News for hackers. Stuff that matters.
Re:Purple Llama by Meta AI by comboy on Thursday December 07, 2023 @11:22AM (#38558216)
> we gotta set the house on fire

Context doesn't matter, they can't afford this being on the platform and being interpreted with different context. I think flagging it is understandable given their scale (I still wouldn't use them, but that's a different story).

Re:Purple Llama by Meta AI by pixelbyindex on Thursday December 07, 2023 @11:47AM (#38558592)
> Context doesn't matter, they can't afford this being on the platform and being interpreted with different context

I have to disagree. The idea that allowing human interaction to proceed as it would without policing presents a threat to their business or our culture is not something I have seen strong enough argument for.

Allowing flagging / reporting by the users themselves is a better path to content control.

IMO the more we train ourselves that context doesn't matter, the more we will pretend that human beings are just incapable of humor, everything is offensive, and trying to understand others before judging their words is just impossible, so let the AI handle it.

Re:Purple Llama by Meta AI by comboy on Thursday December 07, 2023 @11:56AM (#38558721)
I wondered about that. Ideally I would allow everything to be said. The most offensive things ever. It's a simple rule and people would get desensitized to written insults. You can't get desensitized to physical violence affecting you.

But then you have problems like doxing. Or even without doxing promoting acts that affect certain groups or certain places. Which certain amount of people will follow, just because of the scale. You can say these people would be responsible, but with scale you can hurt without breaking the law. So where would you draw the line? Would you moderate anything?

Re:Purple Llama by Meta AI by hansvm on Thursday December 07, 2023 @12:46PM (#38559390)
Scale is just additional context. The words by themselves aren't an issue, but the surrounding context makes it worth moderating.
Re:Purple Llama by Meta AI by bentcorner on Thursday December 07, 2023 @12:43PM (#38559358)
Welcome to the Content Moderation Learning Curve: https://www.techdirt.com/2022/11/02/hey-elon-let-me-help-you...

I don't envy anyone who has to figure all this out. IMO free hosting does not scale.

Re:Purple Llama by Meta AI by ethbr1 on Thursday December 07, 2023 @01:22PM (#38559835)
When the 2020 election shenanigans happened, Zuckerberg originally made a pretty stout defense of free speech absolutism.

And then the political firestorm that ensued, from people with the power to regulate Meta, quickly changed his talking points.

Re:Purple Llama by Meta AI by ChadNauseam on Thursday December 07, 2023 @12:07PM (#38558887)
I agree with you, but don't forget that John Oliver got on Last Week Tonight to accuse Facebook's lax moderation of causing a genocide in Myanmar. The US media environment was delusionally anti-facebook so I don't blame them for being overly censorious
Re:Purple Llama by Meta AI by ragequittah on Thursday December 07, 2023 @12:53PM (#38559460)
John Oliver, Amnesty International [1], Reuters Investigations[2], The US District Court[3]. Just can't trust anyone to not be delusional these days.

[1]https://www.amnesty.org/en/latest/news/2022/09/myanmar-faceb...

[2]https://www.reuters.com/investigates/special-report/myanmar-...

[3]https://globalfreedomofexpression.columbia.edu/cases/gambia-...

Re:Purple Llama by Meta AI by skippyboxedhero on Thursday December 07, 2023 @12:00PM (#38558792)
Have heard about this happening on multiple other platforms too.

Substack is human moderated but the moderators are from another culture so will often miss forms of humour that do not exist in their own culture (the biggest one being non-literal comedy, very literal cultures do not have this, this is likely why the original post was flagged...they would interpret that as someone telling another person to literally set their house on fire).

I am not sure why this isn't concerning: large platforms deny your ability to express yourself based on the dominant culture in the place that happens to be the only place where you can economically employ moderators...I will turn this around, if the West began censoring Indonesian TV based on our cultural norms, would you have a problem with this?

The flip side of this is also that these moderators will often let "legitimate targets" be abused on the platform because that behaviour is acceptable in their country, is that ok?

