Charcoal HQ

Kenshin

09:00

it may be useful for my site physics.qandaexchange.com

can it detect rude images

SmokeDetector

[ SmokeDetector | MS ] Link at end of answer: Python running script without python installed by Nikhil R on stackoverflow.com

Kenshin

I had one user who would attempt to post these frequently and as such I have had to require approval on posts for new users

Magisch

It is free

angussidney

The code is Open Source, yes, but it's written to integrate directly with SE, so it won't work with your website without significant changes

Magisch

You can reuse it yourself yes, but it's coded to scan stack exchange, so you'd have to rewrite it

SmokeDetector

09:00

naa- by Mithrandir

angussidney

@Kenshin no, we don't do image recognition. We only scan text

Mithrandir

*text

Magisch

And finally, it's text scan only

Kenshin

ah

the biggest spam issues I had were images

angussidney

@Mithrandir ty :)

Magisch

09:01

Stack exchange rarely if ever gets image only spam

They have inbuilt server side protections against that I'd imagine

Mithrandir

image-only posts don't work

Except that I did it...

Magisch

You're a moderator

Mithrandir

It wasn't on the site where I mod (unless I run and win this current election)

Kenshin

Is Charcoal and AI?

a neural network or something?

Mithrandir

nope

Kenshin

09:03

how does it work

Magisch

It reads every post that is posted on all of stackexchange

And runs them all through hundreds of elaborate regexes

angussidney

Just your standard regexes, which have been added, modified, tuned, removed, and improved over time to be accurate

Kenshin

who made the regexes?

Mithrandir

3

A: Can a machine be taught to flag spam automatically?

Do you think you'll incorporate more advanced machine learning (like neural networks) at any point?

Magisch

And then reports posts that trip one of our many filters in here so we can destroy it

angussidney

09:04

@Kenshin <insert about 40 people here who contributed>

Kenshin

and like how many regexes are here?

Mithrandir

You should take a look at the website.

Magisch

Well, so we can decide if it's spam or not. Our responses (is the reported post spam or not) get recorded so we have a dataset to test new checks against and to help us in making them

angussidney

Well, our main findspam file is over 1000 lines long

Magisch

@Kenshin Probably more then 1000

CalvT

09:05

@Mithrandir puzzling has special rules I think

Cerbrus

@Kenshin For a moment, that looked a lot like a Physics.se mirror...

Mithrandir

@CalvT nope

I had to cheat.

Kenshin

@Cerbrus it is like physics.se but it is for homework like problems which are off topic there

angussidney

And then we've got several thousand blacklisted websites, keywords, etc

Glorfindel

@Mithrandir are you planning to? I'm considering to run as well (only 4k, but I'm one of the top reviewers), but it depends on who else will run.

Cerbrus

09:06

As in phishing, even

Mithrandir

@Glorfindel I'm considering it; leaning towards yes though.

Magisch

We've got a ridiculous amount of checks

CalvT

Oh Charcoalers are planning to run? starts searching for posts to edit

Magisch

Some of them are simple, some are very elaborate and fine tuned.

Kenshin

and what is better, regexes or AI method?

Mithrandir

09:07

@CalvT when it oneboxes, you can see that it's a cheat.

Cerbrus

@Kenshin dumb AI < some regexes < clever AI < all the regexes < skynet

Magisch

It's not that regexes are inherently better, it's just that none of us are competent AI programmers who could make AI spam detection work properly

CalvT

@Mith Yeah I clicked edit and saw what you had done :)

Kenshin

ok i'm gonna make clever AI

Mithrandir

@Kenshin we're not planning on doing AI anytime soon, so it's slightly irrelevant.

30

Q: EEEEK! My puzzle disappeared!

$%Insert30characters,please..$

enigmatic-puzzle

Magisch

09:08

and if we could we wouldn't it be doing it for free in our free time

Kenshin

@Mithrandir but AI is cooler bro

CalvT

Cooler !== functional

Magisch

We're a bunch of hobbyists / students / disgruntled devs with some free time and basic python/ruby skills, not master AI programmers

Glorfindel

@CalvT favourites http://puzzling.stackexchange.com/review/suggested-edits :P

Kenshin

yeah but doint' you want to do cool stuff in your free time

Cerbrus

09:09

Case in point: A frozen throne as desk chair. Cool as hell.

