Follow Slashdot blog updates by subscribing to our blog RSS feed

 



Forgot your password?
typodupeerror
×
AI Android Censorship Chrome Communications Google Social Networks The Internet

Alphabet's AI-Powered Chrome Extension Hides Toxic Comments (engadget.com) 196

An anonymous reader quotes a report from Engadget: Alphabet offshoot Jigsaw is launching a Chrome extension designed to help moderate toxic comments on social media. The new open-source tool, dubbed "Tune," builds on the machine learning smarts introduced in Jigsaw's "Perspective" tech to help sites like Facebook and Twitter set the "volume" of abusive comments. Using "filter mix" controls, users can either turn toxic comments off altogether (what's known as "zen mode") or show selective types of posts containing attacks, insults, or profanity. Tune also works with Reddit, YouTube and Disqus. Jigsaw admits that Tune is still an experiment, meaning it may not spot all forms of toxicity or could hide non-offensive comments. "We're constantly working to improve the underlying technology, and users can easily give feedback right in the tool to help us improve our algorithms," C.J. Adams, Jigsaw product manager, wrote in a blog post.
This discussion has been archived. No new comments can be posted.

Alphabet's AI-Powered Chrome Extension Hides Toxic Comments

Comments Filter:
  • by Anonymous Coward on Wednesday March 13, 2019 @09:08AM (#58266686)

    Also known as the, "I'm not mature enough to have my beliefs challenged!" SNOWFLAKE mode

    • by lgw ( 121541 ) on Wednesday March 13, 2019 @09:42AM (#58266904) Journal

      Also known as the, "I'm not mature enough to have my beliefs challenged!" SNOWFLAKE mode

      Or, more concisely, "rightthink mode". Soon to be mandated in China.

      • It's really incredible how quickly a product developed in the US, by a US-based company, to sell to US-based culture, automatically gets attributed to China. Almost like the US is quite sophisticated at directing the national narrative such that even our own output is reattributed to our global rivals in less time it takes to think a thought. And if Slashdot moderation points mean anything, this happens to applause!

        • Re: (Score:2, Insightful)

          by AmiMoJo ( 196126 )

          Nah, it's just the freeze peach extremists who consider any failure to amplify and broadcast their bullshit to be an assault on liberty and literal genocide.

          • by AHuxley ( 892839 )
            Freedom and free speech is great.
            • by AmiMoJo ( 196126 )

              Are great.

              • by AHuxley ( 892839 )
                Considering the censorship now been used by brands commenting on what they see as sinful and what needs to be hidden?
                The greatness of the internet was in the ability to publish, comment and share links without censorship.
                • by AmiMoJo ( 196126 )

                  "censorship now been used by brands commenting on what they see as sinful"

                  What is this? Seems like they would do the opposite, e.g. Gillette kept YouTube comments open because the triggered snowflakes just amplified their message.

                  • by AHuxley ( 892839 )
                    When a brand selects what publication, comments, links, words, books, cartoons, reviews memes, movies, grammar, art, politics it think is sinful that's censorship :)
                    Time to enjoy some of that great freedom of speech and freedom after speech.
                    • by AmiMoJo ( 196126 )

                      So... You have everything except 8chan? Even Infowars and Brietbart heavily "censor" content by your standard.

                      Maybe Gab doesn't, but they do remove spam sometimes so I guess freedom is dead on that platform.

                      Oh and Slashdot doesn't let you post certain words, so I guess you feel oppressed here too.

                    • Comment removed based on user account deletion
    • by Anonymous Coward

      I should think it looks for inflammatory language or anything promoting outrage/polarization, which you can pretty much guarantee will be a low value contribution. People can always rephrase in a measured way.

      If they get this right, it will be a serious boon for the internet.

      • They won't get it right. But if there is an option to deemphasize such comments, such as greying them out but still showing them, and providing an easy way to declare whether they are right or wrong in their classification, then they could have the start of a learning classification system that could eventually work as advertised.
        • They won't get it right. But if there is an option to deemphasize such comments, such as greying them out but still showing them, and providing an easy way to declare whether they are right or wrong in their classification, then they could have the start of a learning classification system that could eventually work as advertised.

          Not likely. If you give the users direct interaction with the learning mechanism it will become horribly polluted. How long has /. been around now and we continue to see Troll mods to express disagreement.

