Fork me on GitHub
#clojure-uk
<
2018-09-28
>
thomas07:09:39

morning lovely people!

agile_geek07:09:17

Bore da pawb welsh_flag

agile_geek07:09:49

@otfrom saw your tweet on magit todo's - might give that a look when I have a moment

otfrom08:09:15

the code he is commenting on is handy too as you can create a http://todo.org in a repo and have the todos show up in the magit status. Having both would allow you to do both.

👍 4
3Jane08:09:04

Good moaning, it’s Friday and there’s not enough coffee

thomas08:09:15

@lady3janepl skip the coffee... go directly for something stronger 😉

3Jane08:09:02

What, dropping the production database?

16
3Jane08:09:49

It’s infra team’s turn this week though

alexlynham08:09:43

at my friend's work somebody got a production key and bricked 400 servers earlier this week

3Jane08:09:02

It’s all fun and games until someone loses revenue

parrot 4
thomas08:09:55

oops indeed!

3Jane08:09:31

Guys, if you were to run an experiment with a web crawler and sentiment analysis of the dl’d articles, where would you run it?

thomas08:09:05

or worse... someone dies as a result of something like that ( not impossible if it would be at a hospital)

3Jane08:09:30

There’s so many overhead today, I can’t decide ...

Conor08:09:01

Mechanical Turk, you won't have to write any code 😛

3Jane08:09:51

Model training takes a while though ;)

3Jane08:09:28

I could poke one of those guys that write comparisons of language popularity on GitHub I think

3Jane08:09:55

I’d like to see if it’s possible to compare languages and such by how popular/positively treated they seem to be when people talk about them, rather than how much in use they are.

dominicm11:09:18

Google BigData has GitHub in I think

3Jane08:09:38

(Cue “everything hovers around 0” due to symmetry of likes/dislikes in internet wars I guess XD )

alexlynham09:09:54

@lady3janepl serverless for the crawl, something like parquet/S3 for the storage, some kind of cheap dedicated instance for the rest

alexlynham10:09:17

Just a wild guess at what might be a cheap/fun way to do it :)

Rachel Westmacott10:09:02

I feel like I saw something recently saying it might be cheaper to just buy the hardware yourself than rent GPUs in the cloud - but I didn’t read the article and can’t find it now

jasonbell10:09:39

@lady3janepl https://clojurians.slack.com/archives/C064BA6G2/p1538124031000100 I have a spare machine hanging around for this kind of thing.

3Jane10:09:22

I remember the article - I think the cost was amortised over the time

Rachel Westmacott10:09:58

yes - obviously if you’re just dabbling, it will be cheaper to rent in the cloud for a few hours than buying a new machine up front

3Jane10:09:26

I would, if I did this sort of thing regularly, but I was wondering if there was an equivalent of “cheapo for open source / amateurs”

3Jane10:09:11

Morning! 😄

Rachel Westmacott10:09:38

I suspect it depends hugely on what problem you’re tackling - if you’re using a pre-trained network to analyse new data then it will be a very different scenario to if you’re training a model yourself

3Jane10:09:22

I was going to grab a pretrained model from somewhere, I don’t need huge accuracy

Rachel Westmacott10:09:45

you get some usage on the free tier too with AWS (https://aws.amazon.com/comprehend/pricing/)

3Jane10:09:34

(if you were following the conversation in #jobs-discuss we talked about anonymising CVs among other things; first I wanted to write that, then I wondered whether I could colour-code anonymised technologies depending on their relative popularity)

3Jane10:09:07

(but how to determine whether a technology is “popular”? I could hardcode it, or I could actually check it, thus wondering about what the options are)

3Jane10:09:32

biased sample

3Jane10:09:50

(being “people I have access to” / “groups I hang out with”)

3Jane10:09:41

but yeah, obvs faster 😄

3Jane10:09:38

I wonder how researchers check things like preconceptions (positive or negative) in a reliable way (given that people reporting they like or dislike something do it partly because they are expected to express likes/dislikes, so it’s not wholly accurate.)

alexlynham11:09:14

@lady3janepl re: > I would, if I did this sort of thing regularly, but I was wondering if there was an equivalent of “cheapo for open source / amateurs” Databricks (i.e. hosted Apache Spark) has a community edition on AWS

👍 4
alexlynham11:09:22

you get 4 nodes for free IIRC

alexlynham11:09:31

I have an account & use it to test scala notebooks

3Jane11:09:26

👍 👍 exactly what I was looking for

3Jane11:09:06

btw, anyone going to the spark summit?

alexlynham12:09:25

one of my bosses from my work might be talking so hoping to swing a freebie

3Jane14:09:04

let me know if you do, will come over to say hi - I just found out that I’m going along with one more guy from work

alexlynham14:09:58

nice! I will do

alexlynham14:09:27

oh it's next week?! okay, maybe not then lol

alexlynham14:09:40

but yeah it looks like the head of data eng from my place is speaking

3Jane14:09:31

which talk is that?

danm13:09:09

I should not have gone to the Jura stand of Mancs Food Festival after Whisky Club last night...

Conor13:09:58

Do they still have that stall with the peanut butter icecream @carr0t

danm13:09:23

Er, maybe? It was quite late and I'd already eaten. We were mostly after free whisky

😍 8
Conor13:09:20

Eyes on the prize

danm13:09:01

I'm not sure the whisky was supposed to be free. Other people seemed to be handing over some sort of token in exchange for a small taster dram

danm13:09:21

We turned up a bit en masse, asked what they had got and if any of it was any good, and they just started pouring

😄 8
🚰 4
jasonbell13:09:23

Did someone say free whiskey?

danm14:09:06

Possibly 😉

3Jane15:09:34

unrelated to engineering, but you might find it useful if you’ve not a regular Londoner; it’s certainly given me a lot of ideas of what to do on weekends 🙂

🙌 4
3Jane15:09:05

(a giant map of “if you liked this popular thing, here are other things of the same category which are less well known”)

alexlynham15:09:36

so as I was sitting here reviewing notes on an interview candidate I had an 'aha' moment where the lambda calculus finally clicked

rickmoynihan21:09:01

(fn [f]
  ((fn [x]
     (x x))
   (fn [x]
     (f (fn [y]
          ((x x) y))))))

😁 4
mexisme18:09:47

And people complain there are too many parentheses in Lisp... Pchaw!

practicalli-johnny12:10:40

In lisp code that evaluates, there is always the right number of parentheses 🙂

mexisme21:10:05

haha! lovely!

alexlynham15:09:43

funny how these things happen

alexlynham15:09:33

my subconscious has probably been chewing that one ever since I read Tom Stuart's Understanding Computation back in 2015