Text
the case of the ai classmate
One of the first things that happened after I finally added “MIT ‘27” to my bio was the rush of follow requests from fellow ‘27s on instagram. Too lazy to go through and check them all, I spam accepted and followed everyone back, but in july I found a strange one. A girl named Yuna Yang whos pictures gave me a serious case of Uncanny Valley. *insert pictures here*. Confused, I took to my old robotics team’s discord in search for answers.
Still confused: I went looking in a place where I knew I would find someone who would probably know way more than I expected on the subject: #technology on the MIT ‘27 server (Because who else would want to spend a wednesday night sussing out an insta account)
We (he) then got to analyzing what we could, and found some points of evidence:
Messed up hair
People in the background seemingly missing legs
Things just look too smooth in general
Fuzz, could be from diffusion
Kyle had already stalked the account with an alternate instagram frontend (of course he knows one) and found hidden comments among other things. Now, why were we doing this? Boredom? Protecting others from a potential scam? Who knows, but we were all in at this point.
At this point we created an “AI busters” group chat and dragged in kiera who had commented on the first post:
I also talked to Diego, who pointed out that the facial features across multiple posts seemed to be slightly off, and that no such person existed on the directory, slack, or discord. Clearly, we were dealing with a fake person at this point. Diego also pointed out that the hashtags on the posts were in like 6 different languages. I also found a more powerful ai detection tool that suggested that all of the images on their instagram were 99.5%+ likely to be ai generated. Even a damn picture of grass was AI generated.
(A benchmark picture of me from orientation challenges)
After joking about administering a turing test, I went in to interrogate.
As expected, the account denied everything, even claiming that they use a fake name on social media, but stopped responding once I asked for a kerb. The kerb would have been an end all be all for the situation: if she’s truly a ‘27, she’ll know what it is and send it, proving her existence. If she isn’t, then she wouldnt know what a kerb is, giving herself away. Perfect.
In the end, the account refused to give a kerb to “not compromise their identity” (after a few hour gap), even though they claim to have posted real images of themselves. In the words of Kiera, “basic OPSEC L” (i had to google what that meant)
1 note
·
View note