By Cory Doctorow (GPG 0xBF3D9110957E5F4C) @doctorow.pluralistic.net · 02/06/2026Contrast this with AI: when you prompt an AI to generate words or pixels, you are conveying some intent about the feeling you want the people who experience the model's output to experience. 9/ 1232
By Cory Doctorow (GPG 0xBF3D9110957E5F4C) @doctorow.pluralistic.net · 02/06/2026The problem is that the AI doesn't have any intent of its own - it just has statistical *predictions*, based on other people's intent, which it has analyzed through its training data. 10/ 2414
By Cory Doctorow (GPG 0xBF3D9110957E5F4C) @doctorow.pluralistic.net · 02/06/2026So when the AI expands the three sentences in your prompt into 100,000 words or 1,000,000 pixels, it isn't *adding* any of its intention to the finished work, it's *diluting* the intention you fed to it. 11/ 2649
Philipp Angermeyer @pangermeyer.bsky.social · 02/06/2026This is a dimension of language that LLM proponents don’t understand. They think linguistic communication works like a math equation, when meaning is in fact derived from the interaction between speaker and addressee 000
