Anthropy :verified_dragon: on Nostr: I'm thinking, people keep saying ML is so inefficient for information retrieval and ...
I'm thinking, people keep saying ML is so inefficient for information retrieval and all that.. but I actually think it would be hypothetically feasible to create a search transformer/RAG, that is MORE efficient at its task of simply returning relevant links for the user (especially considering the size of the internet nowadays), than a classical index based search engine. That may seem like a challenge given the extensive optimization of the latter, but I think an ANN could do it even better 🤔
Published at
2024-05-29 20:58:33Event JSON
{
"id": "b9d09273f84e11df6a2c29e2785d91722e8d80272916b7f2e8c1b977a4eb2252",
"pubkey": "e6da4b3c3bae782bfec1f845f8e41da3a15b43920ba660b8e195b43a8ee136a2",
"created_at": 1717016313,
"kind": 1,
"tags": [
[
"content-warning",
"ML"
],
[
"proxy",
"https://mastodon.derg.nz/@anthropy/112526381094416749",
"web"
],
[
"proxy",
"https://mastodon.derg.nz/users/anthropy/statuses/112526381094416749",
"activitypub"
],
[
"L",
"pink.momostr"
],
[
"l",
"pink.momostr.activitypub:https://mastodon.derg.nz/users/anthropy/statuses/112526381094416749",
"pink.momostr"
]
],
"content": "I'm thinking, people keep saying ML is so inefficient for information retrieval and all that.. but I actually think it would be hypothetically feasible to create a search transformer/RAG, that is MORE efficient at its task of simply returning relevant links for the user (especially considering the size of the internet nowadays), than a classical index based search engine. That may seem like a challenge given the extensive optimization of the latter, but I think an ANN could do it even better 🤔",
"sig": "448aed760dc449172b296268d6390fcbd358a68a4f189a856569459482b7b50fdfe57ed3c3b97c3fed2134a5796743424789662c81f1b30834f7717f7a8fc4b3"
}