Twitter is moving ahead with its plan to make Twitter Blue one of its main sources of revenue but ahead of the company is a long battle with bot accounts utilizing artificial intelligence.
Musk took over Twitter back in October, and since then, the Tesla CEO has made a plethora of changes while also dramatically reducing Twitter's total number of employees. One of its main focuses for Musk is bringing value with Twitter Blue, which now gives a blue "verified" checkmark to any account that purchases it. Before Musk's takeover, the verified checkmark indicated the account was authentic, while Musk's pay-wall system has given nefarious actors an easy way to gain access to the "verified" checkmark.
Now, reports are surfacing regarding Twitter accounts operated by bad actors using artificial intelligence systems to generate profile images that imitate real-life people. Since the profile picture appears to be a real-life person, these accounts are given the checkmark. Unfortunately, the operators of these accounts are using their verification to push specific political agendas across the platform. For instance, one account that was found described itself as a "nationalist," "anti-liberal," and "anti-cringe".
So, how were these accounts identified? According to conspirator0, who took to Twitter with their findings, these accounts were discovered by looking for telltale signs of profile pictures being generated by a generative adversarial network (GAN) that is commonly used to make faces for things such as "This Person Does Not Exist".
Additionally, conspirator0 said that if you know what you are looking for the portrait photographs generated by AI standout as the AI may have incorrectly rendered an individuals glasses to make them wonky, placed a weird background behind the subject, distorted secondary faces if they are present in the image and more.
One study found that a dead giveaway was specific highlights in the subject's corneas. As the development of AI technologies continues, the quality of these AI-generated photos will increase, making it much harder for them to be picked out from actual photos taken by humans. The combination of AI photos of humans and nefarious bot/spam accounts on social media platforms is undoubtedly a war that is just getting started.
If you are interested in any more recent news, a team of scientists have printed edgeless human skin that features enhanced biomechanical properties. For more information on that story, check out the below link.