The desire to humble and infantilise African writers is clear.
Certain people can't cope with the idea that the African continent is full of literary talent that isn't tied to Western validation.
This ai frenzy is a way for those people to assert authority or control.
@TuhinChakr you have been on twitter since 2018. before the vauhini vara essay on the caine prize, there’s no mention of african literature on your timeline. how come suddenly you’re an expert on it? can i see your publications on african literature pre-2026?
So, my ABSA account is now working. However, for my birthday on the 8th, there was no coffee and cake. Sort of drove the knife in that Stanchart is really gone. Anyway, let's see what Absa has to offer.
Reading used to be a surefire way to learn the contours and contents of another person's mind. That's gone now that AI has ended the human monopoly on the written word.
To me this loss cannot be compensated by any benefits that may come.
Pangram works quite well for identification as you approach 0% or 100% AI text, but most writing these days falls somewhere in between and detection seems to tend towards overstating AI contamination for for fields where writing conventions have been largely replicated by AI
Pangram cannot tell the difference between AI-generated and AI-edited text. It has no clue where the actual ideas come from. Yet people weaponize that "100%" label to shame writers and wreck careers, hitting vulnerable people the hardest.
It is the exact opposite of everything we’ve spent years teaching about ethical AI deployment.
The irony is that Pangram's founder is drinking so much of his own Kool-Aid that he gets mad when committees rightfully ignore his meaningless label.
This is not true of Pangram. It has been demonstrated to have very low rate of false positive. The statistical signature it relies on is different than the ones that you’ve intuited. Many studies have been done on this - Ive done my own experiments with it too - its quite good. Shame it gets lumped in with detectors with flawed methodologies.