Raw LLM Responses

you're close to what I think but not quite. It's not that take cannot make "happy accidents", or that "they can't notice things"; they can, they might. on the one hand, the space of generation that they can "make mistakes from" or "notice from" or "create from" is limited to what they've learnt as you said. they might be able to put things together even then and come up with something that seems new, as it's the combination of many humans' experience. it's in any case a coming together of cultural endevour. that's not so different from a human; people are also creating things from a pool of a coming-together of what they learnt in their culture. more importantly, LLMs are created by humans to serve humans. they are a tool. the reason humans endeavour to make art is not because it's functional or because they want the end result, as a hobbist myself, it's the very process of drawing or playing music which is fun. im not learning to play jazz because i want the perfect end result. i want to *play*, it's the act of *playing*. it's not that ai cannot play jazz better than me (certainly at this moment it can) -- or that i cannot make serendipitous accidents or come up with a fun combination of things it knows -- it's that it cannot play for me. i play for it's own sake. i draw for it's own sake. it really is as simple as that the cosnequences on the industry will depend on it's actual ability. the videos look promising but whether or not it can produce slop at least as good as human slop, then it might end up replacing slop industry. that's an economical problem more than a moral one, as i don't think it's likely governments will ban ai automation because "it's not nice" (they might slow it down if it causes unemployment to spike maybe) completely different issue in any case. fall more under the category of automation as a whole which has been a topic since the 19t century