He says "companies will behave like companies." I think there is a very different, and perhaps valid, perspective that says companies should be "socially responsible" and proactively try to benefit society. AI companies have the power to set the terms of what AI looks like; if you had that power, wouldn't you try to use it benevolently, and be relatively responsive to complaints from the people you affect? Isn't that the natural, human reaction to being in a widely resented position of strength?
https://en.wikipedia.org/wiki/Kremlin what is the Kremlin anyway? Five palaces, four cathedrals, the residence of the tsars, and the seat of the Soviet government. currently it's Putin's residence and there are also museum tours.
https://www.conspicuouscognition.com/p/wishful-thinking-is-mostly-a-myth claims that people do not typically engage in "wishful thinking". Motivated cognitions exist, but they're mostly beliefs that advance a person's social position, rather than wish-fulfillment fantasies. In fact, some of the most prominent "motivated cognition" falsehoods many people believe in are bad things -- belief in Hell for instance, or belief that current events are more catastrophic than they are.
https://simplex.pub/nonergodic-geometry/ Simplex explains their model of the generator of an LLM's output as a hidden Markov model and observe that the activations in the residual stream match the geometry of the data generators the transformers are trained on.
https://openreview.net/pdf?id=GRCbGii7Ti "Frameless Agentic System Theory", a way of defining a family of boundaries around an agent's observations and available actions based only on a dynamical system of states evolving over time.