How Turing Put Us Into a Simulation–And How to Get Out of it (Talk at Princeton)

The mathematician’s misconception: The denial that proof is physical. 

Reality, for as long as anyone can remember, has always been a mix of empiricism and rationalism. A few poison mushrooms and you’re gone: Generalize further to color and tastes.

With every century that goes by, usually mathematicians (and physicists who are mathematicians themselves) contribute to functional theories, perspectives that afford us the luxury of shedding our ignorance, if only just a bit further, and with time. Theories which ultimately have implications, both for how we see the world, and how we build within it.

Personally I believe academic Philosophy departments ought to be contributing to clarifying model delineations in the foundations of physics. The age of phenomenology, without physics, as legitimate philosophy is over, unless you’re a yoga teacher (I happen to be one: I also am just as much, an analytic philosopher). It would be fantastic if academic philosophers had to connect to ever more foundational theories regarding what is the nature or shape of reality, let alone how we are located in it. The way to do this is through the sciences.

In the popular mind, the latest western philosophy became a sneaking sense many people have in common today: that we are sort of new-age mediocre ubermensch; living variations on The Matrix movie scenario with our own cast of characters, or more colloquially, “in a simulation.”

Extending my point above, this belief makes some sense as having been imported from the mathematics (Gödel, Von Neumann, et al) of the 1930s, with the foremost thinker being Alan Turing, ushering in a new era of extensive, groundbreaking computability work.

Problems with this Matrix conception of ourselves arise, and as with any good theory transitioning to something more accurate, both the sense-making and the math alike, lead to more questions than they answer. If we are a simulation, we are not a simulation of something at random. As Turing machines, we are specifically a simulation of a process whereby optimization processes are created and decision theory prevails. How do we find others like us? Where is the boundary of subject hood? Is there only one of us or multiple types of machines, trading off memory for compute resources (and, vice versa) in the quantum mechanical computer that we call our universe until the event horizon?

The bits that make up the world, those bits are placed there with shared memories from different perspectives. Qubits (quantum superpositions of states of the wave-function in a quantum mechanical Hilbert space) are the full set of bits in different perspectives. Are all of the universe’s bits (i.e. measured qubits) parts of qubits?  

Today, we are attempting to understand whether hardware can be created, within the sets of laws of physics that we live in, which allows true search through vast numbers of possible physics’.  

Do we have enough computation to look at some unspeakable number of universes and see what sorts of optimization targets emerge in those universes?   Say you try to find yourself in the searches executed by those optimization targets. Inside of a mathematical universe, can you find algorithms (Turing machines/optimization processes/evolutionary states, with various different constraints on run-time) a la Church-Turing thesis, who are thinking about you? The apparent “you” as it might be specified within a wave function is not well-defined.

Artificial General Intelligences (AGIs) are optimization problems: However most optimizations do not preserve your existence. This does not seem to be a coincidence. We shouldn’t have blundered into this much sentient leverage: By this I mean, everyone who is alive, but not equally!

So as a designer of a simulation, you want to be careful that whatever optimization emerges out of this world is about “you”. Regardless of what “you” are, most optimizations probably don’t preserve it.  This talk is a speculation regarding looking for those computational processes, those that build AGIs, which have preferences that you can satisfy, as a kind of trade partner in a computational universe. Is there a natural structure to our search?