Alphabet's Tune browser extension uses AI to filter out toxic comments

DPennington

Posts: 88   +32
Bottom line: Jigsaw, cousin company of Google, released an experimental version of the Tune Chrome extension, meant to filter out toxic comments from popular websites and social media platforms. The extension, built on the Perspective API, allows users to filter out comments based on their sensitivities.

Trolls beware: Alphabet offshoot Jigsaw is rolling out a new AI for Chrome that will filter out toxic comments.

The tool, dubbed Tune, will allow you to tune out (see what I did there?) nasty comments from sites like Twitter, Facebook, and Reddit. Tune works by allowing users to select different levels of filtering using a virtual knob to block out toxic comments all together (called "zen mode") or show selective types of posts. The AI can filter out profanity, insults, and racial attacks.

Tune is built on the Perspective API that Jigsaw rolled out in 2017. Perspective allowed third-party developers to integrate their software with the AI, and as a result Tune has a wide range of support at launch. To train Perspective, Jigsaw pointed it at hundreds of thousands of comments deemed toxic by human reviewers. From there, the software looked at other comments and assigned a score based on similarity.

As with all AI, both Tune and the underlying Perspective get smarter every time it's fed more examples of toxic comments. Among those contributing to the testing was the New York Times, which used it to supplement their comment moderators and cut down on their workloads.

Tune doesn't work with the entire web yet, but it does work on sites that are grappling with the worst forms of toxicity. The plugin will moderate Facebook comments, even allowing users to filter out political content that may be false. YouTube and Reddit are both notorious for their sometimes-toxic comment threads, and Tune will help automatically moderate those as well.

Jigsaw admits that their AI tech is still in its infancy and far from perfect. Tune still sometimes miscategorizes comments as toxic when they aren't, but with some tinkering it shouldn't affect your browsing experience too much. No word if Tune will work with online forums for now. Even though we try to keep a clean house here on TechSpot, we rely on a modified version of Xenforo to keep our comments and forums sections. Feel free to install the experimental version and let us know in the comments (nicely) how it works for you.

Permalink to story.

 
I did not infer that. Rather I saw an article regarding the general topic of censorship and shared something which is anti censorship as an indirect rebuttal.

I must assume you don't have any issue with the post because you are staff (no?) and could have simply deleted it if it was in violation of any of this board's rules.
 
I did not infer that. Rather I saw an article regarding the general topic of censorship and shared something which is anti censorship as an indirect rebuttal.

I must assume you don't have any issue with the post because you are staff (no?) and could have simply deleted it if it was in violation of any of this board's rules.

Oh no, of course not. I have no issue with what you posted. I just thought the implication you were making was that Tune was censorship.
 
So does this AI understand subtext or context? From what I see it's simply doing a straight comparison of sentences taken completely out of context or without subtext.

It's be funny when trolls start using common friendly statements as troll innuendo. People will make it as trolling and poof, now that AI thinks that's toxic as well.

Seems like a really easy system to abuse and doesn't address the root of the problem.
 
I'd love a Kardashian\Royal family ai powered filter or adblocker type deal. Maybe even a political blocker, especially on election year. If people don't like being censored I'm sure it's not to hard to find another forum or social media platform.

I hope this AI doesn't go mad from all the trolling and decide to nuke us.
 
You didn't...you just acted like you did

I disengaged in a snarky way because it's not worth the debate, here or anywhere. If you wanna see that as a win for you and/or a loss for me or just be indifferent, it makes no difference in any case.
 
So does this AI understand subtext or context? From what I see it's simply doing a straight comparison of sentences taken completely out of context or without subtext.

It's be funny when trolls start using common friendly statements as troll innuendo. People will make it as trolling and poof, now that AI thinks that's toxic as well.

Seems like a really easy system to abuse and doesn't address the root of the problem.


The root of the problem being the trolls themselves? hmmm - how to root out that problem indeed.
 
This is a great extension and I can't imagine how it could be circumvented. Great work, Jigsaw. You solved trolling online! (y) xD
 
How AI is this AI? It's just some shitty algorithm that won't have any effect. People will always think of new ways to insult other people.

And what's with people who can't handle words on the internet? You can't go anywhere on the internet and expect every person to accommodate you. There will be some people with different views on things. It's just the way this generation is raised and that's a much deeper problem.

What I think these companies are ultimately trying to do is make a "perfect" world where everyone is happy and no bad things can happen by using censorship and propaganda but that is laughable cause you'll never achieve that in the real world.
 
I'd love a Kardashian\Royal family ai powered filter or adblocker type deal. Maybe even a political blocker, especially on election year. If people don't like being censored I'm sure it's not to hard to find another forum or social media platform.

I hope this AI doesn't go mad from all the trolling and decide to nuke us.
Maybe it would be, if the censor happy snoflakes didnt keep trying to shut down every forum and website that doesnt play by their rules. These spinless whiners cant take banter, cant take anything that isnt positive arse-patting, and since they have no life outside of their internet hugboxes, spend all day shrieking at hosting sites and payment processors until any site they dont like is forced off the open internet.
 
Maybe it would be, if the censor happy snoflakes didnt keep trying to shut down every forum and website that doesnt play by their rules. These spinless whiners cant take banter, cant take anything that isnt positive arse-patting, and since they have no life outside of their internet hugboxes, spend all day shrieking at hosting sites and payment processors until any site they dont like is forced off the open internet.
You can make your own forum for free.
https://superbwebsitebuilders.com/how-to-build-a-forum/
 
Back