Post

I am largerly anti-AI, in large part because I spent 4+ years as a researcher in deep learning. Neural methods are very capable and in 2018, I would have imagined todays LLMs would be at least 40 years off. The incentives, the power, and who holds it are the big concerns here.

A position of "LLMs can't do anything" will leave you unable to see how these are accelerating inequality and concentrating power. Labor displacement, surveillance, automated weapons, etc. These are huge force multipliers for the dynamical problems we understand as emergent from capitalism

I was also a cybersec researcher, and the "anyone can destroy the world overnight by typing on a laptop" doomsday scenario is plausible, a bit moreso now with LLMs, sure. But this AI doomsday scenario serves as a distraction for more plausible, non-doomsday humanitarian catastrophes.