#ossasepia | 2019-09-20

← 2019-09-19 | 2019-09-21 → ↓

lobbes: damn whaack, dat connection. You ever consider shared hosting with pizarro? Best pipe I've ever experienced in terms of lack of drops (can't remember the last time my auctionbot disconnected)

lobbes: slap an irssi on it, and boom, problem solved

diana_coman: whaack_pura_vida: ffs, what are you doing there? do I need to kick you out so you set it up properly or what?

diana_coman: http://logs.ossasepia.com/log/ossasepia/2019-09-19#1003117 - sounds like you crack the bash+curl+diff, well done; do the write-up and publish it on your blog, who knows what it grows into/when you'll need to ref back to it.

ossabot: Logged on 2019-09-19 17:31:02 shrysr: diana_coman: hmm - i tried the raw log for id > 1000000 - both visual and diff inspection seem to indicate they are still the same, i.e in reference to http://logs.ossasepia.com/log/ossasepia/2019-09-19#1003089 ?

diana_coman: and yes, the differences will start appearing only at the point where each bot relied on his own recording of lines.

diana_coman: !o uptime

ossabot: diana_coman: time since my last reconnect : 1d 0h 27m

jfw: http://logs.ossasepia.com/log/ossasepia/2019-09-19#1003081 << I was; specifically 'I will get back to i

ossabot: Logged on 2019-09-19 12:37:02 diana_coman: jfw: weren't you saying something about getting back on Wednesday aka yesterday?

jfw: t on Wednesday', by which I meant the looking around. Which I did. But I confess it occurred to me I should at least post a "here I am, reading", and it now seems to me I certainly should have since I brought the day up in the first place, so for neglecting that I'm sorry.

diana_coman: jfw: aha, a sign of life is all it takes; no worries; and ask questions whenever things are not clear.

jfw: I worked on catching up with the current log from were I'd left off, but didn't make it through the 18th, a pace that has me worried about finishing a full readthrough any time soon. I know I have room to improve in staying focused. If anyone has suggestions on eating log faster (while still comprehending adequately) I'm all ears

diana_coman: jfw: so how many words/hour did you get to read there or how do you measure it?

diana_coman: jfw: re focused - do you mean while reading (ie you start reading and then just wonder off) or what?

jfw: haven't figured words/hr here, probably should. Yeah, mind wanders

diana_coman: jfw: is this usual for you (ie on ~anything you read) or just log-specific?

jfw: depends on the material I think. Might just be I tried to pack in too much this time

jfw: I've previously measured ~1k words/hr on #trilema by the btcbase word counts, with some notetaking

diana_coman: jfw: well, "mind wanders" when specific like that to one thing is essentially "not engaging" ; now if it's not engaging because too tired or because doesn't yet make sense, only you can tell at this stage

diana_coman: iirc you had a question re what does it mean to train here - take the question and read the log with this goal ie as you read try to extract and refine gradually the answer to this question

diana_coman: it helps to give you some grounding and a path to engage the whole

diana_coman: jfw: I wouldn't have thought that #o logs are harder to read than #t but dunno - do you find them harder to read?

diana_coman: (there were a few days of long logs in here for sure, but not really *all* days are that long)

jfw: Not harder, no, but perhaps feeling a need to be more thorough

jfw: Heading to bed shortly; I'll be at it again tomorrow and will see if I can be more observant of what brain is up to.

diana_coman: whaack_pura_vida: !

diana_coman: are you even there?

shrysr: http://logs.nosuchlabs.com/log/ossasepia/2019-09-20#1003128 << yeahh...as you said it was not as difficult as i thought, though curl has a lottaa functions and I havent gone into them at all... write-up already underway in parallel, literate programming style. I went a little further by implementing conditionals and now plugging in a for loop .. never done these with bash. I guess it shd be all done today

snsabot: Logged on 2019-09-20 03:02:46 diana_coman: http://logs.ossasepia.com/log/ossasepia/2019-09-19#1003117 - sounds like you crack the bash+curl+diff, well done; do the write-up and publish it on your blog, who knows what it grows into/when you'll need to ref back to it.

shrysr: and i can get back to analysis 2.

diana_coman: shrysr: you certainly need a for loop if you want to be able to actually compare a full set of lines between some given indices, ofc

diana_coman: re curl - it's not like you need to fully know *all* of it right now, lolz; take it easy, there is time and you'll probably need it again sooner rather than later - when you do, you'll get to know it better and so on.

