this post was submitted on 11 Oct 2023
2002 points (98.3% liked)
Technology
59446 readers
3632 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
I’m skeptical… how are the fake reviews identified and how do you avoid flagging real ones?
They’re just building Fakespot into the browser so the same way Fakespot does, by analyzing the user who posted the review
What does "analyzing" mean?
They detect when a whole bunch of reviews are posted at exactly the same time, or are posted on a fixed schedule, or use extremely similar language, or with a brand new account...
Basically they use spam-detection techniques on reviews.
What does “techniques” mean?
It’s that Lego that’s slightly more advanced
Algorithms.
What does "algorithms" mean?
stahp
ppl gonna stop answering my real questions and I'll be tech illiterate firebrand
Heuristics.
Probably ai lol
It could be something like that (hint: they already deployed an offline neural network in Firefox with which you can translate web pages), and the idea would be to detect AI-generated content.
Well I hope they're going to do better at detecting AI content than anyone ever has before because nobody's done it well at all so far.
There's an inherent problem here that AI produces results similar to what it's trained on and it was not trained on robotic input it was trained on natural human language online.
Well it will be, because it's detecting AI-generated content indirectly. What it's directly detecting are bot posters, which are much easier to spot.
"AI detectors" have the uphill job of having to figure out whether something is generated by looking only at what was generated. Fakespot and tools like it get to use the metadata, which has many telltales that bots aren't even trying to hide.
I think for me personally they can fuck right off with this. It's unwarranted and invasive. Maybe some fat asses need to get off the couch and stop ordering so much shit online. ( any perceived negativity here is my disappointment in Mozilla not negativity directed at you)
IDK chief. It seems like one of those things that are hard to do in theory as you said, but relatively easy in practice.
I mean just about any human who has played a bit with ChatGPT nowadays is able to identify ChatGPT generated paragraphs within a few words. I don’t suppose it would be much harder for a machine.
Therein lies the issue though. If its not hard to detect, then right after that, its hard to detect again, because the previous fix has been trained out/around. The harder we work to develop detection, the harder we work to ensure detection avoidance is advanced in parallel.
Elsewhere in this thread someone explained that its just integrating FakeSpot into the browser, which uses basic email spam detection techniques to detect fake reviews by analyzing how the reviewer posts. Is there a set schedule they post reviews by, what else have they reviewed, how new is the account, etc. A 2 day old account with 20 reviews would be an obvious source of fake reviews for example