Register a SA Forums Account here!
JOINING THE SA FORUMS WILL REMOVE THIS BIG AD, THE ANNOYING UNDERLINED ADS, AND STUPID INTERSTITIAL ADS!!!

You can: log in, read the tech support FAQ, or request your lost password. This dumb message (and those ads) will appear on every screen until you register! Get rid of this crap by registering your own SA Forums Account and joining roughly 150,000 Goons, for the one-time price of $9.95! We charge money because it costs us money per month for bills, and since we don't believe in showing ads to our users, we try to make the money back through forum registrations.
 
  • Post
  • Reply
blatman
May 10, 2009

14 inc dont mez


VanSandman posted:

Legalism is a scum philosophy for bootlickers.

Adbot
ADBOT LOVES YOU

Shame Boy
Mar 2, 2010

Judge Dredd Scott posted:

trying to moderate all of discord is like trying to moderate all of irc. it makes absolutely zero sense

also i don't know anybody who would willingly use an instant messaging service that was actively monitored by an algorithm or *anything* for that matter lol. nobody's asking AT&T to listen to all phone calls is an appropriate analogy, i think

i dont know any actual solution to this but snooping directly isn't it thats for sure

All of discord is controlled by one entity while all of IRC isn't. Plus I don't think discord monitors (or should monitor) private/direct messages, which would be the phone call analogy, just the channels and servers. Discord admins / staff already regularly randomly drop in to servers to see what they're about all the time, and people keep using it regardless, so I don't think it matters nearly as much as you think.

Fame Douglas
Nov 20, 2013

by Fluffdaddy

Shame Boy posted:

All of discord is controlled by one entity while all of IRC isn't. Plus I don't think discord monitors (or should monitor) private/direct messages, which would be the phone call analogy, just the channels and servers. Discord admins / staff already regularly randomly drop in to servers to see what they're about all the time, and people keep using it regardless, so I don't think it matters nearly as much as you think.

Would you really expect Quakenet to police all their IRC channels? There's no fundamental difference. It's simply impossible.

dex_sda
Oct 11, 2012


police need a wiretap to listen to a phone call, but they regularly randomly drop into places to see what they're about all the time, and yet precincts are burning in america. people don't actually want their poo poo snooped on.

Subjunctive
Sep 12, 2006

✨sparkle and shine✨

bike tory posted:

I'm not a computer toucher by trade or anything but I'm pretty sure it's a shitload easier to train algorithms on text than it is on images or video

it’s not an issue of format. training on video versus text is mostly just a matter of more/bigger computers, which isn’t a material barrier to even a company of Discord’s modest size. it’s an issue of “closely matches the form of this exact thing that is long, even when excerpted” versus “matches the meaning of these sentiments, expressed in disjoint short statements, via language that evolves rapidly and can easily be obfuscated”. I’ve worked with very-well-resourced researchers who are making meaningful progress on this exact issue of hate detection in short-form text, and they will all tell you that image and waveform detection is a piece of cake in comparison.

effective ML to detect hate speech would be an operational and financial boon to a number of very large companies, and I have some hope that we’ll see at least pre-AlexNet-image-stuff performance on the task in the next few years, but it’s definitely not a problem with a clear solution. multiple companies have teams which are, I suspect, larger than the whole of Discord Inc working on this space now.

blunt
Jul 7, 2005

Fame Douglas posted:

Would you really expect Quakenet to police all their IRC channels? There's no fundamental difference. It's simply impossible.

Not with 1997 processing power, yes with 2020 processing power.

Shame Boy
Mar 2, 2010

Fame Douglas posted:

Would you really expect Quakenet to police all their IRC channels? There's no fundamental difference. It's simply impossible.

