One thing that's been irking me about the adoption of AI is the homogenisation it brings. Not only in sentence structure, like the tell-tale 'It's not X, but Y', but in the loss of local phrases and culture. For instance, a colleague highlighted someone's work, but their message was clearly 'AI'ed and it felt slightly false and unnatural (while their English is excellent, it isn't so clearly US English). I found it quite sad that they trusted the AI more than themselves and felt the AI rewording gave the best chance of 'landing' the sentiment in our global team.

More than that loss of 'local colour' (we're all going to speak American eventually anyway), it got me thinking about the plurality of thought, and how our thought is governed by our words. What if there's a good reason Italian is structured the way it is, or that German has a perfect word for something that we simply don't have in English? We aren't just losing local dialects or 'flavour' but whole ways of perceiving and interpreting the world, through the adoption of AI trained on a subset of a subset of the Anglospheric world. Even if you're using ChatGPT in French, the LLM is trained on English language data, and thus the resultant French will be in an English (well, US English) style.

I know many on the Fediverse take a dim view on LLMs and GenAI anyway, but this isn't 'just' a technology or jobs issue, we could be losing far more of our diversity this way.

#genAI #genaiuse #llm_ai #Culture
Hi AoIRistas, after having to sit @AoIR conferences out for many years (😥), I am reeeaaallllllly planning on attending #AoIR2026 🤗🤞 Papers are already in preparation, but I'm also happy to plan/join roundtables, fishbowls, and pre-conferences. Feel free to reach out, notably re:
#STS #GenAIuse #AIgovernance #AIpolicy #criticalAIstudies