Re:Purple Llama by Meta AI by ethbr1 on Thursday December 07, 2023 @01:28PM (#38559899)
I mean, most of FAANG has been US values being globalized.

Biased, but I don't think that's the worst thing.

But I'm sure Russia, China, North Korea, Iran, Saudi Arabia, Thailand, India, Turkey, Hungary, Venezuela, and a lot of quasi-religious or -authoritarian states would disagree.

Re:Purple Llama by Meta AI by ragequittah on Thursday December 07, 2023 @01:47PM (#38560128)
>I mean, most of FAANG has been US values being globalized.

Well given that we know Russia, China, and North Korea all have massive campaigns to misinform everyone on these platforms, I think I disagree with the premise. It's spread a sort of fun house mirror version of US values, and the consequences seem to be piling up. The recent elections in places like Argentina, Italy, and The Netherlands seem to show that far-right populism is becoming a theme. Anecdotally it's taking hold in Canada as well.

People are now worried about problems they have never encountered. The Republican debate yesterday spending a significant amount of time on who has the strictest bathroom laws comes to top of mind at how powerful and ridiculous these social media bubbles are.

Re:Purple Llama by Meta AI by ethbr1 on Thursday December 07, 2023 @02:24PM (#38560645)
It's 110% US values -- free speech for all who can pay.

Coupled with a vestigial strain of anything-goes-on-the-internet. (But not things that draw too much political flak)

The bubbles aren't the problem; it's engagement as a KPI + everyone being neurotic. Turns out, we all believe in at least one conspiracy, and presenting more content related to that is a reliable way (the most?) to drive engagement.

You can't have democratic news if most people are dumb or insane.

Re:Purple Llama by Meta AI by ragequittah on Thursday December 07, 2023 @02:50PM (#38560959)
Fully agreed, but the conspiracies are now manufactured at a rate that would've been unfathomable 20 years ago. I have a friend who knows exactly 0 transgender people in life who, when talking politics, it's the first issue that comes up. It's so disheartening that many people equate Trump to being good for the world because they aren't able to make off-color jokes without being called out anymore, or because the LGBTQIA+ agenda is ruining schools. Think of the children! This person was (seemingly) totally reasonable before social media.
Re:Purple Llama by Meta AI by skippyboxedhero on Thursday December 07, 2023 @06:44PM (#38563388)
It isn't US values, it is values from the countries where moderators are hired.
Re:Purple Llama by Meta AI by ethbr1 on Thursday December 07, 2023 @10:06PM (#38564980)
The fact that everyone should be entitled to say whatever pops into their mind is a pretty US value.
Re:Purple Llama by Meta AI by slantedview on Thursday December 07, 2023 @11:29AM (#38558329)
Have you seen political facebook? It's a trainwreck of content meant to incite violence, and is perfectly allowed so long as it only targets some people (ex: minorities, certain foreigners) and not others. The idea that Facebook is playing it safe with their content moderation is nonsense. They are a political actor the same as any large company, and they make decisions accordingly.
Re:Purple Llama by Meta AI by comboy on Thursday December 07, 2023 @11:40AM (#38558473)
I have not, I'm not using it at all, so yes that context may put parent comment in a different light, but still I'd say the issue would be comments that you mention not being moderated rather than the earlier one being moderated.
Re:Purple Llama by Meta AI by giancarlostoro on Thursday December 07, 2023 @12:00PM (#38558799)
I think this is how they saw my comment, but the human who reviewed it was clearly not doing their job properly.
Re:Purple Llama by Meta AI by mcpackieh on Thursday December 07, 2023 @12:02PM (#38558815)
> Context doesn't matter

It's fair to say you've lost the plot.

Re:Purple Llama by Meta AI by VHRanger on Thursday December 07, 2023 @01:19PM (#38559791)
As commenter below said, this sounds reasonable until you remember that Facebook content incited Rohingya genocide and the Jan 6th coup attempt.

So, yeah, context does matter it seems