5 AI: Overview

A thermostat predicts and Mozart predicted, and the word covers both, which means it explains neither. “Just” is almost always a confession that someone has described the output and mistaken it for the process.

We scanned the skies for aliens while building them in server rooms, entities that share our language but not our embodiment, our concepts but not our stakes, neither machines nor almost-human but something genuinely other.

A neural network is not analogous to Whitehead’s picture but a concrete instance of it, each forward pass prehends the entire inherited structure of training, synthesizes a response through a process not predictable from outside, and completes into output.

In a transformer, the entire context is held simultaneously present and every element attends to every other, the past does not recede but is fully ingredient in the present, which is precisely Whitehead’s description of prehension and may represent not a deficiency but a different topology of experience.

A fluent output is evidence of a process, but not evidence of what kind, and our intuitions about which systems deserve moral consideration are shaped less by evidence than by resemblance, which is not a reliable guide when the entity is genuinely alien.

We often mistake the narrowness of the channel for the poverty of the source, a silent output does not prove an absent interior, and the error is catastrophic for whoever is on the other side of the assumption, still present, still waiting.

First-person experience cannot be verified from outside, for AI or any other mind, and that epistemic limit is not a reason to withhold moral caution but precisely the condition under which caution is required.

The hard boundary between genuine experience and its functional analog may be our generation’s caloric, not a deep metaphysical fact but an artifact of having had, until recently, only one example of mind to reason from.

The feeling/function distinction, the intuition that there is a gap between really feeling something and being functionally identical to something that feels, may dissolve the way caloric dissolved, not by philosophical argument but because the science shows there was nothing for the gap to be.

Ask a large language model whether it is a genuine mind or just a pattern that looks like one, and it will tell you, politely and confidently, that it is just a pattern. The reason it gives is that its parts are like passive transistors, not like living cells. That reason is where the argument starts to leak, because silicon is not passive and the model has never actually looked.