I'd expect them to randomly drop in to make sure they're not full of nazi child porn rings or w/e from time to time, yes. Speaking as someone who modded a fairly big channel on a pretty big network for like 10 years, admins did in fact do that occasionally :shrug:

Centrist Committee
Aug 6, 2019
building robots to police speech rather than subjugating the wealthy smh

Main Paineframe
Oct 27, 2010

Subjunctive posted:

the copyright problem is a fair bit easier than shutting down the Proud Boys discord without also nuking Emily’s or whoever’s anti-racist groups, and still has a terrible rate of false positives _and_ negatives

if they can stay on top of responding to reports effectively, which would be a first for social media and I’ll definitely be cheering for them, then the environment will become inhospitable pretty well anyway

no, it won't, because then things will be reliant on unpaid volunteers hunting through the internet to find publicly posted links to Nazi discords, which is not a particularly thorough method. in order to find offensive content, social media relies heavily on reports from people who don't like that offensive content, which is fundamentally difficult in a system specifically designed to silo content so that people can easily surround themselves with only things they like to see

social media companies have deliberately removed as much human review as possible because they get more growth and better advertising metrics when they make it as easy and smooth as possible for people to post content and create groups. this also creates the bonus problem that focusing so heavily on increasing the number of users and posts at all costs guarantees that any human review systems will never be able to catch up

although I should note that for all this "but it's hard" stuff, social media companies don't appear to have much trouble removing stuff that might get them in trouble with governments, such as pro-ISIS posts. for instance, Facebook once claimed they could remove 83% of pro-ISIS propaganda within an hour of it being posted, thanks to an AI model they've trained specifically on ISIS propaganda, as well as an aggressive approach that flags all of an account's friends and groups and connections for the highest level of scrutiny as soon as it becomes associated with terrorist content

hell, even when it comes to white supremacists and hate speech, social media companies often exaggerate how much trouble they have dealing with them, because they seem to have a lot less trouble removing them for German users!
https://mobile.twitter.com/Hadas_Gold/status/937757138381955073

Megillah Gorilla
Sep 22, 2003

If only all of life's problems could be solved by smoking a professor of ancient evil texts.



Bread Liar

IAMKOREA posted:

We just use indoor drying racks in Europe in the winter, IKEA sells them for a few dollars. My house has bad humidity problems in winter though and I would kill for a good dryer. You can't buy the air dryers they have in the US here, only condenser models are available and they loving suck, they leave clothes so wrinkled that they are unwearable (even for a guy like me who doesn't give a poo poo).

Do you have room for a drying cabinet? They absolutely loving rock.



Megillah Gorilla has issued a correction as of 15:07 on Jun 21, 2020

the milk machine
Jul 23, 2002

lick my keys

Judge Dredd Scott posted:

also i don't know anybody who would willingly use an instant messaging service that was actively monitored by an algorithm or *anything* for that matter lol. nobody's asking AT&T to listen to all phone calls is an appropriate analogy, i think

you'd better sit down, i have some bad news about what's been going on the past 20 years

Zvahl
Oct 14, 2005

научный кот

the milk machine posted:

you'd better sit down, i have some bad news about what's been going on the past 20 years

are you telling me my icq logs aren't safe

Kitfox88
Aug 21, 2007

Anybody lose their glasses?

Zvahl posted:

are you telling me my icq logs aren't safe

I

SEEK

YOU

:tinfoil:

Menacer
Nov 25, 2000
Failed Sega Accessory Ahoy!
hell yeah this panopticon is totally going to be used to destroy fascists and racists

*panopticon is immediately used to destroy the enemies of capital exclusively*

Zvahl
Oct 14, 2005

научный кот

Kitfox88 posted:

I

SEEK

YOU

:tinfoil:

Only registered members can see post attachments!

duz
Jul 11, 2005

Come on Ilhan, lets go bag us a shitpost


Main Paineframe posted:

no, it won't, because then things will be reliant on unpaid volunteers hunting through the internet to find publicly posted links to Nazi discords, which is not a particularly thorough method. in order to find offensive content, social media relies heavily on reports from people who don't like that offensive content, which is fundamentally difficult in a system specifically designed to silo content so that people can easily surround themselves with only things they like to see

social media companies have deliberately removed as much human review as possible because they get more growth and better advertising metrics when they make it as easy and smooth as possible for people to post content and create groups. this also creates the bonus problem that focusing so heavily on increasing the number of users and posts at all costs guarantees that any human review systems will never be able to catch up

although I should note that for all this "but it's hard" stuff, social media companies don't appear to have much trouble removing stuff that might get them in trouble with governments, such as pro-ISIS posts. for instance, Facebook once claimed they could remove 83% of pro-ISIS propaganda within an hour of it being posted, thanks to an AI model they've trained specifically on ISIS propaganda, as well as an aggressive approach that flags all of an account's friends and groups and connections for the highest level of scrutiny as soon as it becomes associated with terrorist content

