sgeo> IRC is a rather esoteric medium, so yes
23:58:14 <Lyka> oh, sorry. eye-rolling is genetic in my family