    • by TimothyHollins ( 4720957 ) on Wednesday March 13, 2019 @12:21PM (#58267692)

      I very much doubt the AI can spot challenging views. It is however capable of identifying most slurs and invective, making it effective in reducing the number of low-content posts and pointless insults. Rest assured that you are still free to post Trump propaganda or whatever you wish as long as you are capable of doing so without adding personal attacks.

    • There's "oh noes my safe space!" and there's the GNAA and it's Russian equivalent.

      I use spam filters to kill junk comments all the time. Why wouldn't I?

      Not saying it can't be tricky to get out of a bubble. I'm following Bernie Sander's campaign pretty closely and barely noticed Kamala Harris; though to be fair that's mostly because she's quietly raising funds from mega donors at the moment...
  • Glasses (Score:5, Funny)

    by religionofpeas ( 4511805 ) on Wednesday March 13, 2019 @09:08AM (#58266688)

    I hear they are working on improving the technology so it can be used for peril sensitive sunglasses.

  • Fucking idiots (Score:1, Informative)

    by Anonymous Coward

    I for one welcome our censoring AI overlords.

  • by Anonymous Coward

    You're selecting for passive aggression, dog-whistles and subtle memes.

  • So this will end all false outrage right. There is no reason to still be reading whatever you don't agree with.

    • Re: (Score:3, Insightful)

      by AmiMoJo ( 196126 )

      Outrage is big business on YouTube, they will never ban that. People like Carl Benjamin have popular weekly shows that are nothing but outrage. One single tweet expressing mild concern at a trailer spawned nearly 100 outrage videos [youtu.be], and that's pretty normal.

      • Outrage is big business on YouTube, they will never ban that.

        I wondered who would mention YouTube first. They could probably do a better job of blocking toxic comments by just shutting them off on YouTube entirely...

      • Outrage is popular in normal news media as well. How quick you forget Covington and Smollett getting major coverage. Terry Crews had to apologize for saying fathers are important on the View.

  • by Anonymous Coward on Wednesday March 13, 2019 @09:13AM (#58266710)

    1. Toxic to whom? Who but me can decide that for me? (Nobody, if I'm still an individual and not a passive-thinking swarm entity.)
    2. Why would I want to ignore them, given that there are still real people behind them. (Even when they use automation to repeat them.) Those people have a reason they post that. Maybe they are mentally ill. Maybe they have been traumatized. Maybe they are right, but contradict our society's wrong expected norms. Maybe we just don't like how they make us feel. Like disfigured people.

    MAYBE then we should fix the underlying causes, instead of looking away and letting it grow, becoming a problem for Tomorrow Homer.
    Maybe those comments remind us that we should lift our lazy asses and fix this rotten world that is broken only because we don't do shit about it!
    Maybe we want to ignore that we're ashamed of that too.
    Maybe, .. hopefully, ... that won't work.

    • Most people who deal with mentally ill people online are not qualified to help them.

    • The fundamental cause is that unlike in meatspace, on the internet one bored asshole can appear to be 50,000 people and drown out a city of reasonable people. But do you really want to fix that problem at the source, since the fix would probably be some form of ID requirement or pervasive spying or prosecution for speech? I'd rather we build filters.

  • by Anonymous Coward

    Yes of course, at first they started to read all emails in the world in order to block spam messages, now they want to control everything people post over internet. Good job Google! Very nice! My site would never have any Facebook or Google trackers or services. I wrote my own spam filter and it works really fine, so privacy of clients stays intact.

  • Just what we need. (Score:2, Interesting)

    by Anonymous Coward

    Great, just what we need: MORE silencing of dissenting voices. This time under the cry of 'toxicity.'

    Toxic: adjective - any subject, word, phrase, or idea which a person of left-leaning political views disagrees with.

    Except when it's leftists poking their noses into (or destroying) right-leaning areas, it's not toxicity, it's 'diversity of thought' and "you need your echo chambers broken up to prevent radicalization!!!"

  • If you need any more evidence that Alphabet trolls Slashdot, down voting everything that doesn't agree with their company goals, look no further than articles like this one. Everything comment even slightly critical of their new Snowflake mode, will be down-voted nigh instantly.
    • Still waiting....

      • I would, but I've been shadowbanned from moderation. Feh.

