Discover and read the best of Twitter Threads about #SundayShenaniGANS
Most recents (8)
There's some weird stuff going on with the engagement on this tweet from @ilkersenock, including multiple retweets from accounts with GAN-generated faces. Let's take a look. #SundayShenaniGANs
cc: @ZellaQuixote
cc: @ZellaQuixote
The accounts with GAN-generated faces that retweeted the @ilkersenock tweet in question are part of a network of 18 accounts created on February 17th, 2023 with GAN-generated faces. All of their content thus far is retweets of @ilkersenock tweeted via the Twitter Web App.
Despite the fact that one can produce an endless number of unique GAN-generated faces, this network contains duplicates: @BerkYil82206927 and @GokerCelik5 use the same image, as do @DenizSe18153981/@KorayEr34506957, @GokerKaya7/@TugraAgaoglu3, and @TunaDem21703166/@UmutKorkmaz8.
Meet @Noah__Abrams, @AbbyFeldman19, @MarkPerlman14, and @SamHoffhoff, a quartet of allegedly liberal Twitter accounts created in January 2023 with GAN-generated faces. #SundayShenaniGANs
cc: @ZellaQuixote
cc: @ZellaQuixote
GAN-generated faces produced by popular tools such as thispersondoesnotexist.com and generated.photos have the property that the major facial features (particularly the eyes) are in the same location on every image. This becomes obvious when the images are blended together.
In addition to the facial feature positioning, there are several additional indicators that these faces are artificially generated, such as the mismatched lenses on @SamHoffhoff's glasses, @MarkPerlman14's mismatched ears, and @Noah__Abrams's surreal background.
Meet @prof_freedom, a Twitter account with 49,000 followers (including Sebastian Gorka) and a black-and-white GAN-generated face. #SundayShenaniGANs
(GAN = "generative adversarial network", the AI technique used by tools like thispersondoesnotexist.com)
cc: @ZellaQuixote
(GAN = "generative adversarial network", the AI technique used by tools like thispersondoesnotexist.com)
cc: @ZellaQuixote
One fingerprint of unmodified GAN-generated faces (at least, those made with widely-available tools) is that the major facial features (especially eyes) are in the same location on every image. This becomes obvious when one blends multiple GAN-generated faces together.
There are other signs that @prof_freedom's profile image is not a real photograph of a person's face. The most obvious is the unrealistic teeth; there are also some strange things going on where the hair meets the background.
Here's a look at a lineup of six interesting Twitter accounts with GAN-generated face pics. #SundayShenaniGANS
("GAN" = "generative adversarial network", the AI technology used by thispersondoesnotexist.com and similar tools)
cc: @ZellaQuixote
("GAN" = "generative adversarial network", the AI technology used by thispersondoesnotexist.com and similar tools)
cc: @ZellaQuixote
First, we have @GypsyCeltic (ID 1307840989617434624), an account with a GAN-generated face and a propensity for posting false statements about COVID vaccines. This allegedly Florida-based account is followed by Ron DeSantis campaign operative @ChristinaPushaw. #FloridaGAN
Next up is @dogoym (ID 1346315171229143044), an antisemitic troll account with a GAN-generated pic and a name that references the antisemitic dog whistle "The Goyim Know". According to one of this fake account's replies to Elon Musk, ~30% of Twitter accounts are fake.
It's a great day for a quick look at a variety of suspicious Twitter accounts with GAN-generated faces. #SundayShenaniGANs
(GAN = "generative adversarial network", the technology used by thispersondoesnotexist.com and similar sites to produce fake faces)
cc: @ZellaQuixote
(GAN = "generative adversarial network", the technology used by thispersondoesnotexist.com and similar sites to produce fake faces)
cc: @ZellaQuixote
First up, we have @DrAdamAneevit, an antivax account with 38K followers and a GAN-generated face. Although this account's bio states that it is a "pArOdY", plenty of people are eating up its misleading tweets about COVID and vaccines.
As it turns out, @DrAdamAneevit (permanent ID 949263131800424449) wasn't always called @DrAdamAneevit. Wayback machine archives reveal that this account was previously a "news" account named @/ForFactsSake101.
(archive: )
(archive: )
Meet @Fernandozi2u, @Itzel2xqry, @Jenniferypvw, @Jensen3eigi, @Maximpxh, and @Raelynq1uo, six accounts with GAN-generated profile pics and identical biographies created in May 2021. #SundayShenaniGANs
cc: @ZellaQuixote
cc: @ZellaQuixote
The aforementioned six accounts are part of an astroturf botnet consisting of 2841 similarly-named Twitter accounts with repetitive biographies created in May 2021. These accounts (allegedly) send all of their tweets via Twitter for iPhone.
All 2841 accounts in this network use GAN-generated face pics as their profile images.
(GAN = "generative adversarial network", the AI technique used by thispersondoesnotexist.com and similar tools to generate fake face pics.)
(GAN = "generative adversarial network", the AI technique used by thispersondoesnotexist.com and similar tools to generate fake face pics.)
It's a holiday weekend, so let's look at a mostly-dormant botnet that follows YouTube political commentator @Timcast. #SundayShenanigans
cc: @ZellaQuixote
cc: @ZellaQuixote
We found a total of 1249 very similar looking accounts following @Timcast, created in batches at various points in 2016. All follow a three digit number of accounts but have almost no followers, tweets, or likes of their own.
Most of the tweets accounts in this network have tweeted are retweets, and this is where things get weirder. The tweets being retweeted all have far more retweets than likes, with many having ratios of 100-to-1 or greater.
It's a great day to explore bots that attempt to game retweet-to-win contests! (and possibly ensnare them into retweeting portions of this thread. . .) #SundayShenanigans #SundaySpam
cc: @ZellaQuixote
cc: @ZellaQuixote
We began by downloading 24 hours' worth of retweets containing the word "win". Although most of the traffic is organic, there does appear to be quite the variety of automation apps in play.
We then downloaded recent tweets from each account that used an automation app to retweet a tweet containing "win", and kept accounts matching these criteria:
• at least 50% of tweets are automated
• at least 90% of automated retweets contain "win", "giveaway", or "contest"
• at least 50% of tweets are automated
• at least 90% of automated retweets contain "win", "giveaway", or "contest"
