[dead]
Serendipity is too important to be left to chance (1996)
21–30 of 37 posts
Re: Serendipity is too important to be left to chance (1996)
#22When i read the title i thought the idea may have been to engineer coincidence, which seems like a contradiction, but i wonder if is not impossible.
I think it's possible. One problem that I have with Remembrancer, as well as all approaches based on the same basic idea of 'show the user a list of similar items', is that this is fundamentally guaranteed to be mostly useless because the user will already remember many similar items and have deliberately omitted them. Sometimes you will forget, and then it can be quite helpful, sure, but for the most part, this retrieval strategy is inherently limited.
What you really get serendipity is from re-encountering things you don't know or have forgotten. And we know how to predict if you have forgotten something: spaced repetition and the forgetting curve! To manufacture serendipity, you simply track 'flashcards' of important things, but optimize for the opposite of the normal spaced repetition goal: you want to review only things you are predicted to have forgotten by now.
https://gwern.net/note/statistic#program-for-non-spaced-repe...
Then, as a Remembrancer agent, you make it show preferentially items that are similar but you have probably forgotten. (The Remembrancer agent can be treated as a fraction of a review, on the grounds that even just showing you the title briefly can trigger serendipity and if it didn't provoke any reaction from you, then that is evidence it is not useful and that title should be downweighted and show up less.)
Re: Serendipity is too important to be left to chance (1996)
#23I am sure some of you here will enjoy the etymology of the word serendipity.
Apparently the word was inspired by the fairy tale "The Three Princes of Serendip" [1], the latter being the Persian name of Sri Lanka. In the story, the princes would constantly find interesting things by accident.
[1] https://en.m.wikipedia.org/wiki/The_Three_Princes_of_Serendi...
Re: Serendipity is too important to be left to chance (1996)
#24I'm working on a system similar to this to satisfy my own needs but can open up to brave alpha users if requested. I've a concept of boards/themes where you can add structured information like people (to remember contacts in context), bookmarks, thoughts, tasks and even RSS feeds to stay up-to-date on a theme (e.g battery tech). Emails and calendar will be coming soon so that your personal corpus is voluminous and of…
Years ago, I submitted an admittedly half-baked version of this to YC. The concept is interesting, but the implementation options are: a) require considerable and routine manual data entry from the end-user b) suck up every bit of data from e.g. email, phone, messaging apps, banks and end up with some privacy/security disaster Anyway, there are all sorts of things you could do with that data, like trawling for novel…
Use b) with a local datastore and attached LLM and it sounds like a great solution to me.
I would 100% give this a try if someone already has something rigged up on GitHub or try building it myself if not.
Should be easy to run on a modern machine and I hear open source LLMs are constantly improving.
Re: Serendipity is too important to be left to chance (1996)
#25Re: Serendipity is too important to be left to chance (1996)
#26When i read the title i thought the idea may have been to engineer coincidence, which seems like a contradiction, but i wonder if is not impossible.
Similar to the system gwern mentioned in a sibling comment, I have implemented a "spaced inbox" system [1] that shows me old notes that I have written. The idea is to encourage me to keep adding to and refining ideas I have (but of course, the scheduling can be tweaked to only show notes that I am likely to have forgotten, as gwern suggests). Perhaps coincidentally, both my spaced inbox system and the Remembrance Agent in the OP are designed to work with Emacs (though mine does not force the user to use Emacs).
Re: Serendipity is too important to be left to chance (1996)
#27PKM [1] concepts seem to have taken root in the mid 90s at places like UCLA and MIT.
I think in a way social media is the PKM backdoor. People share near-everything about their personal lives in exchange for likes. On the flipside, you have a pretty good record of what you've done, where you've been and your feelings at the time.
I think LLMs will give new breath to PKM.
[1] https://en.wikipedia.org/wiki/Personal_knowledge_management
Re: Serendipity is too important to be left to chance (1996)
#28I'm working on a system similar to this to satisfy my own needs but can open up to brave alpha users if requested. I've a concept of boards/themes where you can add structured information like people (to remember contacts in context), bookmarks, thoughts, tasks and even RSS feeds to stay up-to-date on a theme (e.g battery tech). Emails and calendar will be coming soon so that your personal corpus is voluminous and of…
Have you tried Confluence ?
Re: Serendipity is too important to be left to chance (1996)
#29Re: Serendipity is too important to be left to chance (1996)
#30Earlier quoted context omitted.
Emacs is very awkward to use in a mobile phone sadly.
It's not as seamless as I would like, but I manage pretty well with Emacs in Termux and files synced via Syncthing.
Orgzly https://orgzly.com/
a nice way to do org mode editing on the go. Gives you notifications for reminders/deadlines and you can press on tick boxes to toggle them etc. I use it with syncthing.