hell, even when it comes to white supremacists and hate speech, social media companies often exaggerate how much trouble they have dealing with them, because they seem to have a lot less trouble removing them for German users!
https://mobile.twitter.com/Hadas_Gold/status/937757138381955073

twitter has already said they don't automatically do any sort of filtering of white supremacist content because it flags too many politicians and their goal is to remain unregulated

voiceless anal fricative
May 6, 2007

Subjunctive posted:

it’s not an issue of format. training on video versus text is mostly just a matter of more/bigger computers, which isn’t a material barrier to even a company of Discord’s modest size. it’s an issue of “closely matches the form of this exact thing that is long, even when excerpted” versus “matches the meaning of these sentiments, expressed in disjoint short statements, via language that evolves rapidly and can easily be obfuscated”. I’ve worked with very-well-resourced researchers who are making meaningful progress on this exact issue of hate detection in short-form text, and they will all tell you that image and waveform detection is a piece of cake in comparison.

effective ML to detect hate speech would be an operational and financial boon to a number of very large companies, and I have some hope that we’ll see at least pre-AlexNet-image-stuff performance on the task in the next few years, but it’s definitely not a problem with a clear solution. multiple companies have teams which are, I suspect, larger than the whole of Discord Inc working on this space now.

Literally all you'd need is a bot that flagged channels when the percent of certain words hit a certain threshold, and for a human mod to then drop in and review the channel. I guess they'd need to keep the list of hate words updated but that wouldn't be particularly challenging.

I'm not talking about having an AI that polices everything that's said effectively and automatically bans specific racists in every channel of every server. Just about having something that can detect the servers that white supremacists use to hang out and organise, and banning those servers and the users involved. This is basically what they already do, except they rely on user reports rather than having a bot that trawls through.

It also wouldn't be perfect but the point isn't to eradicate 100% of hate speech, just to make the platform difficult for nazis to use so that they'll move somewhere else for their organising.

Doggles
Apr 22, 2007

https://twitter.com/Newsweek/status/1274129201025757185

Centrist Committee
Aug 6, 2019

bike tory posted:

Literally all you'd need is a bot that flagged channels when the percent of certain words hit a certain threshold, and for a human mod to then drop in and review the channel. I guess they'd need to keep the list of hate words updated but that wouldn't be particularly challenging.

I'm not talking about having an AI that polices everything that's said effectively and automatically bans specific racists in every channel of every server. Just about having something that can detect the servers that white supremacists use to hang out and organise, and banning those servers and the users involved. This is basically what they already do, except they rely on user reports rather than having a bot that trawls through.

It also wouldn't be perfect but the point isn't to eradicate 100% of hate speech, just to make the platform difficult for nazis to use so that they'll move somewhere else for their organising.

racism isn’t a technology problem, just look at how these forums react every time the mods try to do anything, posts like these make me uncomfortable because they seem to imply you can completely ignore politics

Subjunctive
Sep 12, 2006

✨sparkle and shine✨

bike tory posted:

It also wouldn't be perfect but the point isn't to eradicate 100% of hate speech, just to make the platform difficult for nazis to use so that they'll move somewhere else for their organising.

why is it better for them to use a different platform? I mean, I get that it’s less stress and brand risk for Discord as a company, but why is it better for the subjects of hate for their persecutors to be using ICB or ntalk than discord? It can only be a matter of time before they realize that they want E2E more than they want the fancy roles and bits that they get on Discord anyway.

Main Paineframe posted:

although I should note that for all this "but it's hard" stuff, social media companies don't appear to have much trouble removing stuff that might get them in trouble with governments, such as pro-ISIS posts. for instance, Facebook once claimed they could remove 83% of pro-ISIS propaganda within an hour of it being posted, thanks to an AI model they've trained specifically on ISIS propaganda, as well as an aggressive approach that flags all of an account's friends and groups and connections for the highest level of scrutiny as soon as it becomes associated with terrorist content

I’m trying to find the paper about the ISIS detection stuff, but it might not have been published after all. I’ll ask if there’s a pre-print floating around. Propaganda for a single organization is much more like copyright targets in that it’s designed to be published widely and has consistent structure and content, versus low-volume private conversation that supports hate. I’m not sure how having such a system with a 15% failure rate meaningfully reduces their exposure to any legal risk, but I’ll take your word for it; government involvement was never part of any discussions I had around unwanted content except for CEI.

