Andries du Toit - School of Government, UWC

@andriesdutoit-uwc.bsky.social

Postcolonial statecraft; contentious politics; political belonging, deliberative practice. All this, and the polycrisis. Views not my own.

None of this is to deny that the technology has some uses in these capacities (though not ones I've felt any need for myself). But I worry that we are starting to see intellectual labour as just a means to an end, and as something that can be shortcut.

And as for writing, don't get me started. Most often, the process of writing is not a laborious business of transcription, but a process of thinking itself. I'm not about to outsource my thinking to a machine.

...I mean that even with careful prompts I would not be confident LLMs could perform the function I'd want. How do I know that any bland AI summary of a paper is going to extract what might be of value to *me* in it? I don't even know that myself before I read it...

I'm often surprised at how much researchers are using LLMs for intellectual work. It's not that I disapprove, or at least not on principle. It's more that I'm surprised at the trust being placed in the process. And by that I don't just mean LLMs can give wrong information...

My book, A Blow to the Head, in which I reflect on my own experience of the violence of white supremacy and the complexities of repair, is on the Longlist for the #Canex Prize for publishing in Africa. What a surprise, and what an amazing honour!

Writing Africa@writingafrica.bsky.social · last yr.

CANEX Prize for Publishing in Africa 2025 longlist announced #LiteraryAwards #Fiction #Nonfiction #Publishers www.writingafrica.com/canex-prize-...

Chatbots — LLMs — do not know facts and are not designed to be able to accurately answer factual questions. They are designed to find and mimic patterns of words, probabilistically. When they’re “right” it’s because correct things are often written down, so those patterns are frequent. That’s all.