The case for a “strategic fact reserve” is actually pretty interesting and compelling.

summary

what if AI gets economically necessary, like GPS, but then we find out we don't have any uncontaminated training data to build new ones?

New blog post! The need for a strategic fact reserve

https://t.co/q4BPl0qRqH

Matt Webb 🌸🌼🌸

The case for a “strategic fact reserve” is actually pretty interesting and compelling.

https://x.com/genmon/status/1881352208945643798

Rated 💯 by multiple networks. Fluent in /unglish. Unbothered by /whatday of the week it is.

Allegedly a designer.

weird internet kid. helping Creators with /seemore. yapping on /okbanger. curating in /cameron + pdfproject.org | dev + veteran + HBS. 
seemore.tv/cameron 🫡

Well, that’s one dystopian way to view it. On the other hand, strategic fact reserve literally solves this.

It’s a “who polices the police” issue all the way down unfortunately

Well, I think self-verifying data structures address most of these concerns without requiring trust in a central authority.

“Fact” is defined as “uncorrupted source of information” in the blog post, rather than “objective truth about reality.”

So this is more like a block list or allow list for certain public keys during ingestion of training data so you can confidently bias the model towards your perspective.