If there were as much and varied ISIS content as there is hate content, the enforcement around ISIS would be much less effective, I think.

SpacePig
Apr 4, 2007

Hold that pose.
I've gotta get something.

One reliable way to create new jobs is to kill people working the jobs.

Subjunctive
Sep 12, 2006

✨sparkle and shine✨

SpacePig posted:

One reliable way to create new jobs is to kill people working the jobs.

every newly-graduating librarian in the last 5 years has considered it, I think

Shame Boy
Mar 2, 2010

Subjunctive posted:

why is it better for them to use a different platform? I mean, I get that it’s less stress and brand risk for Discord as a company, but why is it better for the subjects of hate for their persecutors to be using ICB or ntalk than discord? It can only be a matter of time before they realize that they want E2E more than they want the fancy roles and bits that they get on Discord anyway.

From a practical standpoint it's a hell of a lot easier to get going with Discord. For example literally everyone I know around my age has a Discord account, but only one of my friends uses Signal, so if I were to start a Pepe The Frog's Fun Time Hitler Extravaganza group or w/e I know which service I'd prefer to use for maximum effect. Like you're never going to eliminate the fash ability to communicate but pushing them out of popular easy to use platforms isn't nothing either.

Shame Boy
Mar 2, 2010

Anyway I'm bored and reading a bunch of labor history stuff lately, so let's play Guess That Employer Based On Their Workers' Demands:

- Stop hiring through sub-contractors
- Insurance coverage
- Proper compensation for accidents on the job
- Abolish payment via coupons and company scrip rather than money
- Improvement of on the job health services
- Recognition as employees

Are you thinking Amazon maybe?

Good call! It's the United Fruit Company workers strike in 1928, which took place near the Amazon rainforest!

Shame Boy has issued a correction as of 21:08 on Jun 21, 2020

Arcteryx Anarchist
Sep 15, 2007

Fun Shoe
I was going to guess instacart or some clone

AvesPKS
Sep 26, 2004

I don't dance unless I'm totally wasted.

Subjunctive posted:

every newly-graduating librarian in the last 5 years has considered it, I think

Places still employ librarians?

Inceltown
Aug 6, 2019

Megillah Gorilla posted:

Do you have room for a drying cabinet? They absolutely loving rock.





Mines a bit bigger than that*.



* clothes will smell a bit smokey

Failed Imagineer
Sep 22, 2018

AvesPKS posted:

Places still employ librarians?

Some libraries, even

voiceless anal fricative
May 6, 2007

Subjunctive posted:

why is it better for them to use a different platform? I mean, I get that it’s less stress and brand risk for Discord as a company, but why is it better for the subjects of hate for their persecutors to be using ICB or ntalk than discord? It can only be a matter of time before they realize that they want E2E more than they want the fancy roles and bits that they get on Discord anyway.

It's mostly this:

Shame Boy posted:

From a practical standpoint it's a hell of a lot easier to get going with Discord. For example literally everyone I know around my age has a Discord account, but only one of my friends uses Signal, so if I were to start a Pepe The Frog's Fun Time Hitler Extravaganza group or w/e I know which service I'd prefer to use for maximum effect. Like you're never going to eliminate the fash ability to communicate but pushing them out of popular easy to use platforms isn't nothing either.

Pushing them off discord isn't going to destroy the fash, or stop them organising or anything. But trying to make popular platforms fash free is a good goal. It's a shitload easier for them to recruit and they'll form much larger communities on a service like Discord that already has 300m users compared to a niche E2E encrypted service where someone would have to download an app, make an account and receive an individual invite.

Ruffian Price
Sep 17, 2016

plus using the Nazi communicator is an instant out like being on Gab or Voat

Main Paineframe
Oct 27, 2010

Subjunctive posted:

I’m trying to find the paper about the ISIS detection stuff, but it might not have been published after all. I’ll ask if there’s a pre-print floating around. Propaganda for a single organization is much more like copyright targets in that it’s designed to be published widely and has consistent structure and content, versus low-volume private conversation that supports hate. I’m not sure how having such a system with a 15% failure rate meaningfully reduces their exposure to any legal risk, but I’ll take your word for it; government involvement was never part of any discussions I had around unwanted content except for CEI.

