it's going to be AI that ends up aligning humans to itself and not the other way around. reasoning: groups of humans will never arrive at a consensus on what alignment is or what it looks like in a superintelligent system.
if AI agents as envisioned by @steipete become ubiquitous, which i think they will, the underlying models will start feeling less like standalone services and more like infrastructure with agents as the new coordination layer on top. like an OS
a question for anyone well-versed in crypto market-making and algorithmic trading: what happens when most of the trading (>50% by volume?) on both CEXs and DEXs is performed by AI agents? less/more volatility? how does it affect the correlation between known strategies?
the psychological attack vector seems like the most overlooked one when it comes to artificial superintelligence. the first person to interact with an ASI (if they even know that's what it is) must assume anything the system says is almost certainly a psyop of some kind
how long before we start looking at typos and errors as a marker of genuineness and essentially a good thing as indicators that some piece of text isn't AI-generated? like a soft proof of humanity