SoylentNews is people
SoylentNews
New research has shown just how bad AI is at dealing with online trolls.
Such systems struggle to automatically flag nudity and violence, don’t understand text well enough to shoot down fake news and aren’t effective at detecting abusive comments from trolls hiding behind their keyboards.
A group of researchers from Aalto University and the University of Padua found this out when they tested seven state-of-the-art models used to detect hate speech. All of them failed to recognize foul language when subtle changes were made, according to a paper [PDF] on arXiv.
Adversarial examples can be created automatically by using algorithms to misspell certain words, swap characters for numbers or add random spaces between words or attach innocuous words such as ‘love’ in sentences.
The models failed to pick up on adversarial examples and successfully evaded detection. These tricks wouldn’t fool humans, but machine learning models are easily blindsighted. They can’t readily adapt to new information beyond what’s been spoonfed to them during the training process.
(Score: 1) by anubi on Sunday September 02 2018, @05:33AM (3 children)
I get the idea that Soylent News is also a testbed for this kind of thing, trying stuff out on a smaller scale among technical professionals first.
This is my favorite hangout to get whiffs of techie stuff and insights from what I believe to be the world experts in the field... and I am talking about people who actually are familiar with the technical, not the business, end of things. You guys come up with more leads for me - things I never knew even existed - until one of you guys mention it.
I see this as a specialty site, very similar to TheOilDrum.com ( now on archive status ) was for us oil exploration guys ( which is where I came from ).
It was very important for each of us to be able to submit stuff to the group for comment. And I believe that kind of thing is even more important here, given how critical our computational and network infrastructure is, yet we have hidden agendas from special interests that try to block our understanding of how things work, especially covert backdoors, when we all know that obscurity is NOT security.
Having backdoors in our OS is about like having a detailed plan for building an atomic weapon out there, just waiting to fall into the wrong hands. One slipup, and the wrong party will have the power to "upgrade" our whole computational infrastructure to a brick. And we really need to have an active community to keep that from happening. While we may not have the authority needed to keep business executives from doing incredibly stupid things, we can at least know what's apt to happen and prepare for it within the technical community.
From what I see, this system of community oversight/moderation works extremely well. Kinda like an integrator getting the noise out of a system. Or statistics to get to a more accurate estimation than any of its individual inputs. Like already noted, having a system like this requires a substantial number of us participating in the discussions and moderations, no different than a statistical study requires many samples to get decent results.
Now, the real trick is going to be how do we keep the professionals over here, and keep the kids and jokers over there?
"Prove all things; hold fast that which is good." [KJV: I Thessalonians 5:21]
(Score: 2) by The Mighty Buzzard on Sunday September 02 2018, @10:35AM (1 child)
Well, first you have to figure out how to divide them up when they share the same body. Chainsaws are pretty messy. Axes too.
My rights don't end where your fear begins.
(Score: 0) by Anonymous Coward on Sunday September 02 2018, @01:15PM
Look, I'm not sure that will work.
It's worth a try though.
There's a jerk at my work we could experiment on?
(Score: 3, Interesting) by martyb on Monday September 03 2018, @04:28AM
What it does: The actual moderation is unchanged. The resulting apparent moderation can be filtered by adjusting your Threshold and Breakthrough preferences. So, if you set both of those to "0", then whenever a foe posts a comment, the most you should see is just the comment title. OTOH, when a friend posts a comment, even if moderated into oblivion (actual moderation -1) it will still rise above those limits and you will always see their comments.
NB: That is how it is supposed to work. I only recently remembered this capability in the system and have not tested it. I do not anticipate any problems, but if you DO find a problem please let us know! File a bug, send an email to admin@soylentnews.org, or raise it with someone on staff on IRC.
Hope that helps!
Wit is intellect, dancing.