It's all so complicated, I don't know what to think

Feels that way

I think about model collapse daily. It actually gives me great hope. A reductive view of intelligence would mean it's possible to apply model collapse to human thought too. I think there is something essentially different personally, but I'd be interested to hear your thoights on this episode: https://podcasts.apple.com/us/podcast/read-928-urban-bugmen-and-ai-model-collapse-a-unified-theory/id1359544516?i=1000746133548

I think it's a perfect application, and the more important question, really. I read Kingsnorth recently, and he puts a nice bow on a lot of reading I've been doing.

For anyone interested, has been going through the recent Epstein dump and highlighting interesting documents. Sadly, much of the contents seem to be basically hearsay, but there is some seriously messed up stuff in there.

What

Maybe, I bet there are ways to get better guarantees via zk proofs or something

I think advertising ia bad, but it will happen anyway, but nostr's architecture limits how bad it can be, and those of us who are principled can still find business models that work, especially with bitcoin in the mix. Maybe I will write about this, I have some ideas.

There are a few I think. I know divine has been looking into scaling their relay

I do think pain can help. It's a chicken and egg problem.

It would have to be a server run by a client with cryptographic keys for sending notifications to the app. For implementation, see https://github.com/coracle-social/nps https://github.com/coracle-social/npb

why not both

This is hardcore, I aspire to this level of personal study

Epistemology is the other big one, and it's the road to the others

It's hard to talk about AI. It's just such a massive, multi-faceted topic

A rant for your reading pleasure

They have chosen which trees to cut down

My anti-AI takes aren't popular, weird

Being anti-technology is like being anti-tree. Technology is non-negotiable, it's just a fact. But individual technologies, like forests, can change over time. And individual applications of a technology, like trees, which are particular, contextual, and concrete, can be modified to suit the user and his environment.

LLM-assisted coding can be used to good effect, but I think it's pretty clear at this point that its impact has so far been a net negative on productivity in real terms.