diana_coman: shrysr: does your bash script have parameters? ie so one can run it ./check 100 100000 logs.ossasepia.com/ossasepia logs.nosuchlabs.com/nosuchlabs ?

diana_coman: as a general approach: you first figure out how to do the thing you want manually (as in one full step of it) and once you got that, you proceed to write the bash script to automate it for the more general case.

diana_coman: as a bonus with the above approach, it has a baked in clear "stop" - since you know what it should do, you also know when it's done and you don't end up wasting your time adding useless "functionality"

shrysr: re parameters, no not yet - i did plan to make it that way though since that seemed most convenient... and thought that wd be the last step when the base is working, i.e pretty much as you said.. i guess you will see the steps in the post as well, started with a single curl and worked up to variables, i.e the log links, istart and iend and conditionals. yes, i am trying to first plot out steps and then

shrysr: stick to them and not bloat it. I've enjoyed it actually, a lot better than following a typical data-hag tutorial tbh.

diana_coman: shrysr: heh; you know, I seriously doubt you'll enjoy actual "data analysis work" as much as you convinced yourself you will.

shrysr: tbh i've shared that doubt since a long time. Re-emphasised by this erp stuff. no system can 'enlighten dumbos entering anything they like. ML is the only part that is actually exciting, or even the backend stuff they call data engineering nowadays is better than data analysis. tbh: as i've thought more abt it.. particularly in light of tmsr + eulora - i think i'm no longer.. terribly focused on enjoying

shrysr: it. From the JD exploration so far - it still supports that there are way more companies in the comp sci / analytics field in canada. I've had a lazy eye on mech engg job positings fwiw but you know - nothing has really changed in terms of skills they want and what I've managed to scrape together here...v/s being unemployed before.

shrysr: also as discussed earlier.. vancouver does have R+python significantly higher than calgary, so I've been adding them this week. well, more like calgary has negligible 'R'. Does help some of the companies are based in both calgary and vancouver.

diana_coman: that's fine, the point was simply to not feed illusions re what it'll be like, nothing more.

diana_coman: btw, should probably add that ML is *also* unlikely to be as fascinating as it seems from a distance; the *promise* of ML is great, the reality however is rather... far from the promise.

shrysr: :))... I'm inclined to agree on that as well from what little i know/have done. Relatively fascinating compared to cleaning data!! perhaps its more like human-whipping-machine-to-create-impression-of-some-retarded-learning atm, that too where the human is not retarded? easier to call it ML :D. Perhaps just the dangling carrot ahead of data-cleaning stick. Why use ML n exotic crap anyway if an excel sheet

shrysr: brings ROI? hehe, why use excel sheet if scribbled note sufficient to make money? (was asked this recently). I meant exciting as in - theres some theory atleast, and one can atleast afford to cook up small experiments on relatively affordable gear away from mine... presuming interesting data available. Even if salt-minez dont give a shit...all for fun, that can perhaps lead to something useful ... someday?

diana_coman: shrysr: as you just experience, cooking up interesting stuff takes simply your computer and a meaningful purpose :D

diana_coman: experienced*

diana_coman: the theory - such as it is - is, well, Maths

diana_coman: and yes, sure, it *is* interesting and at any rate, totally useful to be actually able to make sense of things, certainly

diana_coman: just... not much "machine learning", no

diana_coman: the power of the machine as much as there is stands mainly as brute power really

diana_coman: fascinating perhaps, why not; but brute power, without reason; to be used and employed as best as you can think of (and the better you can think of using it, the better for you) but not to be asked to figure things out for you

diana_coman: there is this distinction between data (raw figures of any kind), information (data+context) and knowledge; the machine works with data; perhaps one can argue that in some carefully designed cases, it may attain "information" - or rather mimic it to some degree

diana_coman: however, that's about as far as it goes.

diana_coman: the text mining field is probably a good example in this sense - for all the effort spent and all the various models put forward, the truth is that they all "work" only within very, very narrow spaces and for very limited definitions of "works"; and even that, only *if* carefully fitted (over-fitted really) by a human.

diana_coman: not to bring you down or anything, quite on the contrary, to not feed false hopes that will just come crashing down like a ton of bricks.

diana_coman: I did my Master's precisely on "Data Mining", precisely because "amazing", precisely on "using ML for breast cancer detection because OMG, wouldn't that be great"

diana_coman: it was a good cure indeed but a cure for my naivity regarding the domain, not of that much use to those affected by breast cancer, what can I say.