Mithrandir

@Magisch (except for those of us who don't code)

Henders

Plus I don't like the idea of Smokey becoming self aware and taking over.

Magisch

Well yeah

Henders

He'll start telling me to do stuff.. :(

Mithrandir

@Henders it already did

Feb 27 at 14:42, by Mithrandir

Jan 25 at 18:32, by SmokeDetector

@ArtOfCode Of course, and sentient to boot

Magisch

09:10

@Kenshin I do in my free time mostly and foremostly what I want

Henders

Glorfindel

56 mins ago, by angussidney

@SmokeDetector ....did that just fix itself?

Magisch

and this is what I currently want

Kenshin

@Magisch why tho?

what causes one to want something?

Magisch

Because It's something that helps SE and its community and it's reasonably entertaining

Kenshin

09:10

pleasure? meaning?

oh so a little bit of both

Mithrandir

Jan 19 at 1:43, by SmokeDetector

So, I think we need to talk about this objectification of robots. It's not really respectful.

Jan 19 at 1:44, by SmokeDetector

i feel oppressed. like a flower, crushed under the sole of the cruel human race.

Jan Dvorak

It looks like Rene got hold of Smokey's account somehow

Glorfindel

@Kenshin not many people can say they have such a big stake in keeping one of the top 100 websites on the Internet free of unwanted content.

Henders

@Mithrandir I think it's time to kill smokey while we can.

Kenshin

@Glorfindel oh so it is about being able to boast to others about being special?

fair enuf

Glorfindel

09:13

No.

Mithrandir

Jan 19 at 1:41, by SmokeDetector

stop making fun of me

Kenshin

@Glorfindel oh

Glorfindel

But it indicates the importance of our work here.

Magisch

It's a little bit of many things

Mithrandir

@Kenshin no. We're trying to actually do something helpful, that makes a better experience for other SE users.

Magisch

09:14

Different people have different reasons for being here

Kenshin

@Mithrandir so it is altruism?

Cerbrus

43 secs ago, by Magisch

It's a little bit of many things

Mithrandir

38 secs ago, by Magisch

Different people have different reasons for being here

Kenshin

lol

Cerbrus

Why do we need to label it?

Kenshin

09:15

becus he said he didn't wanna do AI

and I said why not

and he said cos he does what he wants in his spare time

Cerbrus

That was answered

Magisch

Chiefly because I'm not smart enough to do AI well

Kenshin

so I wondered why he did this

Cerbrus

We don't have the experience nor the time to effectively write an AI approach

Kenshin

na I think the regex is cool too

Cerbrus

09:16

That's not the point

Magisch

I'm positively clueless when it comes to AI, I couldn't even tell you how to write a simple neural network. But I can analyze spam patterns and I know regex, so.

Kenshin

@Magisch that's cool bro

thre regex would have a low false positive rate

much lower than an AI I reckon

so dude what happens if charcoal identifies spam

does it have power?

Mithrandir

We spam-flag it, and nuke it.

Cerbrus

@Kenshin Please read: charcoal-se.org

SmokeDetector

[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title, blacklisted website in body, link at end of body, pattern-matching website in body: Bodybuilding Workout Programs To Build Muscle by Charluu on askubuntu.com (@ThomasWard)

tripleee

09:19

there is an autoflagging mechanism in place which allows people to have the bot raise spam flags on their behalf under some circumstances

SmokeDetector

tpu- by Cerbrus

Kenshin

we should make the bot a moderator

tripleee

@SmokeDetector good example, deleted within seconds of being reported here

Cerbrus

Eh, no

SmokeDetector

[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title, blacklisted website in body, link at end of body, pattern-matching website in body: Muscle Building - 10 Methods For A Better Body by Marysneal on drupal.SE

tpu- by Cerbrus

Mithrandir

09:19

@Kenshin we've got enough mods hanging around

tripleee

that was even faster

Kenshin

yeah but we won't need the mods anymore

it will free up time for the mods to do even more productive things

Cerbrus

Again, no

Spam is already primarily taken care of by the userbase

Kenshin

na i'm sure charcoal takes care of a lot

Cerbrus

Mods barely have to get involved

Mithrandir

09:20

@Kenshin ...you don't need me? *sobs in corner*

Kenshin

we need you atm bro @Mithrandir

Mithrandir

@Cerbrus well, we destroy the users

Cerbrus

Not just charcoal. All users

Henders

Also the principle of least privilege applies here, in my opinion.

