The second law of thermodynamics has always bothered me. Over time, energy spreads out and less remains to do work. It posits that the universe is destined to die in a slow dissipation of heat—what’s been called the heat death. Yet look around, and you find a world full of complexity: meerkats, memes, and minds. Why? How?
The title of this thesis, Complex Investigations, references Ludwig Wittgenstein’s Philosophical Investigations and Stuart Kauffman’s Investigations (who was referencing Wittgenstein), both of which begin from a place of questioning and lead us down a path, one that must be walked, toward a better understanding. But the comparison speaks to approach, not substance—I do not claim my writing or ideas match theirs in any capacity, only that it’s “of a kind.” In that sense, this isn’t a typical academic work in form or argument.
As I contemplated this work, I realized a large part of what I wanted to do was think through the questions raised by Thomas Nagel in Mind and Cosmos. Nagel argues that natural laws do not account for the emergence of minds, meaning, and value—things we see all around us. These features, he claims, are “not just an afterthought or an accident or an add-on, but a basic aspect of nature” (Nagel 2012, 20). The science of complexity points, I think, toward a potential response to his claim. What follows is, in many ways, an outline for a broader project of naturalizing intelligence, language, and minds through the lens of complexity.
Nagel’s intuition is that any such explanation would likely be teleological—that is, the universe would have a goal that moves toward higher and higher levels of complexity. If the answer is not teleological, then he argues it would have to be non-deterministic and physically biased toward the creation of higher orders of complexity (Nagel 2012, 92–94)—in essence, something like a fourth law of thermodynamics (or a first law of complexity?). This is also one of the main questions Kauffman pursued in his Investigations.
To that end, the best way to think about what follows is as a large chain-of-thought prompt written with a relatively high temperature (amount of noise added), designed for a large language model (LLM) to suggest questions for further research and, for fun, candidates for a fourth law of thermodynamics. The work takes the form of numbered sets of statements, both to structure my thoughts and to structure the prompt for the model (LLMs love outlines). There’s a growing school of thought that we should be writing for the AIs; in a way, all writing is programming now.
The numbered sections build on one another over time. The goal is to weave together concepts from biology, computation, language, artificial intelligence, and complexity in a way that hangs together. Section 1.1 introduces a simple non-teleological complex system. Section 1.2 seeks to define what can and cannot be known, which is important for the subsequent sections. Section 1.3 expands the complex system into life. Section 1.4 introduces the importance of induction and how it relates to complexity and intelligence. Section 1.5 frames language, and relatedly meaning, in terms of complexity. Section 1.6 extends the ideas to minds. Lastly, section 1.7 introduces the idea of complexity completeness.
As the argument progresses, I try to sketch what an answer to Nagel might look like more broadly and explicitly address meaning and minds in the later sections. Interlaced throughout the document is a series of images, generated with various AI programs and then manipulated by me through a recursive process using prompts that reference complexity and evolution. To me, they are a good example of the recursive combination of an envelope and noise. In this way, they touch on the last of Nagel’s questions—value, or at least aesthetic value. A short discussion on the images, along with some similarly produced paintings, can be found in the appendices.
The work is intentionally noisy for two reasons: first, it echoes the ideas it explores (composable bits in a noisy envelope), and second, I want wide variability within the prompt to increase the temperature of what comes back. To that end, certain statements simply bring relevant sources or people into the context window—as we’ll see, meaning is in the richness of the relationships.
On the topic of AI, I should disclose up front how and whether it’s been used. My personal rule is I don’t use AI to initially read or write—only to refine understanding and writing after I’ve done both. At its best, AI should accelerate our understanding, not replace it. I have also used AI extensively to clarify my thinking and validate my understanding. As one example, I had a long discussion with ChatGPT about the system I described in section 1.1 and whether it was actually Turing complete (we decided it was, or close in principle). And for the record, I loved the em dash way before AI was even a thing.
Once complete, the main text was used as a prompt to generate both a list of questions (A3) and candidates for a fourth law (A4). The list of questions can be viewed as the starting point for a more formal academic project (i.e., 2.0). Picasso is reported as saying computers are useless because all they can do is give you answers. This seems to have changed. The questions in the appendix test whether it has.
I try to ground the discussion in science and, from there, ask if science can tell us anything more generally about philosophy. Being neither a scientist nor a philosopher, I am on dangerous ground from the start. The work also lacks the precision and detail that either domain would demand as a piece of pure academic writing (unlike Newton, I feign hypotheses). As the work progresses, it ventures farther and farther out on a limb. And admittedly, we start fairly far up the tree.
My hope is that if I combine enough of these bits in a noisy envelope, it will give me some closure. But this is just one of many paths we could carve out of a near-infinite combinatorial space. Parts of what follows are most certainly wrong. After all, we’re talking about the complex.