Raw LLM Responses

Look, AI doesn't have to be expensive, although many mega corporations certainly want it to be. A few years back when machine learning and neural networks were kinda a cool new thing we had Google pump millions and millions into AlphaZero (AI chess engine) which went on to "dominate" the old champion Stockfish 8 (well, they were actually pretty close in strength as most of their matches ended in a draw). Then soon after that we got an open-source LeelaChess that quickly got on par with AlphaZero. Then the devs behind Stockfish also decided to use a similar approach and nowadays Stockfish 17 is fully AI and both Stockfish and Leela are stronger than AlphaZero was a few years back (abandoned project so no updates) and they achieved all this while being free. Alright so, AlphaZero was always closed to the public, but even if it was a product that could be bought, who in their right mind would buy it? Ok, human pros routinely use engines to evaluate their play, and sure, ideally you'd want as strong an engine as possible, but what's the use of a very expensive option that is slightly (probably unnoticably to a human) better than cheap options? So you're right and I agree that these billion dollar AI investments are not sustainable and built on hype, but that's not because AI is inherently expensive and unsustainable, it's because the big corpos make it to be ridiculously expensive when it doesn't have to be. I imagine we might see a similar thing play out with AI that is trendy right now, like LLMs and image generators. Current ones are expensive, but who's to say there won't be a open-source option that performs similarly to the expensive option (hello, Deepseek) and when enough people internalize this, the AI bubble will burst, but AI as a technology will still continue on its path. There's so much stuff I wanna say about your video, but 1) this is already long and 2) anything I might say has probably been said already. But I'll say even as someone who's generally pro-AI, I'm glad that something like Nightshade exists. If AI devs want to make an image generator, they can feel free to make their own training data instead of taking others'. Or they can commission a team of illustrators.