A clip from Stray got me banned from Twitter
fanbyte.com
A clip from Stray got me banned from Twitter
1–10 of 577 posts
Re: A clip from Stray got me banned from Twitter
#2Its clear that the broken element in this process is not the AI banning the writer of the post, but the poor presumably human review
Re: A clip from Stray got me banned from Twitter
#3A few people commenting on the post say they to had a similar bans when clipping through the world and a grey screen was shown. Maybe the solid gray is triggering the AI to pick any policy violation that had gray in it once...
Re: A clip from Stray got me banned from Twitter
#4Its clear that the broken element in this process is not the AI banning the writer of the post, but the poor presumably human review
Good human reviews are, well, nearly impossible at the scale of Twitter or Google. It would require employing hundreds of thousands of reviewers, training them, and paying them decent salaries for one of the most horrible jobs in the world. So, everybody is trying to automate as much as possible in this area. I am sure that in this case, if there is a human even involved in the escalation, it takes them a few seconds to decide on each case, with poor decision quality.
Re: A clip from Stray got me banned from Twitter
#5Its clear that the broken element in this process is not the AI banning the writer of the post, but the poor presumably human review
I'm not so sure I agree.
The typical argument for AI moderation in these situations is something along the lines of "Twitter is so big how could we possibly moderate at this scale without AI?"
Does a company have a right to scale faster than it has the ability to properly moderate? Should they?
Looking at a more extreme example: If General Motors used AI to design cars and they increased the likelihood of fatalities, would we feel any different?
Re: A clip from Stray got me banned from Twitter
#6Its clear that the broken element in this process is not the AI banning the writer of the post, but the poor presumably human review
An AI arbitrarily banning someone for a completely innocuous post isn't broken?
Re: A clip from Stray got me banned from Twitter
#7It seems like the problem with many of these tech companies (and many modern companies in general) is that they've scaled faster than they have the ability to operate properly.
"How can we possibly moderate properly without AI?"
"We don't want to pay the money it's going to cost to moderate effectively and we're going to cross our fingers that we can solve problems caused by technology with more technology."
Re: A clip from Stray got me banned from Twitter
#8> But there’s also a larger thing here where Twitter apparently has moderation tools so powerful that they can immediately trigger bans for things other than what they’re actually looking for. The actual execution leaves a lot to be desired here, but the itchy trigger finger cannot be denied. Somehow they can do this, but racial slurs are generally pretty okay, white supremacy is largely ignored, and every Twitter troll I’ve reported for harassment gets a three-week investigation and then a “We find they did not violate our rules” email.
This is one of the things that annoys me most about major sites like Twitter or YouTube. Inconsistent enforcement of rules and little to no avenue for appeal.
Re: A clip from Stray got me banned from Twitter
#9Its clear that the broken element in this process is not the AI banning the writer of the post, but the poor presumably human review
Human review is pretense, it costs more to do anything else. A human is told to process these as fast as possible, so the issue can be resolved. With these incentives, it wouldn't be reasonable to actually watch the videos in question. That's not what the system is designed to reward.
The human is there to rubber stamp the AI decision as fast as possible, so that Twitter can tell you your case has been reviewed by a human and settled. The important metric is speed, because that directly translates into cost.
Re: A clip from Stray got me banned from Twitter
#10> Overnight, I received a response that my appeal had been denied. So a human being, someone who works at Twitter dot com, looked at that video, looked back at the rule it was breaking, looked once again at the video, and went “Yeah, this all checks out.”
My guess is that appeals are either processed by some simple code or offloaded to some cheap labour service - such as Amazon Mechanical Turkey or some sweatshop in a developing country.