Especially when LLM becomes ever harder to optimize + vast energy consumption + larger AI requires more outsourcing of hardware and data service, it might be time to consider smaller and strategically regulated AIs.

This policy piece is also a companion project to the AI whitepaper that I wrote in collaboration with a larger group of amazing scholars. Both texts oppose the idea of “winner-takes-all race to construct the biggest general-purpose AI model.”

Link to the PDF ver. of the full white paper - www.uottawa.ca/research-inn...

I’m on the subway and these ads are as creepy and off putting as the product itself.

Of course, so many works have already offered this argument