asciilifeform: diana_coman: the 'machine learning' thing always seemed to me as a rebranded/repainted resurrected corpse of 1980s 'expert systems' fad. was always curious re the exact chronology of this resurrection/rebranding, but never quite enuff to dig

diana_coman: asciilifeform: afaik the expert systems actually made it to some degrees esp in medicine; but it's still pretty much the sad "can search faster through a large pile of match-this-to-that" rather than any machine learning

shrysr: diana_coman: no, i get you. the above is in fact imho the truth, which is almost negligibly conveyed anywhere in the funnel... (which perhaps why its a shit-funnel fwiw). i mean... i got this impression after just initial reading... and comparing to how i've seen. my understanding is data mining quite popular in bio-tech related fields because you cant ethically experiment easily on humans/even animals...

diana_coman: shrysr: and because you need to sift through a ton of data; ie machine-power needed and useful, certainly; now why call that though "machine learning" is a different thing.

shrysr: a guy who with me was among the firsttt to join that DS-dream-job around March 2018 - was a PhD in neuro-science at Calgary uni. iirc post doc for many years too. I wanted to ask him wtf he needed dsdj for...and never got around to it. He recently joined some bank :) intend to get in touch atleast now for tips.

asciilifeform: diana_coman: 'expert system' was enuff of a hype wave that reached sovok, i have a late-'80s ru text on subj. idea was that human expert manually programs a (what would in modern parlance be) 'chatbot' , which then diagnoses disease etc. afaik vanished w/out a trace, other than a small trickle of use in insurance industry today

shrysr: asciilifeform: i believe it is becoming more than a trickle of use today in the insurance industry. its apparently 'cheaper' to get a chatbot to pop out your insurance docs and basic stuff (than run call center)... but ya lol google has some service wherein you build chatbots.. i tried a while ago and essentially exactly the same as you described, feed in Q&A. iirc adobe use alexa in boardroom meetings

shrysr: "alexa, what are the quarterly profits" or something.

asciilifeform: shrysr: the orig item was moar of a 'enter symptoms, get 'which antibiotic' as output' than the current-day 'eliza' clones

asciilifeform: shrysr: are you familiar with the 'eliza' story? imho it is instructive.

shrysr: no? i was just typing what you meant by eliza actualy.

asciilifeform: shrysr: see also

asciilifeform: 1966 (!) . 'Secretaries and nontechnical administrative staff thought the machine was a "real" therapist, and spent hours revealing their personal problems to the program. When Weizenbaum informed his secretary that he, of course, had access to the logs of all the conversations, she reacted with outrage at this invasion of her privacy. Weizenbaum was shocked by this and similar incidents to find that such a simple program could so

asciilifeform: easily deceive a naive user into revealing personal information.'

asciilifeform pretty curious how shrysr managed to learn to program but avoid finding out about 'eliza' -- it was used as example in just about every book on the subj in '70s-'90s , i had one for 'commodore 64'

diana_coman: asciilifeform: they recycled the name as usual; ie still "expert system" though not exactly as promised but "who is there to notice"

shrysr: asciilifeform: i just.. kindda 'did stuff', and tried to learn backwards. idk if you'd call what i know programming. I did not 'do a book' w.r.t programming... at the most in snatches only. coding was more of a hobby and way to reduce manual work.. and have these 'automated systems' doing my work while i did ... whatever else lol. fwiw untill some years ago - i discarded the 'historical' bits i.e took in

shrysr: only upto a point i got an idea of 'then and why now'..

shrysr: asciilifeform: and the eliza story is actually quite instructive i think.

shrysr: I guess i always knew i dont have a structured knowledge base w.r.t programming/comp sci. I think today ...i am often infuriated that i dont have that, particularly when i started moving more into code and data-hag. The only step i took to rectify was starting the SICP course last year, which i abandoned after some lectures thinking I need to

shrysr: learn mainstream shit first to get a job. if diana_coman reccs - i will re-pursue.

asciilifeform: shrysr: some of the most effective folx i've worked with, did not have any traditional schooling; this aint fatal

shrysr: http://logs.ossasepia.com/log/ossasepia/2019-09-20#1003183 << did the 'cure' happen during your masters thesis or did you find out later, i.e did this shape your PhD, and what was that on ?

ossabot: Logged on 2019-09-20 13:39:21 diana_coman: it was a good cure indeed but a cure for my naivity regarding the domain, not of that much use to those affected by breast cancer, what can I say.

