Theme 1 – LLMs are pure statistical next‑token predictors, not evidences of understanding
- “this is just creative writing rehashing one side of the hard problem… restating the dogma that Turing wrote his most famous paper to debunk.” – bbor
- “guessing words is the entirety of what an LLM is engineered to do.” – jimbokun
- “What else could it be doing? That is literally the mechanism of how a model works.” – vhantz
Theme 2 – LLM output lacks the intentionality, uniqueness, or “soul” of human‑created work; it is merely an averaged blend of many minds
- “How is a fictional character the product of someone's mind, but a character generated from a massive database of words from other people's minds is not?” – delichon
- “The product of an individual's mind is not the same as the mathematical average of the products of everyone's minds. The latter is obviously going to lack any of the uniqueness of the former.” – bakugo
- “The AI generated character is a weighted average of many characters written by humans… cannot have any direct bearing on reality.” – hermitShell
Theme 3 – When used as a tool steered by knowledgeable humans (or wrapped in systems), LLMs excel at logical/correct tasks but fall short on insight, style, and deeper understanding
- “The more you know about a subject the better you can prompt AI… a 'mech suit for your brain' is the best analogy I've heard.” – api
- “LLM output is literally randomly sampled… we can expect the whole system… to be able to cite its sources and go digging… the LLM‑system can become as accurate as our best sources.” – eru
- “LLMs are surprisingly good at logic but roughly about as good as expected on information accuracy… they are very impressive on difficult arbitrary logic (like coding).” – howunfortunate