Slashdot Mirror
Zuckerberg Testimony: Facebook AI Will Curb Hate Speech In 5 To 10 Years (inverse.com)
An anonymous reader quotes a report from Inverse: After a question from Senator John Thune (R-SD) about why the public should believe that Facebook was earnestly working towards improving privacy, Zuckerberg essentially responded by saying that things are different now. Zuckerberg said that the platform is going through a "broad philosophical shift in how we approach our responsibility as a company." "We need to now take a more proactive view at policing the ecosystem," he said. In part, Zuckerberg was talking about hate speech and the various ways his platform has been used to seed misinformation. This prompted Thune to ask what steps Facebook was taking to improve its ability to define what is and what is not hate speech.
"Hate speech is one of the hardest," Zuckerberg said. "Determining if something is hate speech is very linguistically nuanced. You need to understand what is a slur and whether something is hateful, and not just in English..." Zuckerberg said that the company is increasingly developing AI tools to flag hate speech proactively, rather than relying on reactions from users and employees to flag offensive content. But according to the CEO, because flagging hate speech is so complex, he estimates it could take five to 10 years to create adequate A.I. "Today we're just not there on that," he said. For now, Zuckerberg said, it's still on users to flag offensive content. "We have people look at it, we have policies to try and make it as not subjective as possible, but until we get it more automated there is a higher error rate than I'm happy with," he said.
"Hate speech is one of the hardest," Zuckerberg said. "Determining if something is hate speech is very linguistically nuanced. You need to understand what is a slur and whether something is hateful, and not just in English..." Zuckerberg said that the company is increasingly developing AI tools to flag hate speech proactively, rather than relying on reactions from users and employees to flag offensive content. But according to the CEO, because flagging hate speech is so complex, he estimates it could take five to 10 years to create adequate A.I. "Today we're just not there on that," he said. For now, Zuckerberg said, it's still on users to flag offensive content. "We have people look at it, we have policies to try and make it as not subjective as possible, but until we get it more automated there is a higher error rate than I'm happy with," he said.
If humans can't even decide what is "hate speech", what makes anyone believe that an AI system can?
They're already doing a pretty good job of censoring conservative beliefs. Most of my conservative friends on Facebook have had their account disabled for time-out periods or even outright banned. They're doing a good job already.
They do have some automation. I saw a friend post a Pepe picture, and he was banned immediately for that hateful act.
Correction: "Facebook AI will curb free speech"
What Zuckerberg will likely be "curbing" and what Facebook tends to ban:
- claims that racism is not the primary cause of poverty and criminality in minority communities
- pointing out that gender is not a social construct
- using a non-preferred pronoun with a transgendered person
- rude or critical statements about Hillary Clinton (if you use the words she/her, it's automatically misogynist hate speech)
- speech critical of illegal immigrants or advocating the expulsion of illegal immigrants
- speech critical or disapproving of Islam
- anything containing derogatory words for progressive protected classes (but not derogatory words for straight white males)
Of course, the net effect will be that Facebook turns even more into a progressive bubble. And while that may be comforting to progressives, it makes it hard for them to understand why their favorite political candidates or policies don't catch on among Americans in general.
It won't. I've seen video games ban the word "jew" simply because somebody decided that a word filter was needed because the word is often used pejoratively. I found out recently I'm 2% Ashkenazi jew, and I wouldn't be surprised if some AI caught the word "nazi jew" out of that. Worse, is that hate speech is a constantly evolving thing, and the words and double-speak deliberately change on a routine basis. This is why hate speech rules are so fucking stupid: Since they're going after a constantly moving target, it's impossible to legislate or filter without making deliberately vague rules, and you can easily break said rules without realizing it at all. And how can you be expected to know that something is illegal when there isn't even a written law against it?
This gives the government plenty of room to get away with abuse. If you don't already have a reason to arrest somebody, you can just create one on the spot. The UK already does this.
It's even more problematic than that. If they have an AI algorithm which can understand human language they have the ability to filter out anything they do not like, not just hate speech and to misquote Agent Smith from the Matrix: "what good is the right to free speech if you are unable to speak?".
This might be termed politically desanctioned speech (politically incorrect speech being the sloppy cousin) where you are not protected by free speech laws, but originally, hate speech was a exacerbating condition that made "hate crimes" worse. It was not a thing unto itself. Of course, it has morphed into a thing by itself because no one remembers the original purpose. For example, vandalism is a crime, but not necessarily a hate crime. What makes it a hate crime, worse than a regular instance of vandalism is "hate speech" or the motive behind the crime. The Supreme Court has ruled this is not protected by the first amendment because you lose your first amendment protections by committing the underlying crime. This was a dubious ruling, but hasn't been revisited. Of course, it's hard to commit vandalism online unless you are a hacker, but easy to invoke "hate speech", so everyone wants that policed despite there being no underlying crime. Of course, you aren't protected by the first amendment because the courts have not ruled that FB or Google hangouts or whatever are "places of public accommodation". The ACLU used to fight for expansion of the definition of public accommodation, and FB would probably fit the bill, but they have taken it upon themselves to police "hate speech" instead. Of course, politicians would love all this to be policed so they can pressure FB and Google to stamp out political speech they don't like either, and advertisers would love it stamped out so they can't be tagged with boycotts and the like, but if malls have to put up with free speech, then FB probably should too.