in the near future you’ll be able to, if you want to, plug GPT-4 into bluesky and describe in natural language the types of posts you want to see and how often - ThreadSky

dystopiabreaker.xyz • 710 days ago

in the near future you’ll be able to, if you want to, plug GPT-4 into bluesky and describe in natural language the types of posts you want to see and how often

Comments

socallmemaeb.bsky.social•710 days ago

youre so lucky theres no block function yet

galumay.com.au•710 days ago

Honestly, that would be enough to make me delete Bluesky.

dystopiabreaker.xyz•710 days ago

?

other people having the option to use GPT-4 to filter the posts is already the case

galumay.com.au•710 days ago

Scary shit.

dystopiabreaker.xyz•710 days ago

in fact, it has already been built: https://bsky.app/profile/jem.fm/post/3jugbg3sv432m

gls.bsky.social•710 days ago

I’d be curious to see live up to date embeddings offered by someone. If you got that you could plug in an open source model on device to generate longer search queries on the vector database.

dystopiabreaker.xyz•710 days ago

the scaling of this also might be fine in the long run since there will likely be coincidences of wants (people sharing feed preferences can deduplicate GPT time)

p-e.bsky.social•710 days ago

another optimization is have GPT translate to a DSL as a top-level filter and then use GPT as a final check for smaller % of posts

myco.bsky.social•710 days ago

So happy to see you here

gojomo.bsky.social•710 days ago

strong agree. here, let me test linking to a 🐦 thread on related ideas: https://twitter.com/gojomo/status/1595912687623761920

gojomo.bsky.social•710 days ago

tho under user control, LLM-powered filtering utopian, under others' control could be the most onerous 'netnanny' ever: https://twitter.com/gojomo/status/1582446736693334016

(eg, can you phrase your posts & texts with enough 'river crab' slang to sneak it by "the great LLM firewall of China"?)

Comments

Posting Rules

Reply