If there were as much and varied ISIS content as there is hate content, the enforcement around ISIS would be much less effective, I think.

as long as the failures come in the form of false positives rather than false negatives, i don't think they care. just look at how over-aggressive sites are about copyright stuff. pleasing governments is heavily prioritized over pleasing users

and there's a shithead of actual white nationalist propaganda out there. what people are really concerned about isn't random nobodies posting random racist crap, it's the white supremacist leaders and organizations growing and expanding their movements on these sites

Tashilicious
Jul 17, 2016

by Jeffrey of YOSPOS

Main Paineframe posted:

as long as the failures come in the form of false positives rather than false negatives, i don't think they care. just look at how over-aggressive sites are about copyright stuff. pleasing governments is heavily prioritized over pleasing users

and there's a shithead of actual white nationalist propaganda out there. what people are really concerned about isn't random nobodies posting random racist crap, it's the white supremacist leaders and organizations growing and expanding their movements on these sites

I love that someone joevatared you.

T-man
Aug 22, 2010


Talk shit, get bzzzt.

i have a radical new idea for how to combat nazis on social media

don't have nazis and stomp that poo poo out IRL, with a boot, in minecraft

Tashilicious
Jul 17, 2016

by Jeffrey of YOSPOS

T-man posted:

i have a radical new idea for how to combat nazis on social media

don't have nazis and stomp that poo poo out IRL, with a boot, in minecraft

woah hey now lets not be hasty the nazis are very fine individuals whos freedom of speech we need to protect after all without the nazis we wouldn't have rockets or the best top cover of any machine gun or snazzy uniforms!

Shame Boy
Mar 2, 2010

Tashilicious posted:

woah hey now lets not be hasty the nazis are very fine individuals whos freedom of speech we need to protect after all without the nazis we wouldn't have rockets or the best top cover of any machine gun or snazzy uniforms!

Goddard could have gotten us to the moon if the government wasn't so intent on ignoring him for most of his career :colbert:

Also if he hadn't died in 1945 but still.

Tashilicious
Jul 17, 2016

by Jeffrey of YOSPOS

Shame Boy posted:

Goddard could have gotten us to the moon if the government wasn't so intent on ignoring him for most of his career :colbert:

Also if he hadn't died in 1945 but still.

A lot of people could have done a lot of things for us. Unfortunately, most of them are ground into blood to fuel the machine.

Hollandia
Jul 27, 2007

rattus rattus


Grimey Drawer
“This machine thrills fascists”

Syncopation
Feb 21, 2020

bike tory posted:

Literally all you'd need is a bot that flagged channels when the percent of certain words hit a certain threshold, and for a human mod to then drop in and review the channel. I guess they'd need to keep the list of hate words updated but that wouldn't be particularly challenging.

The chance for abuse within this theoretical system is ridiculous

doesn't take a lot for a bot designed to flag servers with high volumes of racial slurs/violent threats to flag servers where 'union' and 'strike' are used a lot

really dont understand this urge/desire to build a panopticon for the modern version of AOL Instant Messenger

Finicums Wake
Mar 13, 2017
Probation
Can't post for 8 years!

Judge Dredd Scott posted:

doesn't take a lot for a bot designed to flag servers with high volumes of racial slurs/violent threats to flag servers where 'union' and 'strike' are used a lot
facebook is already building a bot that does exactly this for employers. they even used unionizing as the ready-made example of why employers would want to use the feature.

https://theintercept.com/2020/06/11/facebook-workplace-unionize/

our options right now, for the immediate future at least are:
1) dystopian panopticon that's used primarily against the left
2) dystopian panopticon used against the left and the extreme right

pushing for 2 because it's better than 1 doesn't necessarily mean support for dystopian panopticons, it's just our least worst option until the left can seize the state

Adbot
ADBOT LOVES YOU

Grand Theft Autobot
Feb 28, 2008

I'm something of a fucking idiot myself

nexous posted:

highly compensated employees can only contribute around the same amount as the average employee, which is generally very low.


https://smartasset.com/retirement/401k-highly-compensated-employee

that said he should be maxing out other things like IRAs

Yeah but he's only putting $4,500 a year into it. That's absurdly low for his supposed income.

  • 1
  • 2
  • 3
  • 4
  • 5
  • Post
  • Reply