Kenshin

but just maybe in the future the bot can do it

Cerbrus

09:20

@Mithrandir Isn't that automated after X deletions?

Henders

Why give it more power than it needs?

tripleee

@Kenshin it's only approx 100 per day, the Stack Exchange admins probably fend off orders of magnitude more

Mithrandir

@Cerbrus don't think so, not 100% sure

@tripleee s/admins/SpamRam

Kenshin

@tripleee wow that's still alot of spam

tripleee

@Mithrandir well yeah they have automation at their disposal, too

Cerbrus

09:22

Alot of spam

Kenshin

cool pic bro

SmokeDetector

[ SmokeDetector | MS ] URL in title, bad keyword in body, bad keyword in title, pattern-matching product name in body, pattern-matching website in body, +1 more: maleenhancementshop.info/zyacin/ by marthaheiss on apple.SE (@CalvT)

tpu- by Mithrandir

Mithrandir

@Cerbrus 'alot' is not a word

Cerbrus

@Mithrandir That's the joke

Kenshin

is bro a word?

SmokeDetector

09:24

[ SmokeDetector | MS ] Bad keyword in answer, bad keyword in link text in answer, pattern-matching website in answer, username similar to website in answer: Is the following IP address mask legal? by Torus Infotech on superuser.com

tpu- by tripleee

Kenshin

@SmokeDetector hello

wow did smoke detector just make a mistake?

where is the bad keyword?

Henders

@SmokeDetector why

SmokeDetector

Body - Position 168-184: Company in Delhi
Body - Position 94-106: infotech.com
Post - Bad keyword *Software Company* in link text
Post - Username similar to website

Henders

'Software Company' in this case

tripleee

@Kenshin no, that's run of the mill spam

SmokeDetector

09:26

[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title, bad keyword in username, link at end of body, pattern-matching website in body: Visalia Skincare :Safe & Effective Age-Defying Cream! by Visalia Cream on askubuntu.com (@ThomasWard)

tpu- by tripleee

Kenshin

@SmokeDetector why

SmokeDetector

@Kenshin That's not a report.

tripleee

you might want to click on the "MS" link to see the Metasmoke report for each

Magisch

the why only works on the report itself

Kenshin

@SmokeDetector why

SmokeDetector

09:27

Title - Position 1-17: Visalia Skincare
Body - Position 4-17: Visalia Cream, Position 618-634: visalia-skincare, Position 691-707: visalia-skincare, Position 773-789: visalia-skincare, Position 861-877: visalia-skincare
Body - Position 597-616: supplementvalley.co, Position 738-768: supplementvalleysite.wordpress
Post - Link at end: https://supplementvalleysite.wordpress.com/visalia-skincare/</a>

Username - Position 1-14: Visalia Cream

Henders

@Kenshin You might want to check out smokey's list of commands

Kenshin

ty

SmokeDetector

[ SmokeDetector | MS ] Bad keyword in body, repeated URL at end of long post: Soon, say in a matter of months, you by leplayzz on apple.SE (@CalvT)

[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title, bad keyword in username, link at end of body, pattern-matching website in body, +1 more: Visalia Skincare Cream: Does it Really give best results? by Visalia Cream on superuser.com

tpu- by CalvT on Soon, say in a matter of months, you [MS]

