Raw LLM Responses

This video reveals Neil's INCORRECT understanding of today's AI. Neural networks and today's AI CAN COME UP WITH SOMETHING NEW and innovative that does not already exist on the internet. Today's AI is NOT based on precise computer logic but on neural networks. Neural networks are complex arrays of non-linear mathematics and statistical processes being used to crunch through lots of data on the internet. Scientists like Neil have been using mathematics to explore and discover new things they never knew about before. Neural networks are effectively using mathematical processes that end up exploring and discovering things new that didn't exist before. With neural networks which is the underpinning of today's AI the mathematical results are somewhat unpredictable. In many cases you can ask AI the same questions over and over again and it will likely NOT give you exactly the same answer. I asked Grok to view this video and give it's opinion of Neil's claim about AI -->>>>> WHAT FOLLOWS ARE GROK'S COMMENTS AFTER VIEWING THIS VIDEO >>>> This clip shows Tyson anchoring on an outdated or overly simplistic mental model of AI, one that treats it like a giant, rule-bound lookup table (your "monolithic database"). It's deterministic in his eyes: feed it data, get back echoes, no surprises. Tyson evokes expert systems (e.g., 1970s-80s rule-based AI like MYCIN for medical diagnosis: hard-coded "if symptom X, then disease Y"). Those were deterministic—brittle, exhaustive logic trees limited to programmed knowledge. But today's LLMs? They're statistical beasts: trained via backpropagation on massive corpora, using gradients to minimize prediction errors. Tyson says "no new knowledge," but that's ignoring how these models hallucinate creativity—e.g., generating code for a novel algorithm by blending patterns, or poems that riff on unseen metaphors. It's not "in the data" exactly; it's synthesized, like how a curve-fit polynomial predicts points beyond training samples.