Oddbean new post about login | logout And you don't even need to read the full book - find the content that stands out and then you see how it fits with the rest of the book
Yes, you can find a related part by comparing with other parts and then click through to the whole. That's more efficient than trying to compare two wholes.
indeed/ legal libraries etc. so many poss.
Yeah, case law searches. International and cross-language.
awesome concept
"Show me all sections where 'Ghosts' in Shakespeare's collected works are mentioned and cross reference those with any mention of 'Ghosts' in his contemporaries collected works that also mention the word 'blood'." Same type of request made about molecular motors like Kinesin in, oh IDK, all published molecular biology papers published since 1980 could be made. That type of granularity would cause explosions in academic/private R&D thought.
And not only that, but its not just one specific kind. You can attach embeddings and labels to not just wiki or note styled content, but any others. Then, once you have a nice collection of ideas, you can group them together in a kind 30040. Any related ideas to written papers or books that has been written on nostr? what about for nostr development jtself?
Sounds like the embeddings themselves carry the seeds of any future classification system anyone could want and sprout depending on what they want without actually asking for a 'legacy' classification system. It's dynamic, not static. How far off am I?
🎯 personalized classification systems competing generalized classification systems niche classification systems Put all the sources, broken down into parts, on Nostr and then just start classifying everything. Then people can choose the classification that they want to use, to navigate the way they want to navigate. decentralized classification organic association patterns
This work is nothing new, its being done to analyze existing systems, to maximize money and for themselves. We need it on nostr - so anyone can navjgate and learn the most high signal ideas put out by anyone. The system itself is living, evolving with specifications and people converse and argue about changes to the spec here anyway. Rather than navigate it by thread or anything hard coded just meander and find what is most meaningful. Nothing is against hard coded links, but who knows what connections are there even if they aren't explicitly connected.
Dewey, Peirce, and the rest of the Pragmatists might roll over in their grave. Honestly, I think they meant well with classification systems but I think those systems have held human thought back for far longer than was good. There was no way for "Emergent Classification" to exist under the Dewey Decimal or Library of Congress Classification systems. Hell, I see this being applied to things like lichen and fungus classification because those guys have been scratching their heads for well over a century trying to figure out what's related to what.
Yup. This concept stems originally from biology journal analysis.
I know because I was there. I am THAT OLD fr. 😂
Did my undergrad in cell and molecular biology. Lots of journal reading here.
THAT EXPLAINS IT. 😂 We were amazed that anyone else understood anything we were saying because until now, it's been us talking excitedly amongst ourselves in public, while everyone is like https://c.tenor.com/90dFwxuNWpMAAAAC/tenor.gif
I tested a software project decades ago, that allowed you to navigate scientific journals using word bubbles.
Any efficacy at all?
Yes, it was actually really cool. You could type in the name of a protein molecule, for instance and see a cluster of articles containing that name, with articles closer to the center having the most references and the bubble size indicating how often it contained that word. Or you could type in nothing and just brows around the entire collection, by clicking on words. You didn't have to even know what you were looking for, when you "entered the library". You could just look around and see if anything was interesting. But I left the company and I thought the idea had been abandoned. I only saw visual data navigation after that with Kibana and Graphene, but the visualisation wasn't as emersive. And then I saw some stuff on graphical databases. And then Liminal showed up here and was like, listen yo, I got this idea from biology... Turns out, the idea is still alive and it's more advanced now. And we're not only going to use it for biology. 😁
Historical data would be another fascinating thing. like, I was listening to a podcast, yesterday, from a German scientist who moves to DC and analyzed movements in interest rates and the White House visitors book. There's a correlation and he found the causation: Fed Chairman being ordered to appear in the Presidents' office.
Yes, we were already working on stuff like this in the 90s. But Nostr plebicizes it and actually humanizes it, again, as one Nostrich in Helsinki or Nairobi with a keen interest in Topic XYZ can do these categorizations about whatever obscure topic they are interested in. That's why we're so excited. We want _everyone_ to be able to be a librarian. Everyone gets to assemble and categorize their own library or choose to use Npubbunchofnumbers' library.
That's why I now refuse to perform categorizations for other people. I am giving everyone the means to categorize things for themselves. Anytime you run into something that looks like a Nostr spec, you can just add a "Nostr spec" label to it and move on with your life. Anyone who searches the labels for "Nostr spec" will see everything you tagged. Or you could use the labels from yourself and your WoT to auto-update a Nostr-based register of Nostr specs. In fact, let me go build that. 😂
💯 Another thing we're making possible is making the source data itself more granular, by breaking it into pieces and stringing the pieces back together in order. Same as was done with DNA in the human genome project. Split it up, create an index of the parts, analyze and label each part in parallel, string it back together according to the index. Label the index according to the labels of the parts. That's called "high throughput analysis". DNA and events are both just character sets. 🤷♀️
