🚧 Work in progress 🚧

Despite all the hype, AI still feels underrated1. It will so clearly change everything beyond our current understanding, and the movement is only accelerating.

An increasingly urgent sense I am getting is that AI will increase the ”offensive” powers of individual humans and other AIs to a deafening level. In this world, defensive technologies will actually become offensive, as a way to shield and filter our vision and our perception and our world from digital junk and misinformation.

In that sense, AI has a natural pairing with CG, and I am very excited to explore this intersection in 2024 and onwards.

I think signatures are our first and most important line of defense/offense against AI. As a way to connect data to real humans or trusted institutions and filter our information to VIP sources. ZK makes a killer pairing with signatures, as we can use ZK to make complex and private computations over signatures to prove more useful predicates about data.

AI project proposals

SXG proofs & lobbying

  • This website signs its data with SXG!
  • I want to make proofs about what I say on other parts of the web!
  • And then let’s do it for your website (and Medium, Substack, Mirror, and all major media sites…once we lobby them all to also add SXG!)
  • Specifically to prove that data did not come from an AI, but from a real human author
  • We are running out of time to join the anonymity set of “real human writers” on these platforms LOL

Technical support with Morpheus implementation

  • Promising attempt at watermarking AI produced content, led by yush and Daniel Bessenov
  • Uses perceptual hashes
  • Stability AI has expressed interest in implementation & testing

Footnotes

  1. Borrowing a phrasing from my friend Nate