LLMs are effectively just functions. The prompt is an argument. You could build a stack frame from the prompt, LLM model, an HTTP endpoint as the return address, and a distributed vector DB for the heap.
With this in place you need a runtime. You could probably build the runtime on Nostr.
A system like this could be ran by millions of people in a decentralized way and although would have higher latencies than centralized models like ChatGPT it might be more powerful.
Prompts could be paid for with #bitcoin via #lightning or #ecash. Miners can supplement their revenue by providing access to high performance compute and still earn bitcoin.
#ai #mining
MachuPikacchu / Machu Pikacchu
npub1r6…4gmmd
2024-06-23 12:00:58
Author Public Key
npub1r6ggl0qazvwp02rlxgrf75lkfazuwhu35tmdg0u25eqsjax6243qh4gmmdPublished at
2024-06-23 12:00:58Event JSON
{
"id": "b5ed2bd266a526f0f7c2182ffa8351f598f5a74331619f3c88ba8ad09d9e80de",
"pubkey": "1e908fbc1d131c17a87f32069f53f64f45c75f91a2f6d43f8aa6410974da5562",
"created_at": 1719144058,
"kind": 1,
"tags": [
[
"t",
"bitcoin"
],
[
"t",
"lightning"
],
[
"t",
"ecash"
],
[
"t",
"ai"
],
[
"t",
"mining"
]
],
"content": "LLMs are effectively just functions. The prompt is an argument. You could build a stack frame from the prompt, LLM model, an HTTP endpoint as the return address, and a distributed vector DB for the heap.\n\nWith this in place you need a runtime. You could probably build the runtime on Nostr.\n\nA system like this could be ran by millions of people in a decentralized way and although would have higher latencies than centralized models like ChatGPT it might be more powerful.\n\nPrompts could be paid for with #bitcoin via #lightning or #ecash. Miners can supplement their revenue by providing access to high performance compute and still earn bitcoin. \n\n#ai #mining",
"sig": "26bb7071e8736cbc20762e1d714a28f35acb7fa093a16602ee832f4297429a983d509ea31200eb0dda62678d16e3290e884513ea5af4105300346325798915aa"
}