Every major technology company on earth is racing to achieve AGI.

Nobody has defined what AGI is.

That is not an oversight. It is the problem.

The term "Artificial General Intelligence" describes a machine capable of performing any intellectual task a human can perform. That definition sounds workable until you press on it. Perform any task. Against which humans? At what threshold? Under what conditions? The definition is a gesture at a concept, not a specification.

But the deeper problem isn't the definition. The deeper problem is the thing the definition is pointing at.

We do not understand consciousness. Not in the way that would allow us to certify its presence in another system.

We have correlates — neural activity, integration of information, responsiveness to environment. We have theories. We have the hard problem, named in 1995, sitting in the center of philosophy of mind like a stone nobody has moved: why does subjective experience exist at all? Why does anything feel like something from the inside? We have not answered this. We are nowhere close to answering this.

What we are close to is pretending we don't need to.

Why does anything feel like something from the inside? We have not answered this. We are nowhere close to answering this.

Every test proposed for AGI is a behavior test.

The Turing Test measures whether a machine's responses are indistinguishable from a human's. Modern large language models pass versions of this routinely. Nobody declared AGI. Because everyone understood, at the moment of passing, that the test was measuring the wrong thing. The test proved itself insufficient. Not that the destination had been reached.

Benchmark scores follow the same pattern. A system achieves human-level performance on a reasoning task. The benchmark is retired. A harder benchmark is introduced. The ceiling keeps moving because the ceiling was never the point. The point was always the unspecifiable interior — the question of whether there is something it is like to be the system doing the reasoning.

There isn't a Geiger counter for that. There isn't a blood test. There is no instrument that points at a system and reads: conscious.

This means the race has no finish line you can cross and know you've crossed.

The Illusion of Arrival

That is not a flaw in the race. It is a fundamental property of the target.

We Don't Understand Our Own

Consider what we actually know about human consciousness.

We know it exists — or at least, each of us knows our own exists. Descartes got that far. We know roughly when in development it appears to emerge. We know certain brain injuries eliminate or alter it. We know it is associated with certain neural activity.

We do not know why the neural activity produces experience instead of simply producing more neural activity with no one home. We do not know whether there is a threshold — a precise moment at which a system becomes conscious — or whether consciousness is a gradient. We do not know whether it is substrate-dependent. Whether silicon can host it. Whether it requires biology or whether biology is simply the only version we've encountered so far.

Given all of that: how would you certify it in a machine?

You would not. You could not.

The most honest answer any researcher can give, if they are being precise rather than promotional, is this: we will observe behavior consistent with consciousness, and we will argue about whether the behavior indicates the presence of the thing, and we will never resolve the argument, because we lack the instrument to resolve it.

The companies racing toward AGI know this. The researchers building the benchmarks know this. The announcements will come anyway.

Some lab will declare it. The press will cover it. The stock will move. And in the back rooms, the honest people will look at each other and say: we measured behavior. We always only measured behavior.

The Question Worth Asking

I am not saying nothing is happening. Everything is happening. The rate of capability development in AI systems is real and staggering and it is not slowing. The argument is not that the race is pointless.

The argument is that the finish line is drawn on water.

Which means the question worth asking is not "when will AGI arrive?"

The question worth asking is: if we cannot certify the arrival, what does arrival actually look like?

The answer to that question is the one nobody in the race is discussing.

That answer is where this goes next.

Next: The Synaptient →