      • by nip1024 ( 977084 )
        Apparently you need to refresh the page. Look at all the comments that have already been zeroed. Notice anything?
        • by macraig ( 621737 )

          Since I don't filter for score and hide nothing, I don't know why I didn't see the other posts. There's zeroed posts, to be sure. Is there a pattern to the content of them, or am I selecting for bias? How big a cabal would it take to pull that off? I haven't bothered to moderate in an age, but a single user gets at most five moderating opportunities, right? Could half a dozen pull it it off? They'd have to build up the karma to be eligible to moderate. Are people whoring their karma to the highest bi

  • by jellomizer ( 103300 ) on Wednesday March 13, 2019 @09:27AM (#58266806)

    The problem is with every 10,000 trolls out there is there one spark of genius, with an opposing idea, that is well thought through and should be considered.
    While I would like that that idea to be shared to help diversify our frame of thought, there is the other 9,999 trolls, just meant to enrage us, think it is a stupid joke, repeating the same old disprove message, and lie to us enough times where we think it is true.

    Sometimes we need negative speech, we need to alert people of a major problem, even if it hurts someone feelings, or goes against the cultural norm. Not all problems can be solved with a careful compromise, sometimes you are right and the greater population is wrong.
    However probability is the case is you are in the wrong, and you are just trolling.

    I would rather see, increased education in spotting fake news and trolling, learning to ignore or block message. Learning to be conscious of your personal biases, and not jumping onto the bandwagon, just because you bias says this is good, as you go on the sliding scale toward evil.

    • by nip1024 ( 977084 )
      Disagreeing with someone isn't Trolling.
      • by jellomizer ( 103300 ) on Wednesday March 13, 2019 @10:56AM (#58267294)

        It is when it becomes a personal attack.

        For example if you call the person stupid for their point of view, make assumptions of their education and parentage, sexual preference...

        You can disagree with someone without trying to dehumanize them, and also trying to dehumanize a group of people.

        In the 2016 Election Clintons biggest Faux Pas was calling Trump supporters "deplorables" because that was trolling on her end and stating "This group of people I don't care about their concerns and I will classify them as sub human" While Trump in my opinion is the bigger troll, he was playing a different game, Trumps game was to get People to Hate Clinton (Who has a lot of political baggage), while Clinton game was to make more people like her. Her statement, caused a lot of people on the fence to dislike her more then ever.

      • by Shotgun ( 30919 )

        From your mature thinking, it is obvious you are not getting mod points.

      • Disagreeing with a troll isn't Trolling.

        FTFY

    • by AmiMoJo ( 196126 ) on Wednesday March 13, 2019 @10:10AM (#58267084) Homepage Journal

      The problem is with every 10,000 trolls out there is there one spark of genius

      The is the same argument used against email spam filters. 1 in 10,000 mails might be important and you will end up missing it. Even so, most people prefer to have the spam filter on.

      Some people won't use this filter. Some, perhaps those with kids or more limited time or who just don't want to deal with trolls today, will turn it on. That's fine, it's not their responsibility to listen to everything being said. For those who want to browsing at -1 is available.

    • Easy enough to thwart this filter: Present your "genius" spark with civility and respect to the recipients.

      • Exactly this! The toxicity detector doesn't know whether you're posing difficult questions and challenging prior conceptions. And it sure as shit doesn't know how people are going to feel about a comment. It's detecting the types of language used in toxic comments. That's it. Keep the naughty words to a minimum and you'll be golden.
    • /. does something like this (albeit less sophisticated) with the old "Natalie Portman / Hot Grits / Greased up Yoda Doll" spam posters. If you're one of the older /.ers you probably remember a time when the GNAA (don't google it at work) made the site basically unusable. You're not just dealing with bad actors, you're dealing with bad actors with scripts written by asshats with programming skills.

      At some point you have to have automated comments. There's some stuff even /. won't let you post (real post
      • Which is why you can set the message threshold to whatever level you want. You're not going to find many GNAA posts at +1 or above.

        Or are you talking mostly about 'trolls' that you disagree with?

  • Echo chamber (Score:5, Insightful)

    by DarkOx ( 621550 ) on Wednesday March 13, 2019 @09:31AM (#58266842) Journal

    So the big fear is that these 'online communities' become echo chambers, that re-enforce ideas. The response of the high minded folks at Google apparently is to make sure you can take your echo chamber with you everywhere you go.

    What is a toxic comment anyway. My guess is its any idea Google execs don't agree with.

    • Re: (Score:3, Interesting)

      by AmiMoJo ( 196126 )

      There is a demo of the system with examples of comments that are considered "loud" here: https://www.perspectiveapi.com... [perspectiveapi.com]

      Scroll down a bit. You can drag the slider around to see the filter in action. In the example climate change skeptics are allowed even on the most conservative filter setting, as long as they can express themselves in civil language.

