Raw LLM Responses

It is actually funny how little people know basically nothing about current generative AIs and the technology behind it. Basically all of these are LLM/BERT based (look it up in the Wikipedia) and this is pure statistical math. Statistical math can not think. It is also funny how so many people fall for "argumentum ad verecundiam" (argument from authority, look it up in the Wikipedia), we actually make fun of these people ("eminence over evidence"). And I'm always surprised how people do not recognize, that these LLM/BERT based AIs never reach 100% in standardized tests (!!!!!), where the solutions are known (!!!!) and actually part of the training data (!!!!!!!). Remember how in the past graphics drivers recognized famous games like Quake3 or benchmark tools like 3dmark? And the results changed drastically by just renaming the games and benchmark binaries? It is the same shit again - and I'm one of the developers who work on llama.cpp and ollama. That are nice and helpful tools, but this has nothing to do with "intelligence", that are just good and complex heuristics. Or to cite (kinda) Weizenbaum (one of the fathers of AI, the creator of ELIZA): "People are stupid, they believe every nonsense" after he tested ELIZA on university students. I mean just look at how OpenAI declared what AGI is. They basically said, it is AGI if it produces a revenue of 100 billion US dollars... WTF?!? Are you nuts? And that should tell you everything about these AI-bros you need to know.