Very interesting read. I completely agree with the devaluation of the arts in the UK and other places. As for AI, I struggle even with the term itself. I think it's mostly science fiction. My understanding of LLM is they 'learn' by doing pattern extractions on big data, which can then be assembled generatively. But basically, there's no creativity, just a sophisticated plagiarism of such a large selection of data that it feels original.
Stability Audio uses 800,000 songs for its training, and it sounds terrible. I played around with it for about 5 minutes before getting bored with its really bad piano 'compositions'. I haven't found meaningful audio from Lyria, or interacted with it. I'd be surprised if it sounds any better. I don't have the quote to hand, but a major VC fund stated the other day that unless fair use is applied to LLMs then the industry basically bust. I see no reason why fair use should be applied, so if that makes the industry go bust, so be it. AI is not an economic life vest for Silicon Valley.
I'm sure LLMs have their uses. I doubt they'll change any fundamentals of music composition. Perhaps their most destructive quality would be for young musicians to take them seriously, and not bother to put the hard work into studying. Overall, I think for the most part it's a really stupid technology and a huge waste of computer power, that unless it can find some justification to exist should go the way of Crypto and the Metaverse.
Yes, I also am concerned that if music is made a completely unviable job option, it will become even more deprioritised in schools and musical skills will become quite fringe
Very interesting read. I completely agree with the devaluation of the arts in the UK and other places. As for AI, I struggle even with the term itself. I think it's mostly science fiction. My understanding of LLM is they 'learn' by doing pattern extractions on big data, which can then be assembled generatively. But basically, there's no creativity, just a sophisticated plagiarism of such a large selection of data that it feels original.
Stability Audio uses 800,000 songs for its training, and it sounds terrible. I played around with it for about 5 minutes before getting bored with its really bad piano 'compositions'. I haven't found meaningful audio from Lyria, or interacted with it. I'd be surprised if it sounds any better. I don't have the quote to hand, but a major VC fund stated the other day that unless fair use is applied to LLMs then the industry basically bust. I see no reason why fair use should be applied, so if that makes the industry go bust, so be it. AI is not an economic life vest for Silicon Valley.
I'm sure LLMs have their uses. I doubt they'll change any fundamentals of music composition. Perhaps their most destructive quality would be for young musicians to take them seriously, and not bother to put the hard work into studying. Overall, I think for the most part it's a really stupid technology and a huge waste of computer power, that unless it can find some justification to exist should go the way of Crypto and the Metaverse.
Yes, I also am concerned that if music is made a completely unviable job option, it will become even more deprioritised in schools and musical skills will become quite fringe