What are the implications on on device models and how does that change the economics of AI?

What are the implications on on device models and how does that change the economics of AI? https://9to5google.com/2024/03/07/pixel-8-gemini-nano/

Writer. Building @pinatacloud. Working on simple static site hosting https://orbiter.host \ https://polluterofminds.com

I wonder if mobile models make for better compression. in the same vein, I wonder what the minimum requirements are for critical feedback to a world model.

this may not just be inference cost trending to zero, but also valuable crowdsourcing of incremental gains (that can be incentivized).

𝖒𝖆𝖓𝖉𝖆𝖙𝖚𝖒 𝖆𝖉 𝖆𝖘𝖙𝖗𝖆

This is interesting. It’s part of why I’m bullish on on-device LLMs