Looking forward to AI advances being applied to recommendations (books etc).

Programming, philosophy, history, internet, startups, crypto.

Investor: a16zcrypto.com

Author: readwriteown.com

Can’t wait for “bring your own trained reco engine” for each service I use, so I don’t have to train their silo’ed model from scratch.

A key questions will be who owns a users taste data: user (via blockchain), or silo’d service (eg Goodreads).

My sense is Tik Tok and Spotify are two of the better recommendation algorithms. It tends to work when 1) business model is aligned with user delight 2) quick consumption & feedback to algorithm.

There’s a great line in the recent BJ Novak movie Vengeance about how playlists and albums have disappeared in how people discover music. 

It is purely algo-driven. Either through Tiktok or Spotify suggesting it.

Like the AI looks at Twitter and only shows you the most "valuable" tweets, or videos, or whatever content with respect for your time. Just hard to define what "valuable" "delightful" "informative" etc. in a vacuum -- although maybe not an insurmountable challenge.

In the future, I could see there being some sort of AI system that basically recommends everything for you, but hard to say what that looks like in practice from a product standpoint. Additionally, seems difficult and maybe even impossible to define an objective function.

I've thought about this quite a bit and feel that the idea of recommendation engines goes deeper than most people realize -- TikTok being the biggest and most used one today.

AI will help every niche find its intended audience.

making /obsidian • writing stephango.com