Профиль

Yeah, I think it's inevitable at scale. Like the process is: 1) user see's something that triggers a reaction in them, 2) they react instinctively with a block (as is the culture - 'block early and block often'). Post author cannot manage the trigger reactions of the wider population.

Like, I think they are very useful in this regard, and can save a lot of time and effort, but you still have to have a deep understanding of what you're working with. In a lot of ways you're doing QA work, making sure the 'intern' didn't miss anything important, so to say.

Like, they assign importance based on repetition. And so there may be critical ideas that are salient for people who read it, but because it's only mentioned once, the LLM doesn't understand the importance. You can adjust this by asking it to focus on 'sleeper insights', but... not consistent.

I've been experimenting with AI, and I'm not sure I think this is entirely correct. How can the LLM know what is the relevant part of what it is shortening for the relevant audience? Very precise prompting may ameliorate this, but easy to make mistakes. And you wont know it when you do.

Yeah, I was concerned about this back when it was implemented. The capacity to end any possibility of interaction forever on often extremely thin bases never sat well with me. I thought a default setting in which blocks timeout would ameliorate the issue, with friction to extend/make permanent.

yeah absolutely, AI def muddies the waters, but I think if you got the settings right, it'd be more helpful than harmful getting there is obv the real challenge

The way twitter functioned during covid was incredible. I'd love to be a part of something that recreates that kind of dynamic across a whole bunch of topics, but without the weaknesses. Roads been rough, but destination remains in sight afaict.

I'm here because I see this as our best hope of creating an epistemic substrate at scale that reflects reality. Which is critically important for the type of civilisation we were aiming to be.

I saw jack announce they were thinking about it on twitter, and when I saw Jay get selected I went and read up on her background and values. I got my invite as soon as why started handing em out in matrix.

Honestly wonder just how far back the pattern goes. Maybe this is just how a certain subset of humanity operates. It is effective. Learned behaviour.

Make a labeller maybe? If you can create a reliable detection mechanism that people can subscribe to, it would be able to scale.

Love this.

Alright.

I think there's real value in this approach. No harm in starting out this way, and adjusting later if needed. Can't do that the other way around.

Been following you a long time, and held similar opinions, but haven't been on social much recently. Where can I get a summary of what's been going on?

those were quite the days, yeah

absolutely in New Zealand we started giving legal personhood to rivers and other natural features to give a legal basis for acting in their interests communitylaw.org.nz/community-la...

It's a good time to see them, they have their pups and can be quite playful around now. Looking forward to your talk at the University of Otago in a few days.

Interesting, that would suggest this could be a filtering mechanism.

Glad to see you raising this. It is the right long term decision, we just have to get there.