shrysr: diana_coman: iirc i read on one of your blog posts that you 'recently returned to programming..' ...trying to find link >> was curious what were you doing before that? and what was it like after getting PhD, as in what did you wanna do? If I may also ask - what/where was that meaning you found?

ossabot: Logged on 2019-09-15 05:47:51 diana_coman: to summarise though: I kept moving (and mentally even more than the physical would directly tell); looking back, at some point I realised that I pretty much traced where the meaning *used to be* until I finally accepted that it's not anymore in any of those places.

shrysr: http://logs.ossasepia.com/log/ossasepia/2019-09-20#1003205 << i guess i am glad to know it isnt. ty.

ossabot: Logged on 2019-09-20 16:26:30 asciilifeform: shrysr: some of the most effective folx i've worked with, did not have any traditional schooling; this aint fatal

diana_coman: http://logs.ossasepia.com/log/ossasepia/2019-09-20#1003206 - during the master's itself really; as a result, I tried the PhD in pretty much ~anything else (the Swiss episode was one attempt) and ended up with an even more obvious (in retrospect) scam: software engineering (that's what my PhD is in); that pretty much made me give up on the whole of it entirely

ossabot: Logged on 2019-09-20 17:49:57 shrysr: http://logs.ossasepia.com/log/ossasepia/2019-09-20#1003183 << did the 'cure' happen during your masters thesis or did you find out later, i.e did this shape your PhD, and what was that on ?

ossabot: Logged on 2019-09-11 08:54:22 diana_coman: and now I remember how 25yo me shocked some Swiss interviewers: "why Switzerland? *shrug* For me the world has only 2 countries: Romania aka home and...the rest of the world"

diana_coman: http://logs.ossasepia.com/log/ossasepia/2019-09-20#1003208 - it was possibly the intro to the Reference Code Shelf mentioning the return to programming; the reason was simply that I wanted Eulora to happen.

ossabot: Logged on 2019-09-20 18:01:16 shrysr: diana_coman: iirc i read on one of your blog posts that you 'recently returned to programming..' ...trying to find link >> was curious what were you doing before that? and what was it like after getting PhD, as in what did you wanna do? If I may also ask - what/where was that meaning you found?

diana_coman: after the PhD I stayed on for a year as post-doc and then stuffed them all and did whatever I wanted to try - just making sure I always found a way to get paid for it, too.

diana_coman: http://logs.ossasepia.com/log/ossasepia/2019-09-20#1003204 - neah, I don't think it's worth it.

ossabot: Logged on 2019-09-20 16:20:21 shrysr: learn mainstream shit first to get a job. if diana_coman reccs - i will re-pursue.

diana_coman: http://logs.ossasepia.com/log/ossasepia/2019-09-20#1003203 - fwiw I went to Uni for 5 years (engineering degree) supposedly to get exactly a structured knowledge base wrt programming/comp sci and I still don't think I actually got that at all, mainly because they hadn't a clear idea as to wtf that was exactly; I made my own structuring later on (and also learnt much more later on); so no, I wouldn't worry at all about that.

ossabot: Logged on 2019-09-20 16:20:19 shrysr: I guess i always knew i dont have a structured knowledge base w.r.t programming/comp sci. I think today ...i am often infuriated that i dont have that, particularly when i started moving more into code and data-hag. The only step i took to rectify was starting the SICP course last year, which i abandoned after some lectures thinking I need to

diana_coman: re meaning found - note that when I accepted no meaning was anymore in any of those places, I hadn't as such found it otherwise; it was missing and therefore entirely up to me to make/find my own, that's all.

diana_coman: in fairness though, out of the "software engineering" thing, at least I got to see all sorts of interesting Italian stuff, Ferrari engineering included; and just to avoid misunderstandings - I don't actually regret any of it, at all.

shrysr: diana_coman: ty. ha! Ferrari.. so cool. I

shrysr: i would think if there was any meaning in the automotive world... it would be @ Ferrari :) .. or perhaps some kindda F1 team.

shrysr: yes it was your reference code shelf, ty. this one i read recently. was thinking it is also mentioned on another post.. but never mind.

asciilifeform: diana_coman: laff if you like, but i always imagined 'ferrari' as a pen&paper shop

jfw: I discovered I had some sleep and work review/planning debts to pay; nearing the end of that and will get back to log catchup shortly.

← 2019-09-19 | 2019-09-21 → ↑