How did this tweet from #Texit supporter and lieutenant governor candidate @TheTexianDM get more retweets than likes, mostly from accounts with GAN-generated profile pics?
Answer: @TheTexianDM's tweet was retweeted by an botnet consisting of (at least) 1081 accounts created in September, October, or November 2021. All 1081 accounts tweet exclusively via the Twitter Web app, and almost all of their tweets are retweets. #astroturf
The accounts in this network mostly retweet cryptocurrency/blockchain/NFT content, but there are exceptions - gaming accounts, musicians, and political accounts turn up as well.
All 1081 accounts in this network use GAN-generate face pics as their profile pics, similar to those produced by thispersondoesnotexist.com.
GAN-generated face pics (at least so far) have the telltale trait that the major facial features (particularly the eyes) are always rendered in the same position on the image. This becomes obvious when one blends the 1081 images used by the botnet together:
Other potential indicators of GAN-generated face pics include nonsensical clothing and hats made of abstract colorful blobs rather than any real material. GAN-generated "sports caps" will often have random colorful debris where the team logo would be on a real hat.
GAN-generated face pics also sometimes have trippy backgrounds that often superficially resemble buildings or outdoor scenes but that turns out to be composed of abstract shapes and blobs of varying colors when one looks more closely.
Additionally, vestigial heads near the edges of the image (sometimes referred to as "side demons") and malformed glasses or goggles that sometimes blend into the face or are unrealistically asymmetrical are also fingerprints of GAN-generated pics.
Permanent IDs for the 1081 accounts in this network, in case anyone wants them. (A few are already suspended.) pastebin.com/5TY22SC8
More information on GAN-generated pics and their use on (mostly) Twitter in this compendium o' threads:
This "account vetting guide" has been floating around Twitter on and off for a couple of years. As both a disinformation researcher and a Twitter user, I find multiple aspects of it problematic and therefore do not adhere to it.
First, the "bot detection" criteria are nonsense that will tell you nothing about whether or not a given account is automated. (All accounts are new and have few tweets at some point, plenty of humans tweet hundreds of times a day and make profuse spelling/grammar errors, etc.)
(Longer thread with more information on the detection (and misdetection) of bots, for those who are interested in more detail on the topic:
The coffee-craving cats are part of a 129 account cryptocurrency spam network. The accounts in this network were created in three batches in August and September 2021, and thus far have (allegedly) posted all of their tweets via the Twitter Web App.
These accounts do three things:
• post repetitive tweets, most of which are conversational-ish stuff like "I like to have a cup of coffee in the morning", but with occasioal spammy replies about crypto
• retweet crypto/blockchain accounts
• follow crypto/blockchain accounts
FYI: Despite the banner that says "BREAKING NEWS" in large friendly letters, the @FordJohnathan5 account is neither a news outlet nor a reliable source of information. The "no disinformation" claim is false, as the account has repeatedly presented falsehoods as #BreakingNews.
This is a lie. Neither of the tweets I screenshotted had been deleted by @FordJohnathan5
at the time of this tweet, and had been up for quite some time prior.
Additionally, I made good-faith efforts to get @FordJohnathan5 to address the false tweets LONG before posting this. I was ignored, as were many others who asked for sources.
Some threads on the use of retweet rooms. Users in these rooms can amplify hundreds of tweets per day hundreds of times each with minimal effort on the part of each individual user.
This network consists of 98 accounts created between August 2020 and October 2021. All have few or no followers, and the majority have female names. All 98 accounts (allegedly) tweet via the Twitter Web App.
The accounts in this network do two things:
• Post original tweets, which are generally duplicated verbatim across multiple accounts in the network. 54.8% of these tweets (1918/3499) contain the hashtag #Accelerationism.
• Retweet, reply to, and quote tweet popular accounts.
Meet @AnthonyWill_BSC and @MrThomasKoning, a duo of supposed cryptocurrency experts and "Gem Hunters" with GAN-generated profile pics and over 180,000 followers between them.
GAN = "generative adversarial network", the AI technique used by thispersondoesnotexist.com to generate fake faces such as those used by these two accounts. The major facial features are in the same position on both images, a fingerprint of unmodified GAN-generated face pics.
Both @AnthonyWill_BSC and @MrThomasKoning primarily tweet cryptocurrency/NFT promotions of varying sorts, with no apparent preference for any specific coin. The two accounts follow each other.