Knowledge
Calling something knowledge can be a political act, even a radical one 1. — Thenjiwe Nkosi
I recently learned more about the Library of Alexandria; more than you do while scrolling through Tumblr at random 2 anyway. It turns out that it survived the famous fire! The library as an institution at least, whether you believe it was primarily a warehouse under their control that burned aside. Other libraries superseded it in size and importance, and it slowly fell into insignificance, before closing due to funding issues and lack of support.
I also recently read a short story by Karl Schroeder called Noon in the antilibrary and though I think it is a little over-pessimistic about the nadir that AI could lead us to3, it really speaks to a worry I have about the web as its developing now. In the story - spoilers! 4 - technology is developed using AI to develop the ability to trivially create entire libraries worth of fake news and media, and eventually easily spin up an entire internet's worth. This would be an apocalyptic scenario for the kind of misinformation most people are concerned about.
Misinformation on the web has always been an issue. The Library of Alexandria example is one of many, where a good story and a bunch of funny memes ended up taking the place of what is true in the minds of me and many others. To be fair, I also had a look at the Wikipedia article for the library and in 2006 Wikipedia didn't know either, so perhaps it's more a popular knowledge proliferation issue.
This is fine in (relatively!) unimportant cases like library trivia. but I often wonder what a trustworthy web would be like. I'm not sure it would be desirable to have a web where there's a single source of truth for everything as it feels very easily manipulated for ill no matter what blockchain advocates would tell you. I think the nuances of the web are a culmination of the nuances of human communication, the effect of imperfect technology and its imperfect usage, and the crushing interests of corporations duking it out.
Corroboration helps a great deal with trustworthiness, so it's very tempting to say that trust should form a part of it, but so many parts of the vastness of human knowledge lack credible sources. There are many things that people have struggled to express or that are purposely suppressed in the world, and so don't show up online. Mimi Ọnụọha's project On Missing Data Sets explores this from the angle of a list of things that are knowable but not currently collated. She highlights the important idea that there are benefits to being unknowable, invisible and illegible.
The people missing due to missing data — whether through death or simply not being heard — often have little to gain by being included. The knowledge extraction and data collection necessary often doesn't serve these people directly. Maybe a pilot program will be run in some other region, and the benefits might trickle down to their ancestors in a generation or two.
"It took us 10 years to reach 1 billion Shazams and now we deliver 1 billion song results every month!" — Shazam
To use a fun and harmless5 example, Bop Spotter, a project collecting Shazam results from random passersby, is collecting a previously invisible data set, creating previously unknowable knowledge of what songs are played at a particular location and time. I also realise as a user of Shazam I had no idea what they did with their data prior to being bought by Apple, and I didnt know what Apple does with it now! Looking it up briefly for this article shows me it was historically to do with predicting "the next big thing" musically on its face.Apple now incorporates Shazam into its charts and artist-facing analytics.
The business case for Shazam aside, being codified by a computer anonymously in this way has few risks6 beyond making companies more money. I do still think most people could easily envision less savoury uses for a similar phone recording based set up, but I'm happy to not know about them. If knowledge is power maybe knowledge of what you don't want to know is powerful too.
Last updated 1 day, 18 hours ago
This piece is part of my attempt at Alphabet Superset, a “6-month” creative challenge (I passed a year in September — with a long break! — and the creator of the challenge hasn’t finished yet either). Other posts so far: abolition, bump, boost, culture, discussion, english, formulaic, gone, home, immortality and jargon.
Feel free to reach out with burning questions, comments or suggestions!
I can't find the source for this beyond Undisciplinary Learning, an art project from Berlin, but I love it anyway.↩
Admittedly my knowledge was topped up by randomly scrolling the Wikipedia article but I don't know why I just believed the fire was the end and the tragedy.↩
Given how LLMs seem to be peaking and there don't seem to be new approaches to the problem of uncanny images and inaccurate text results than "throw in more data and more computing power", plus the fact the LLMs have a whole web's worth of data now but haven't substantially improved, I am pretty confident AGI is unlikely in my lifetime and probably not possible in any meaningful sense. We are all just normal, innocent men and I reserve the right to be wrong!↩
I often worry I overuse memes and images and things accessibility wise and maybe I'll have a whole separate post about that, but let me know if there are guidelines around this that aren't for governments and businesses...↩
I'm afraid to look up the potential harms honestly. Everything considered harmful at this point I swear.↩
I assume there's no way to cross reference Shazams and songs with a low number of Spotify plays but honestly I will never doubt people's detective powers again after a FitBit was used as evidence in a murder case.↩