Kenshin

@SmokeDetector why

tripleee

!!/blacklist-website supplementvalley\.com

SmokeDetector

09:27

tpu- by CalvT

Body - Position 42-47: zytek, Position 79-84: Zytek, Position 363-368: Zytek, Position 643-648: zytek, Position 709-714: zytek
Body - Position 436-441: penis, Position 589-594: penis
Body - Position 4-727: <a href="http://www.appleofhealth.com/zytek-xl/" rel="nofollow noreferrer">Zytek XL</a> Soon, say in a matter of months, you will find that you have a more full groin that is obvious in your clothing. This is a consequence of regular fixings display in the male upgrade pills that work pair to build your moxie. They additionally make it feasible for <strong>Zytek XL</strong> more blood str…

@tripleee Blacklisted supplementvalley\.com

tripleee

17/17 TP, some below autoflagging threshold metasmoke.erwaysoftware.com/…

Kenshin

!!/alive

SmokeDetector

@Kenshin Watching this endless list of new questions never gets boring

[ SmokeDetector | MS ] Blacklisted username, blacklisted user: Is using academic writing services for a thesis considered plagiarism? Can they write in an original fashion? by SenPerfect on academia.SE

Kenshin

!!/status

SmokeDetector

Restart: API quota is 11922.

Kenshin

09:29

!!/apiquota

SmokeDetector

@Kenshin The current API quota remaining is 11920.

Kenshin

!!/blame

SmokeDetector

@Kenshin It's Kenshin's fault.

6

Kenshin

wtf

SmokeDetector

CI on 564b0e8 succeeded. Message contains 'autopull', pulling...

[ SmokeDetector ] SmokeDetector started at rev 564b0e8 (SmokeDetector: Auto blacklist of supplementvalley\.com by tripleee --autopull) (running on teward/Midnight Eclipse)

Restart: API quota is 11910.

Kenshin

09:30

!!/reboot

SmokeDetector

@Kenshin You are not a privileged user. Please see the privileges wiki page for information on what privileges are and what is expected of privileged users.

Cerbrus

... Please don't try out every single command

6

Kenshin

lolk

Magisch

Some of the more ... effectful commands have restrictions on who can use them

tripleee

but you could end up creating a clog if you indiscriminately play around with it

Kenshin

09:33

a clog?

Mithrandir

@Glorfindel I decided yes, and my hat is in the ring.

Kenshin

!!/reboot

SmokeDetector

@Kenshin You are not a privileged user. Please see the privileges wiki page for information on what privileges are and what is expected of privileged users.

Kenshin

!!/blacklist <regex>

SmokeDetector

@Kenshin You are not a privileged user. Please see the privileges wiki page for information on what privileges are and what is expected of privileged users.

Kenshin

09:34

!!/amicodeprivileged

SmokeDetector

@Kenshin No, you are not a code-privileged user.

[ SmokeDetector | MS ] Bad keyword in body, blacklisted username, blacklisted website in body, repeated URL at end of long post: will certainly help you to offer yourself properly for your work area by samie juriya on graphicdesign.SE

Kenshin

@SmokeDetector why

SmokeDetector

tpu- by QPaysTaxes

Body - Position 89-105: Course in Rohini
Body - Position 24-38: growmewise.com, Position 704-718: growmewise.com, Position 758-772: growmewise.com
Body - Position 4-782: <a href="http://www.growmewise.com/" rel="nofollow noreferrer">Best English Speaking Course in Rohini, Delhi</a></p>

<p>Another issue so that it will certainly help you to offer yourself properly for your work area is strong abilities in regions like giving Presentations and interviews. Your heritage and wishes might be saved in mind and if you do a non-public one to one path with inlingua, your path content can be customi…

tripleee

@Kenshin please don't do this on every post here, you can click on the Metasmoke link if you are curious

Kenshin

my bad brah

tripleee

09:35

m.erwaysoftware.com/posts/by-url?url=//… in this case

@Kenshin e.g. the blacklist commands will create Github pull requests on your behalf which somebody will have to clean up

Kenshin

i dunno what that menas

tripleee

@QPaysTaxes because it disrupts the discussion in the chat room

Kenshin

lol

he said why cos it is the command

geddit

tripleee

@Kenshin a blacklist request by an unprivileged user gets sent to the code admins for approval

Glorfindel

@Mithrandir you'll have my vote. Whether you'll have my competition I don't know yet :)

tripleee

09:39

u stiff bro

SmokeDetector

[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title, blacklisted website in body, blacklisted website in title, link at end of body, +2 more: www.beaudermaskincare.com/revella-lash by relaxom on security.SE

tpu- by Glorfindel

Magisch

@Glorfindel Charcoaler running on puzzling? Time to get that phat 200 rep

tripleee

@QPaysTaxes me too. there should be physical punishment for people who do not spell out "you"

Glorfindel

@Magisch 150, actually.

