Stephen Downes

"The fight over training data is not only about what gets taken from the commons," writes Ranjit Singh."It is also about what the system returns to us as the baseline version of culture." A similar point is also made in this paper by John Moravec on AI coloniality. I think this is true if we think of AI - or even a particular AI system - as a single undifferentiated whole. But it never is, and it's important to understand this. AI always works in a context. Sometimes it's a vague and ill-defined context, such as the prompt, "Write a short story about a significant holiday." Even here, a lot of work is being done by each word. Ranjit Singh focuses on the word 'significant', but I focus equally on the word 'holiday', which has the context 'American!' written all over it. Nobody does holidays quite like the Americans. I tried it here (in Canada) and got a story about New Years Day. I asked for a story about a significant writer and got Mary Shelly. For a significant deity, I got Athena (I was hoping for Ganesh, but you can't always have what you want). For a significant city, I got Istanbul. Here's my transcript. None of this disproves the original statements on coloniality, but I think it requires of them a lot more nuance.