jamiepinheiro
Joined 7 years ago
Karma
4
Submissions
7
Member since
Nov 20
Recent submissions
Comment · 6 months ago
Comment · 9 months ago
I was thinking of a similar idea to this, but for news/tweets/posts. As a consumer of media, I might decide to only read media with $x staked, so AI media factories need to be willing to stake that much to reach audiences, and will get penalized when they are wrong…
I imagine a hard problem is building a system to resolve these markets.
Comment · 4 years ago
TLDR
Recently, I fell down the rabbit hole of Cyclic TV Reference Paradoxes. These occur when a chain of fictional TV show references form a cycle. Each show’s reality depends on another being fictional, so a cycle of these dependencies would be a paradox.
I wanted to know if these paradoxes actually occurred and figured it would be an interesting programming exercise to try and find out. It was effectively a graph problem, where each node was a TV show, the directed edges were references between them, and any cycles were instances of this paradox. To build this graph I scraped, indexed, and searched over 40,000 subtitles across ~300 popular TV shows, finding 2000+ fictional TV references.
I built a small web app to visualize the data and perform depth-first searches to find these cycles amongst the references. 72 paradoxes were found!
Here’s a set of screencaps illustrating a paradox across 4 shows: https://imgur.com/a/4Ej6lr2
Here’s a gif of the tool in action: https://imgur.com/a/1a9WqtC
You can try it out for yourself - https://jamiepinheiro.com/cyclic_tv_reference_paradox_finder
I’ve been using things like onrender.com for hosting projects, and once the initial pain of setting up was complete, I found shared infra like logs/metrics somewhat useful. Do you imagine building these in exe.dev? Or does doing so move into “more confusing abstractions” territory?