Companion paper to Madness and Consent. Abstract: Bounded self-certification in Turing machines fails because self-simulation necessarily incurs a strictly positive temporal overhead. We translate this operational constraint into a domain-theoretic framework, defining an operator that advances a finite halting observation from time bound i to i+1. While no bounded machine can achieve a fixed point under this operator, the iterative process forms an ascending ω-chain. The Scott limit of this chain resolves to the least fixed point of the operator, representing an unbounded computation that fully captures the machine's halting behavior. Our construction provides a novel perspective on the halting problem, framing the transition from finite observability to the least fixed point as the continuous deferral of the diagonal.
Contributor to the volume edited by Mattin and Anthony Iles and published by Minor Compositions. Contributed a poem interweaved throughout the book, "I think I was suppposed to die", under the pen name Loss Choi.
I pose the question, what is it like to be racist?, with Sara Ahmed in conversation with
Charles Mills. I consider the possibility that racists are nowhere, cannot experience, and do not
die.
This paper is about uncomputability as freedom. It defines uncomputability as good, from which
follows that computing what is uncomputable is bad.
The conclusion is wrong. I said that once there is uncomputable AI, the human race might as
well go extinct. This is completely wrong, since AI cannot be uncomputable. Flesh is required to be uncomputable, as I discuss in Madness and Consent. My only defense for this position is that I was barely twenty years old when I wrote this essay in 2018.
This is a collection of notes I wrote when I was teaching CS198-97 Philosophy of Computation at UC
Berkeley in Fall 2017. The class is designed to build up intuition about what computation is, how
computation is ontological, and culminates in a proof
of why you ought not compute an arbitrary human.