Boosted by pixxl:
jonny@neuromatch.social ("jonny (nonvenomous)") wrote:
@varx
That's the entire pattern of LLM code that have never changed and only accelerated with model version. It progressively decays into dogshit because there is no actual structure anywhere, it merely imitates the form of structure, the same way it imitates the form of code, and the form of knowing. There has been no breakthrough in handling complex systems with LLMs. The companies that can afford infinite compute simply spend infinite compute on every increasing likes of nonsense.