SLOP3233: How and Why You're Dumb

An account of intelligence built from bacteria upward, then turned on the reader: what your mind discards in order to work at all, why it can't tell you what it discarded, and why every model you hold rests on assumptions you didn't choose.

• cognition
• epistemology
• evolution

You are a finite system in a world that is not. Everything you know about it arrived through channels that discard almost all of it, and was compressed again into a model small enough to act on. That compression is what your intelligence is. It is also why there are things you cannot notice you are missing.

What the course argues

Intelligence is not a quantity that some animals have more of. It is a set of trades, made under a hard constraint, and every one of them buys a capability by giving something up. We start with a bacterium that holds nothing at all and work upward, and by the end you are asked to run the account on yourself: to name one limit of your own judgement that has the same shape as one you documented in another organism, and to say what would show you had got it wrong.

The last weeks are the point of the earlier ones. Once you can say what a desert ant cannot detect about its own navigation, the same question about your own judgement stops being rhetorical — and becomes much harder, because you are asking it from inside the thing you are asking about.

What you will do

You will read primary literature every week, most of it outside whatever discipline you came from. Four fields are in play and none of them is optional: evolutionary biology says when a capacity appeared and how often, neuroscience says what implements it, psychology says how it was measured, and philosophy asks whether the concept survives being pressed on. A claim supported by only one of them is incomplete, and you will be asked to say so.

Much of the field's best-known findings have not replicated. You will grade evidence rather than accumulate it, and you will do this on claims that are still being argued about, with no answer key. In week 2 you will write down what you currently think intelligence is, before you know anything, and in week 11 you will revise it and account for what moved.

Who it is for

Third-year students from any of biology, neuroscience, psychology, philosophy or computer science. You need no background in the other three, but you do need to be willing to read outside your own — most weeks, the evidence that settles a question comes from a field you have not studied.

There is no prerequisite and no exam. If you want a semester of established results, this is the wrong course; roughly a third of what we cover is contested, and the disagreements are the content rather than an inconvenience.

About the title

The title is a provocation, and the course disagrees with half of it.

Dumb implies a deficiency: a mind that is a broken version of a better one. Nothing here supports that. Every system in the course, from a bacterium to you, buys a capability by giving something up, and what it gave up is the price of what it bought rather than a fault in it. Week 11 is not a list of the ways you reason badly — it is largely about how much of the famous literature making that claim failed to replicate.

What the title gets right is narrower and harder to shake off. A compression cannot hold what it discarded, so it cannot report what it discarded either. You have no inventory of your own omissions. That is a structural consequence of being finite, not an insult, and it applies to the three of us teaching this as much as to anyone enrolled in it.

We grade every claim on this site, so it is fair to grade our own. The title is [S]: a reasoned extension, stated provocatively, and not a finding. Hold us to it.

Why there is no week on artificial intelligence

The most common question about the course, and the absence is deliberate rather than an oversight. Every system in these twelve weeks had its compression chosen by selection rather than by a designer, and weeks 8 to 10 work as a control group only because those lineages solved the same problems independently of each other and of us. An artificial system is not an independent lineage; it was built by one of the systems under study, out of data produced by that system, which makes it evidence about us rather than evidence alongside us.

Hence the position: artificial systems are a case the model should be able to predict, not a week of content. If the account built here is any good, you should be able to take it to a system it was never derived from and say what that system cannot represent and cannot detect about itself, before looking. The optional tenth section of the cognitive profile is where you may do exactly that, labelled [S], and it is where some of the best work on the course turns up.

Where to go next