All-bones Jones on Nostr: kinda wish i knew more about the language models LLMs use and how to train them and ...
kinda wish i knew more about the language models LLMs use and how to train them and change their output. they just use probability to find the next word, so couldn't you easily do the same thing for songs? music can be divided up into key phrases with a ton of repeated patterns, it feels like it wouldn't be hard at all to get a midi output for a song
Published at
2024-02-25 21:49:19Event JSON
{
"id": "f5a5229f04fb62288eec2d9615e090fe84c837bfe87aee90c7778f1ee1a4b722",
"pubkey": "d92de9e6281ce4730322a3c2f961f042430119c0c85c42b2de7a9b97f2107334",
"created_at": 1708897759,
"kind": 1,
"tags": [
[
"proxy",
"https://bae.st/objects/8cc2988f-92cc-4e8a-9003-ab207971288a",
"activitypub"
]
],
"content": "kinda wish i knew more about the language models LLMs use and how to train them and change their output. they just use probability to find the next word, so couldn't you easily do the same thing for songs? music can be divided up into key phrases with a ton of repeated patterns, it feels like it wouldn't be hard at all to get a midi output for a song",
"sig": "708a963cf3725b3f4e0281c7944441ea8aa30ae2ad6388a2a256e5dd80db2f36312b46ea987acb6cafbf013fff278d548dc8110ec682cb95e52af5ba9f8789f2"
}