      On the US election example even putting the slider to about 75% still shows the "your[sic] a socialist snowflake!" comment.

      • by DarkOx ( 621550 )

        Okay but

        Donald trump is the worst person to be president. This election really showed how stupid the U.S. is

        gets flagged as pretty toxic. I don't consider it a terribly productive comment but it certainly could be a valid statement in part of a larger post about voter apathy, voting rights, access etc.

        Its a valid string in this conversation! I see this kind of AI us unhelpful as far as the free exchange of ideas go. There is enormous potential to block valid conversation and exchange if you crack it up. If you dial it down to where almost anything is allowed it appears to be no better than a basic pro

    • by Shotgun ( 30919 )

      What is a toxic comment anyway.

      Anything written by Mark Twain. I mean was a white guy that had the audacity to write the word "n1gger". Just ignore the fact that the only characters in "Huckleberry Finn" using the term were completely ignorant blithering idiots, and the ones targeted with it were black and noble. It is the word that makes it toxic.

      I had to mangle the word "n1gger" to get past the toxic filter. How appropriate is that?

    • Just how advanced do you think this algo is? You actually believe that it can understand an "idea"? Unless some major advancements in NLU happened while I was asleep last night, the proposed toxicity detector is going to be a glamourized swear word detector. There is no way it is capable of what you're accusing it of.
    • the big fear is that these online communities become basically useless because of abusive troll posts.

      A toxic comment isn't too hard to identify, especially for somebody as sophisticated as google. Using racial slurs is a dead giveaway. Also excessive cursing (excessive being relative to the length of the post), common insult words like "idiot" and "stupid", etc. As for how to calculate the thresholds, you have users mod comments then you run the modded comments through one of the many common text algor
  • Hope for Slashdot comments.
    • Oh... maybe that explains why, after I fired up Chrome this morning and let it update, yours was the only comment on this story?

  • Of course I am sure that Google considers that to be toxic...

    Censorship of any kind is wrong.

    • Censorship of any kind is wrong.

      So you read /. @ -1? I know I don't unless I have mod points.

    • Censorship of any kind is wrong.

      Not when it is done in good faith to actually facilitate free speech and maintaining protocol for speech to be heard. In that sense, there is a difference between censorship and the moderation that is necessary for the free flow of ideas.

      Where censorship is wrong, is when it is done in bad faith, to tilt an otherwise level playing field in one direction or another.

      I'll agree that there is no reason to trust Google has any intention to moderate in good faith, at all.

  • Wouldn't that render pretty much half of all social media unreadable? The focus on how you can finely tune the threshold is a dead giveaway a little bit tunes it all out.
  • by Anonymous Coward

    He's just going to troll conservative people. We should censor him.

  • by argStyopa ( 232550 ) on Wednesday March 13, 2019 @10:07AM (#58267066) Journal

    I'm sure this won't insulate people even FURTHER into their own personalized bubbles of self-confirmatory groupthink.

    Hint: in a democracy, sometimes people say shit you disagree with

    and

    Hint: Sometimes people say things that hurt your feelings. Sometimes deliberately! It's your job as a grownup to ignore them.

    • Re: (Score:3, Insightful)

      by AmiMoJo ( 196126 )

      In a free democracy no-one forces you to listen to them. The telescreen has an off button.

      • In Amimojo's 'free democracy', the telescreen only has one channel, and if you turn it off Antifa shows up at your house.
  • "Toxic comments" will be whatever Google doesn't like.
    • What? You don't trust Google and their black-box algorithms to regulate speech on the internet? What could possibly go wrong?
  • This is the fruit of research into Dragonfly for Google?
  • There couldn't possibly be negative effects...

  • Per the perspectivesapi.com website, "toxic" is:

    "What's toxic?

    This model was trained by asking people to rate internet comments on a scale from "Very toxic" to "Very healthy" contribution. Toxic is defined as... "a rude, disrespectful, or unreasonable comment that is likely to make you leave a discussion.""

    That definition literally states toxicity is according one's emotional sensibilities and reactions thereof. This is exactly how echo chambers are built.

  • It might work with people who were accidentally being offensive, but if a trolls wants to annoy you, no AI based system is going to stop them.

  • Will always allow its own approved ads.
    People publishing their own ideas, thoughts, links, reviews and comments get hidden by an ad brand.

Dynamically binding, you realize the magic. Statically binding, you see only the hierarchy.

Working...