log☇︎
287400+ entries in 0.181s
asciilifeform: simply that any improvement i expect to see will not be qualitative.
mircea_popescu: what's this to do with your defense of dumb ocr ?
asciilifeform: correcting several MB of ocr text against original.
mircea_popescu: when you have to correct on average 1 error per book, which is where i expect this can be taken, it's just a breeze.
asciilifeform: i can apparently tell that mircea_popescu has never tried it
asciilifeform: doing this correctly is ~same amount of work as manually ocring.
mircea_popescu: there's worse fates to library text than being read once.
mircea_popescu: phf incidentally, how interested are you in the unholy art of probabilistic models ? it occurs to me that making a proper ocr would be a most respectable task, but not really for everyone.
asciilifeform: but that no one will improve on it to the point where i can throw out the original TB and be ~certain~ of no irrevocable loss.
asciilifeform: my contention is not that it is physically impossible to improve on extant ocr
mircea_popescu: they work a lot the fuck better than dumb ocr present currently.
mircea_popescu: same principle as in machine translation, search engines, etc.
asciilifeform: no srsly nobody's gonna proofread a TB of ocr.
mircea_popescu: and stop with the romanticisms already.
asciilifeform: realize that when you have more than a little meat sweat mixed in, you are no longer discussing a computer program
mircea_popescu: makes ten trillion parameters, sets them all to 0, then sets more and more to values
mircea_popescu: asciilifeform self-trained like the markov nonsense works.
mircea_popescu: "hm, i wonder how come these guys keep speaking about modems in non-modem context".
mircea_popescu: if proggy has no fucking idea modems don't go with macbeth a) it's not trained and b) everyone involved is fired.
asciilifeform: if yours does, it is because you tweaked it
asciilifeform: proggy has nfi that shakespeare didn't mention modems.
asciilifeform: think about the kind of persistend turd that one gets with ocr
asciilifeform: problem is that 'fix' is not always mechanically evident, even.
mircea_popescu: self-trained, goes back and fixes typos.
mircea_popescu: there's really no reason a 100% accuracy over 1mb text ocr can't be had today.
asciilifeform: (a turd on every 8th page on avg.)
asciilifeform: even for FICTION it is 99.99% accurate which is to say SHIT
mircea_popescu: i do not think so.
mircea_popescu: more hopeless than ai for go ?
asciilifeform: yes! quite like this.
asciilifeform: (thing is per se almost like ocr in terms of sheer bowel-loosening turd-shedding)
mircea_popescu: i had an issue with an ocr that was TOO informed, kept creating single glyph fl and other wonders.
asciilifeform: this is annoying but the 100x mass saving pays off.
asciilifeform: by transforming 'similar' letter into another.
asciilifeform: incidentally djvu compressor is 'too good' and makes ocr quite a bit more difficult - it sometimes INTRODUCES typos
mircea_popescu: ocr is really a good fit for the markov chains as ai they do
mircea_popescu: google has the patience to fuck with go, hasn't the patience to fix ocr ?
mircea_popescu: that's another thing, must have MUCH BETTER ocr.
asciilifeform: (ocr, contrary to what you've been misinformed, doesn't actually work)
asciilifeform: ain't nobody gonna ocr TB of ru diagrammed b00kz.
mircea_popescu: stum. the sweet juice of grapes.
mircea_popescu: scans as bitmaps are ~in the situation of wine as stum.
asciilifeform: likewise i have a TB+ of scans here. these can exist as bitmap and naught else.
phf: mod6: there was a hosting provider guy behind the domain cock.li, was interviewed by mp, was selling a vps box access
asciilifeform: latex doesn't survive ill-forming any better than html.
mircea_popescu: seriously, this is allowed, "graphics" that aren't svg and text that's not tex ?!
mircea_popescu: incidentally why the fuck is the web not native latex.
mircea_popescu: none of this is acceptable, intellectually.
mircea_popescu: yeah, it's vaguely inhuman, this. the scholar in me is constantly opressed by various shits. 1. "oh we don;t alphabet - pdf" 2. "oh you can't reference text snippet" 3. "oh we forget, what was this ? linkrot ?" and i think there's more
asciilifeform: mircea_popescu: it was where i started - text. and from that ended up in 'computing is braindamaged beyond repair'
asciilifeform: mircea_popescu: sorta what the first article on my blog was about.
mircea_popescu: there's a serious problem with digital handling of text, as we have it atm.
