In 1988, Donna Haraway proposed the concept of "situated knowledges" grounded in individual and community perspectives, as opposed to knowledge grounded in "a conquering gaze from nowhere". Gary Marcus has criticized Large Language Models and what Emily M. Bender and Alex Hanna call their "synthetic text-extruding machines" (STEMs) for lacking a "world model" that would allow such machines to understand what they say. Bringing these ideas together, I wonder if the further issue is that whatever those STEMs might "know" will always be hampered and even made impossible by their lack of a specific perspective to understand the world from. That is, "artificial intelligence" would always suffer from being "unsituated knowledge". (Andrew Shields, #111words, 3 February 2026)