Remember meForgot password?
    Log in with Twitter

article imageOp-Ed: New Chrome AI custom app Tune hides toxic comments

By Paul Wallis     Mar 12, 2019 in Internet
Mountain View - A new open source Chrome app called Tune has been developed by Alphabet/Google subsidiary Jigsaw. It hides and moderates toxic social media comments using machine learning techniques.
According to Jigsaw, the app focuses on text, identifying profanities, attacks, etc. and can be customized by users. The idea isn’t all that new. This is a somewhat more advanced version of older hardwired site editing programs which used to refuse to type specific words. The modern version of this using AI has been working as an anti-spam asset for a while now. Google currently uses an AI-based anti-spam filter called Tensor to protect Gmail, for example.
On those older programs, if you tried to include Summa Cum Laude, it wouldn’t publish that expression because of the word Cum. They were pretty basic, but it was the same general idea. This is a much more advanced version, and giving users some sort of comeback against the trolls is after all a good idea. This software can adapt and tailor itself to user needs, another big plus.
The issues
The idea of reducing the dribbling ravings of online nutcases, paid trolls and Alt Right mediocrities in their occasional out of jail moments is welcome. However, there’s more to this situation, and it’s a lot more complex than most users realize. Given the downright insane level of trolling online, the disinformation campaigns, the hate speech, the pure sleaze, etc., you’d think this was a dream come true for social media users. Anything which removes the morons has to be good, right?
Not quite the whole story, though. There’s another issue which is much less obvious. The question is whether this solves the problem, or makes it worse.
1. The online trolls are easily visible. Even through multiple proxies, fake IDs, etc., they can be located.
2. These proven idiots leave a virtual audit trail, a clear record of everything they do online. That’s how they get convicted. Even the “dark web” isn’t really invisible, just harder to find, works on pretty much the same principles as the basic web. (The dark web has a lot of vulnerabilities which make it anything but dark, but hey, they can find that out for themselves.)
3. Finding them is the critical issue in shutting them down. If the trolls get the idea they simply can’t post, they’ll become harder to track.
4. Intelligence and law enforcement services online track the serious trolls and nutters in various ways. Does Tune affect their ability to monitor them? (There’s a way of getting round that, but I’m hardly about to spell it out.)
Good idea, but needs to evolve, preferably fast
I don’t want to give the impression I’m in any sense anti-Tune. It’s a great way of fixing the miseries of online experiences for so many people, particularly all those poor kids, in its stated form. I just think it will need to evolve and develop itself to manage foreseeable future scenarios.
Think about this:
• AI-based malware, including workarounds for things like Tune, is either here or it's coming. I’d suggest Jigsaw gets moving on finding any vulnerabilities, ASAP. Get ahead of the hackers, and you may well have solved the problems. Also bear in mind that malware code has a lot of "scatology" it has to use to work....?
• Add a security-based version of Tune to antivirus software, and you could well have a huge commercial asset for personal, site, server, and network security, a whole new level of security, in fact.
• Tune could have another extremely valuable life as a repurposed security tool, using the AI to monitor and block malware, or even rework it and send it back to the malware users. (AI based penetration testing and macrosystem security could be a truly great move, and how much do people who think of things like this get paid? I wouldn’t mind getting paid that sort of money to actually solve these problems.)
• AI software can work systemically, and extremely quickly. It could, in theory and practice, “immunize” itself against the attacks it studies, while neutralizing them in real time. An AI system could be turned in to a very effective obstacle course for malware progressions.
If you’re thinking this evolution may involve some very heavy duty number crunching, you’re right. The big difference with AI is in its inbuilt efficiencies. One thing that’s been bothering me since IBM’s Watson is that people seem to have forgotten you can learn from AI, as well as simply set it to work on various tasks.
Tune could be a very important learning tool for a whole range of online malaises, and a massive asset in terms of delivering dashboard level information regarding threats, trends, and the rest of the major site information shopping list, with active supports for site owners and users. The trolls are a minority, and if both users and sites have more rocks to throw at them, the Golden Age of Trolling, and hacking, could well be over.
Then maybe we’ll be able to use the net as intended. Worth a shot, do you think? You can check out Tune at the Chrome Web Store
This opinion article was written by an independent writer. The opinions and views expressed herein are those of the author and are not necessarily intended to reflect those of
More about Google, Alphabet, Jigsaw, Tune AI Chrome app, custom social media filtering
More news from
Latest News
Top News