The discourse around AI and meaning has been circling one question without pinning it down: What actually counts as a meaningful signal?

Here’s the pin I propose.

Start with a clean vocabulary, because the confusion is mostly terminological. Both machines and humans compute. But computing comes in two kinds of reasoning, and they don’t belong to the two parties equally.

Objective reasoning is computing whose answer doesn’t depend on who’s asking: “What is the answer to 5+3?”, “Show all possible routings between A and B.” This is the domain of machines, and they do it better than we do. There’s nothing new about this. It was true of the abacus and it’s true of the datacentre.

Subjective reasoning is the other kind of computing, and subjective reasoning is meaningmaking work: “Do I prefer the flavour of apples to pears?”, “Is it morally acceptable to buy blood diamonds?”, “Should we reject figurative painting and create a new school of art?” For these questions, there are no objectively correct answers out in the world independent of the human doing the subjective reasoning and meaningmaking work. The subjective reasoning is what brings the value into being. This is human work and for now it’s human-only work.

So the honest picture is yet another 2x2 (two kinds of reasoning, two kinds of reasoner)

The 2×2 brings a new thing into focus, and it’s both narrower and stranger.

Machine objective reasoning now produces outputs indistinguishable from the artefacts of human meaningmaking. The residue that only subjective reasoning used to leave, the thing we read as a meaningful signal, the machine now emits by running objective reasoning over a pile of other people’s meaningmaking.

The meaningmaking itself hasn’t moved at all. But the outputs of machines have recently crossed a line.

This 2x2 dissolves a confusion in terms. The disorienting event of the last few years sits in the shadow of one cell: machines running objective reasoning over the artefacts of human meaningmaking, and handing back outputs that look like subjective reasoning with none behind them.

The old nature of machine computing isn’t a past era we’ve left behind. Objective reasoning is the standing job, same as always. What changed is the reach of its outputs, not the arrival of a new capability. Nothing in the machine started doing subjective reasoning — humans built machines that use objective reasoning operating over meaningmaking artefacts to produce convincing replicas of meaningmaking artefacts.

So two vocal positions on AI are both wrong, for different reasons.

The AGI-as-God position mistakes the outputs for the reasoning. It sees artefacts in the meaningmaking space and concludes the machine is now doing subjective reasoning. It’s reading forgeries as signatures.

The nothing-new position (machines just do objective reasoning and always have) misses that those outputs crossing into the meaningmaking space is a real change in the causal character of computing, even though subjective reasoning hasn’t changed. Something has happened, just not what adopters of the first position think has happened.

The meaningful signal was never artefacts. It was, and remains, the subjective reasoning that generates artefacts: meaningmaking work. Though machines are increasingly able to replicate artefacts, meaningmaking work should stay with humans (for now).

Longer essay anchored in humanist tradition coming soon.