Raw LLM Responses

@AsianDadEnergy I don't know if I 100% agree with that. 1. It's very hard to know if our brain need any training data as most of us adults have already absorbed so much data during our lives that none of us would be great test subjects. 2. We can test the ability to reason about new topics but then again I feel like thinking through a problem as a human feels similar to how an LLM agent would do it. Thinking step by step. For example if I hadn't heard about the "Mary had a little lamb" song/poem before then I would have approached it like: Mary - 99.9999% a female name, 99% we are talking about a human girl or a woman with no other context given. Had - assigning the right hand side noun to the left hand side noun. A - showing that a noun is approaching. Little - Adjective that describes the upcoming noun. ... - It could be an another adjective or a noun. Given that we want to finish a sentence it's a noun 99.99999%. Let's guess "Cat" (I would guess cat, dog, house, car, family, school, ... hundred other things before a "lamb" if I didn't have the cultural context). [obviously that processing takes a millisecond for my brain and i don't think word by word what i wrote but if someone asked me to explain how I did it then that's how I'd do it] _______ If I feel like I didn't have enough context then I'd either Google or ask someone. However, the likelyhood of reaching out for more information differs from person to person. Some people are more likely to admit they are wrong, some lie to others to save face, and some lie to themselves as well unconciously (what we would maybe call hallucinating in AI terms). ______ Also, I'd argue that many of the abilities we have through genetics/birth that feel innate to being a human is just a training data through thousands of generations of human lives. We who are alive today have this data in us thanks to the natural selection before us, even if we are not able to articulate our feelings or decisions. Most of us naturally avoid snakes, spiders, big animals... even if we never came across those things. Not because we are super smart but because our brain and chemistry is structured in a way that if we go pet them then we could likely die. Those models need millions of iterations to become useful... well... we did as well... for billions of years life didn't go beyond single cell organisms. We criticise AI for needing so many generations to train itself from zero but we forget that out brains didn't just appear. We are standing on hundreds of millions of years of evolution during which information for survival gets compressed into our genetic code.