first DeepSeek, now a new set of models from ByteDance. hopes the US had a moat it could maintain around AI dominance seem dashed, despite the extraordinary scale (and ecological weight) of US investment and attempts to retain advantage in GPUs and advanced microchips. https://www.aibase.com/news/14931 1/

these Chinese models are much, much lighter and cheaper to run than US frontier models.

will we really need to invest half a trillion dollars in AI and build special power plants (maybe due to Jevon's paradox), or might we just enjoy greener, cheaper applications? https://en.wikipedia.org/wiki/Jevons_paradox /fin