mod6: <+asciilifeform> mod6: that thread wasn't even about living conditions, but about being a hermit << saw that. figured, "hey, i've got a crypt he can live in if he wants..."
mircea_popescu: trinque tbh, i think wikis are a (braindamaged, dysfunctional, uncomprehending) response to the html-is-broken / transclusion issue discussed yest and etc. ☟︎
phf: i have that cockli box by the way. i'm probably going to spin up an instance on it, but if anybody wants to attempt that task, feel free to take over ☟︎
trinque: if they are in my wot, sure
asciilifeform: how to edit ? folks get shell on the box ?
trinque: I don't see how it's better than html shat in a folder somewhere
asciilifeform: mod6: that thread wasn't even about living conditions, but about being a hermit
mod6: so it works for the time being.
mod6: <+trinque> sometime I could hack another field onto it; I'm not certain I like the thing (or wikis) << I don't love wikis either, but we need a place for quick docs.
mod6: <+trinque> it uses the page title verbatim in the URL << ok i think I can just have shinohai change the page title then, thx!
mod6: offer stands if you ever get sick of the the 202
mod6: you think i live in a teepee?
trinque: but we must include them
trinque: they're really a response to "tards can't into html"
trinque: sometime I could hack another field onto it; I'm not certain I like the thing (or wikis)
mod6: <+asciilifeform> mircea_popescu: i lived in a ~hole for 20 yrs. << you can come live in my basement if you're tired of paying rent. i've even got a garden. :] ☟︎
trinque: it uses the page title verbatim in the URL
trinque: mod6: wiki is the code that runs cliki.net
asciilifeform: mod6: this is not even wholly improbable. recall the selection criteria for dingledinity
mod6: he has a van that says 'free candy'
trinque: like the child-fucking kind
mod6: mircea_popescu: btw, did you want me to change tb0t to use '!' instead of '%'?
mod6: <+asciilifeform> (though almost any imaginable mistake would result in MORE-phuctorable mods, vs less) << yeah this seems odd.
a111: Logged on 2016-06-28 00:31 mircea_popescu: there is nothing dingledine can ever do that can redeem him, because the manner in which he thinks is beyond salvation.
asciilifeform: http://btcbase.org/log/2016-06-28#1491739 << same dingledine as officially represents nsa in tor ☝︎
mod6: damnit. the spacebar on this thing is getting toast.
mod6: trinqueis there a way to update this URL so it doesn't contain spaces? i.e.: http://wiki.deedbot.org/The%20Real%20Bitcoin
mircea_popescu: phf it's probably the "fire retardant" retardation
mircea_popescu: where did all teh junes go... loong time passing...
mod6: time for me to work on the SoBA already!
BingoBoingo: Milestones in pipe wrench ownership #2: When loosening a slip fitting the pipe decides to tear instead
phf: ikea sells chairs and chair covers for those chairs. can buy covers in uk, etc. but not in u.s. fancy that
mircea_popescu: he must be killed, publicly, and the carcass left to rot where it fell.
mircea_popescu: there is nothing dingledine can ever do that can redeem him, because the manner in which he thinks is beyond salvation. ☟︎
mircea_popescu: anyway, this paste is a point-by-point example of sinful, evil behaviour.
mircea_popescu: i know, they constantly try to wander in here also.
asciilifeform: btw ~every american outfit bigger than a hot dog stand has an in-house 'shari'.
asciilifeform: insurgency theatre.
mircea_popescu: Framedragger is this one of those guys you're friends with, then ? ☟︎
mircea_popescu: fuck these idiots with hot pokers already omfg.
mircea_popescu: usg agent wondering if the actual people fighting the usg are a "cointelpro style" "attack" ?
mircea_popescu: "we won't bury the corpse until nsa provides a new trough"
asciilifeform: ons. Tor, and our broader freedom movement, are about communities, and about how we want to interact with each other. Let's handle this situation as an example of how we want to do it right.'
asciilifeform: 'It's tempting to wonder if there's some cointelpro-style attack going on. Realistically, we likely do have the attention of governments who are well-funded to attack us. But first, this really doesn't look like a cointelpro op. The complaints come from people both inside and outside the Tor community, and I know some of them. And second, in this case it really doesn't matter. It's no excuse for not taking responsibility for our acti
mircea_popescu: " Things are probably going to get worse before they get better." << ahahahaha.
mircea_popescu: i don't think you fully appreciate just how braindamaged this "laughing gas of the future" actually is.
asciilifeform: not really problem if each node only has to verify tx against list of length 1
mircea_popescu: but then the problem'd have been obvious ?