After losing this post a few times due to problems with commenting on the under construction social media web site WikiTribune 2 / Truth Café, I decided to type it locally. I might as well put it on my web site too as a backup and so that search engines can find it. It's in reply to a post by Horatio Mortimer about a proposal (PDF here) by Monica Anderson (of Syntience Inc.) titled Bubble City Design Proposal / A Twitter Alternative Which is not a Social Medium / It is a Real Time Idea Router, October 31 2022. First, a few short form notes to figure out what it is.
This is a social media system that feeds posts through multi-category sorting and then users subscribe various categories (bubbles). That's been done before with Bayesian statistics; spam filtering is a subset, with just two categories. But here she's waving the AI wand in hopes of better quality categorization (I suspect it will be). So you're basically using the categorization to amplify some editorial decisions, so it's a tool for editors?
Interesting more-like-this action idea. Can rate a whole message or just a word or a semantic phrase (AI defined). Maybe that is the custom filtering? Yup! Also has a fewer-like-this button. And it chains the filter levels together, with the per-user custom one at the end, and the general category and safety filters at the start. And restricts new message flow with "Pacer" to just show top or random messages if too many (and shows them scrolling by in real time - for that passive TV viewing experience). There is a playlist of bubbles to be displayed. All posts are public and used for AI training.
It thinks it's common carrier, but with editors under the hood (someone trained that AI for each category), maybe not! Years of legal fun!
Pay to post. But unlimited if you have the money (my #RatingStone has a weekly maximum to make things fairer), nicknames tracked to a credit card. Readers anonymous, free, just maintains reading state for a few months of inactivity. Claims spammer's duplicate posts are detected, but what if they are AI random generated? At least they'll be paid posts. Perhaps you could filter by how much was paid. #RatingStone goes a bit further and also lets you pass a few points to the reader, if you really want them to examine something.
Too many obvious patents. The USA patent office should go back to requiring a working model for each patent, so people don't blue sky ideas they haven't even tried out. But that's another rant.
So, Monica Anderson's system is actually quite interesting. Will be fun to see if it works out in practice, if she or Google gets around to making a prototype. I'd recommend not just using posts and news-wires as feeds, but include the whole Web (I should patent that!). After all, it's kind of a search/filtering engine but continuously running.
Copyright © 2023 by Alexander G. M. Smith.