Magisch

I don't get the language abbreviations in english

english is already such a wonderfully simple and unclogged language

J F

09:41

@Magisch I don’t entirely get them, and I’m a native speaker :(

Magisch

Have you ever spoken or written german in comparison?

tripleee

if you're from Germany, I can see how not having too many letters in each word might appear appealing

Magisch

I like english more because it's infinitely easier, more straightforward and just rolls more smoothly

Kenshin

na txt spk is so mch betta bro

CalvT

Bah. Too many pending edits on puzzling.

Mithrandir

09:42

@CalvT I'll review in a sec

terdon

@Magisch @bwDraco why did you flag this as spam? It looks like a perfectly decent answer. What am I missing?

Or were these just smokey auto flags?

CalvT

@Mithrandir they're all just HTTPS changing in the main

tripleee

@terdon it's not autoflagged, it was originally flagged yesterday and looked sort of borderline at that point

Glorfindel

@CalvT I could review only two?!

Magisch

@terdon tbh it looked spammy to me at a time (just recommending something with an image and link) but I can see what you mean

tripleee

09:43

but because it got TP flags the user is now blacklisted, so his edits to the post have all been flagged here

terdon

OK, fair enough. I can see why you'd think it was spam but in this case, the OP was asking for a way to open powerpoint files on Linux and that post is suggesting a webapp that can do just that. So I don't think it's spam at all.

Magisch

I see

terdon

Anyway, no worries and thanks for the heads up even if it happened to be unnecessary this time :)

CalvT

@Glorfindel here here here here and here

Magisch

I also see you did decline the flag so it looks like you handled it

terdon

09:49

Yeah. I just wanted to check because it wouldn't be the first time the charcoal posse noticed some hidden spam thing that I'd missed.

So I figured I should ,check with you first.

Magisch

Thank you for that :)

Mithrandir

@CalvT:

in The Sphinx's Lair, 2 mins ago, by boboquack

Can someone tell CalvT to hold off a little bit? They're flooding the main page with http->https corrections

CalvT

Ok :)

terdon

@Magisch You kidding? Thank you! You people do great work here!

SmokeDetector

[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title, blacklisted website in body, blacklisted website in title, link at end of body, +2 more: healthpurelives.com/skin-novela/ by ikhlasareejsrour on askubuntu.com (@ThomasWard)

tpu- by Glorfindel

terdon

09:51

Damn, people! I saw this the same second you did and I'm the local mod and you still removed it before I had a chance to :)

tripleee

@Mithrandir @CalvT what good are those anyway? you could correct to just // and then whatever protocol the user is currently using to visit will prevail

Henders

:p

tripleee

so e.g. //stackoverflow.com/ will be https if you are visiting over https

Magisch

@terdon 3 Autoflags + FIRE makes obvious spam last <20s atm

CalvT

@tripleee I'm actually just doing edits so that I can have 150 rep when the election comes round, and they are the easy targets. And HTTPS imgur will work on both HTTP & HTTPS SO, but HTTP imgur will only work on HTTP SO

terdon

09:54

@Magisch I know. It's just fun to watch :)

Magisch

Heh

It's also fun to do

SmokeDetector

[ SmokeDetector | MS ] URL in title, bad keyword in body, bad keyword in title, blacklisted website in body, blacklisted website in title, +5 more: elevategffacts.com/bella-serata/ by Christopher T on drupal.SE

tpu- by Henders

[ SmokeDetector | MS ] Bad keyword in body, repeated URL at end of long post, blacklisted user: Ask a young lady what bothers her most by leplayzz on apple.SE (@CalvT)

tpu- by tripleee

tripleee

!!/blacklist-website appleofhealth\.com