"This Company Thinks It Can Solve Diversity With 100,000 Fake AI Faces"
Sep 20, 2019 vice.com/en/article/mbm…
The first documented instance of large-scale disinformation using artificially generated images.
"How a Pro-Trump Network Is Building a Fake Empire on Facebook and Getting Away with It"
Nov 20, 2019 snopes.com/news/2019/11/1…
Facebook would soon remove the disinformation network and AI-Generated accounts tied to The BL/Epoch Times.
"Facebook says a pro-Trump media outlet used artificial intelligence to create fake people and push conspiracies"
Dec 20, 2019 nbcnews.com/tech/tech-news…
"Dating apps need women. Advertisers need diversity. AI companies offer a solution: Fake people."
Jan 7, 2020 washingtonpost.com/technology/202…
This article provides some background information regarding my experience being on the receiving end of a harassment campaign waged by a person who had created dozens of AI-Generated accounts.
More than a dozen accounts with StyleGAN profile images have been posing as experts to amplify positive articles about Huawei. Huawei execs sparked suspicion when they amplified the fake accts 60+ times.
"Facebook has removed a network of accounts with AI-generated faces linked to the Russian troll factory responsible for meddling in the 2016 election."
Dec 2021:
Fake account with StyleGAN profile picture generates publicity for white nationalist march— later reveals itself as member of white nationalist group. logically.ai/articles/fake-…
• • •
Missing some Tweet in this thread? You can try to
force a refresh
The 99 accounts in this network have tweeted 14,600 times. 780 of those tweets contain media. All 780 are screenshots of fake tweets supposedly written by Elon Musk. But what's the motivation behind the fake tweets? Let's look. 1/
Here are the images. Elon Musk did not tweet any of these. Each of the 7 fake screenshots encourages users to google specific types of watches, they're apparently so great that you'd be thanking the fake Elon tweets "later".
DON'T GOOGLE THEM. I did it for you 🙂 2/
Googling leads you to two domains; "outerwatches[.]com" and "otcwatches[.]com". Both domains show warnings of a "disease vector" and "scam" on open. The latter of the two domains was blocked and had to be accessed via archive.
The people behind these networks of bot accounts recognize that StyleGAN faces give them the power to mass-generate accounts that appear authentic. A small SEO botnet seems harmless but this same tech has been used by mega corporations and int'l govt's for propaganda campaigns.
I know that I sound like a broken record, but we need to be educated on how to identify AI images en masse BEFORE the tech advances to the point of indistinction.
I'll include some recent examples of how this tech is being used in the next few tweets.
Thread: On May 20, Breitbart threw some red meat to its readership. Pure outrage-bait. The article inspired 8,000 comments, 1,400 of which on @disqus, I spent a couple of days reading and documenting the response to this article. Here's what I found. 1/ @slpng_giants
My focus was on the @disqus comment section. Disqus has a TOS for users and publishers called the "Basic Rules". First rule: "Hate speech and other forms of targeted and systematic harassment of people have no place on Disqus". Let's see how this goes. 2/ help.disqus.com/en/articles/17…
I began documenting a few days ago. The process takes some time and has to be done in small batches to preserve my sanity. 😅
The first thing that stuck out to me was the general tone of the most upvoted comments. I tweeted about it on May 22. 3/
Last summer, a very determined and very angry individual created several dozen fake accounts explicitly to harass and intimidate me and others who often tweet to @slpng_giants.
AI-generated faces were used to make the accounts appear authentic. (1)
It's not the first time that a person has been harassed on twitter dot com and I'm sure it won't be the last.
What's troubling is the tech. Untraceable faces of people who don't exist. The fake faces are meant to give the appearance of authenticity and it works. (2)
Had it not been for this Medium post by @kcimc I'd have had no idea what I was looking at. The tech has changed a bit since, but many of the artifacts in the images can still be identified -- missing earrings, warped backgrounds, misshapen teeth, etc. (3) medium.com/@kcimc/how-to-…
THREAD: I tend to take anonymous commenters with a grain of salt, particularly when they're contributing to the conversation on Breitbart but I feel it's important to draw attention to how Breitbart's audience has responded to the President's "Civil War" tweet. @slpng_giants
The comments in this thread were taken from the first 1/4 of the article's comments. I grabbed about 100 before needing to step away. It's disheartening and alarming to see an entire community drift toward violent extremism without seemingly any accountability.
The comment section on Breitbart is powered by @disqus. In 3 years of documenting Breitbart's comment section, I've flagged about 5,000 comments that violate Disqus' TOS. They know it's a cesspool and they've chosen to ignore it. I'd like to know why. help.disqus.com/en/articles/17…
THREAD: This past week, I ventured into the comments section of Breitbart’s article documenting the timeline of the mass shooting in Christchurch, New Zealand. In total, I collected about 500 comments. Roughly 10% of the total comments on the article. Here’s what I found. 1/
There were a few categories that dominated the comment thread. Twenty percent of the 500 comments collected expressed some form of joy or happiness over the results of the mass shooting where 51 people were murdered. 2/
Forty percent of the 500 comments collected were anti-Muslim, despite the article being a timeline of an attack on a mosque. This kind of Pavlovian response is evidence that Breitbart has trained its audience to voice outrage when the headline includes the word Muslim. 3/