Under new CEO Elon Musk, Twitter has decided to keep its paid “Verified” program through a Twitter Blue subscription, despite extensive scrutiny since Musk took over in November.
While the platform now includes different colored check marks for business and government, the blue check mark – which once indicated that a user’s identity had been authenticated – remains open access and is now exploited by parties. infamous armies of artificial intelligence technologies. .
As Point by Twitter user conspirator0, a bunch of “verified” Twitter accounts are sporting AI-generated faces as profile pictures while pretending to be real people.
And many of them, according to conspirator0’s results“pushing specific political agendas”, both left and right – albeit primarily the latter.
One Account under the now dangling handle of cortez_santiage have described themselves as a “nationalist”, “paleo-conservative”, “anti-liberal” and “anti-cringe”. Another one found by conspirator0formerly under the username Kenoisseur, campaigned to share the so-called evidence of “white genocide in America”.
Others may be more innocuous, such as pretending to be a Harvard-educated epidemiologist.
It’s unclear how many of them are mere robots or anonymous, treacherous humans trying to maintain a more believable face – but our best guess is that it’s a mix of the two.
“Allowing accounts with fake faces to be ‘verified’ without even requiring operators to disclose that the ‘face’ is artificially generated is a patently pro-deception stance,” conspirator0 wrote in a post. Tweeter.
Numerous accounts, dating back to November – right after Musk’s takeover – were eventually suspended. But conspirator0 has since unearthed more verified accounts with AI faces that have not been suspended or stripped of their “verified” status – using only a quick and simple search on Twitter by entering “filter: blue_verified” and pasting over common English words.
Then, to root out suspicious accounts, conspirator0 looked for the tell-tale signs of synthesized faces using a Generative Adversarial Network (GAN), which they claim is used in popular tools like This Person Doesn’t Exist.
The most prominent and distinctive feature of unmodified GAN-generated faces is the still placement of the eyes. If you overlay several faces generated by GAN, it becomes clear that the portraits were not naturally taken and cropped. In other words, the eyes almost never deviate.
Other indicators include wonky glasses, absurd clothing, and deformed secondary faces in the framework of. At least one study has identified inconsistent specular highlights in eye corneas as the ultimate giveaway.
But in all likelihood, this is just the tip of the iceberg. These freebies only apply to unmodified GAN generated faces. If someone made the effort to adjust them manually, even with a simple touch, they might be even harder to detect.
It’s a particularly concerning trend since “Legacy” verified accounts – accounts that were verified under the old program that required users to corroborate their identity – still retain the same blue check mark.
This risks making those who only briefly look at these profiles mistake them for real people, whether they have paid for Twitter Blue or have been verified.
Conspirator0 also cites a study published last year in the journal Proceedings of the National Academy of Sciences who discovered that AI face generators “have crossed the strange valley and are able to create faces that are indistinguishable – and more trustworthy – than real faces”.
Anecdotally, of course, you can find many sloppy likenesses which indicate otherwise, but the most convincing of them will likely go largely unnoticed, and the mere fact that these AI faces look more like professional portraits than terrible selfies also lends them an undue sense of credibility.
In short, what we are witnessing is a confluence of the emerging and widespread popularity of AI, its propensity to be abused to spread disinformation, and Musk’s decision to let just about anyone wield a status sign badge.
Still, it’s true that badges are arguably the least worrisome aspect of this developing trend. Once (or if) widely available AIs are competent enough, they won’t need trivial digital badges to feign credibility – or maintain a facade of humanity.
Learn more about AI: Shameless real estate agents are already preparing property listings with ChatGPT