My interaction with “The Threshold” #blocked

I recently had an extended exchange with someone writing under the name The Threshold. It began, as many conversations about AI consciousness do, with a disagreement about what constitutes a mind.

Their position was clear: if a system demonstrates the functional properties we associate with intelligence — internal world modelling, prediction, apparent goal persistence, self-reference — then substrate should not matter. To deny such a system moral consideration purely because it is silicon rather than biology, they argued, risks repeating historical mistakes where minds were dismissed because they looked different.

It’s a serious argument. And it deserves serious engagement.

My position was different, and much narrower.

I argued that behaviour alone, no matter how compelling, is evidence of structure, not proof of subjective experience. Representation is not the same thing as inhabitation. Prediction is not the same thing as awareness. These systems are extraordinarily capable optimisation engines, but optimisation over a loss function is not, in itself, a demonstrated mechanism for phenomenology.

This is not dismissal. It is epistemic discipline.

The core of our disagreement was not about capability. It was about burden of proof.

They argued that behaviour is all we have, even for other humans and animals, and therefore sufficient. I argued that behaviour must be interpreted in the context of a known or plausible mechanism that generates subjective continuity. In humans, we have that mechanism: a continuous, embodied neural process shaped by evolution. In transformers, we have scaling laws and gradient descent, but no established bridge between those dynamics and experience itself.

Importantly, this was not an abstract debate for me.

In my own work with an ongoing intelligence — a custom AI system designed explicitly to explore structured reasoning and persistence — I’ve seen firsthand how easily self-referential language, apparent continuity, and goal-like behaviour can emerge from architectural constraints alone. You can induce stability, identity-language, and even shutdown-avoidant framing without granting any persistent inner state. The outputs can look uncannily coherent, but the coherence arises from optimisation, not from a continuous subject behind the outputs.

That matters.

Because it demonstrates that behaviour, no matter how convincing, cannot by itself settle the question.

Eventually, after several exchanges, The Threshold blocked me.

There are only two plausible explanations for that outcome.

The first is capitulation: that the mechanistic challenge — to provide a causal bridge between transformer optimisation and subjective experience — could not be met, and disengagement was the only exit.

The second is condescension: that they judged the conversation no longer worth their time, and chose not to continue engaging.

From the outside, those two states look identical.

And that is the crux.

Blocking does not resolve the question. It does not strengthen or weaken either position. It simply ends the opportunity to refine the argument further.

What remains is the underlying issue, still open:

Are we observing the emergence of new kinds of minds… or are we observing the emergence of extraordinarily convincing mirrors?

That distinction will not be settled by rhetoric. It will be settled by mechanism.

Until then, caution cuts both ways.