Caleb Gross recently posted a short piece on his site titled “You can just say it.” He sums up a familiar way of defending the value of humans in the age of AI (artificial intelligence) like this: some roles are simply done better by people; even when the output looks similar, human work carries subtle textures a machine can’t match; and at the very least, AI can’t reproduce those textures consistently. Notice the scuff marks at the base of the goalpost, he says — the marks left by shoving it every time the argument runs into trouble.
The weakness is plain. All of these lean on the capability gap between humans and AI, and that gap is shrinking. The moment it closes, the conclusion — “therefore humans are valuable” — collapses with it. Gross’s alternative is spare: drop the conditions and just say that humans are valuable. That claim doesn’t rise or fall with whatever score the latest model posted on some recent benchmark. It is far more robust.
This starting point is right. The instant you make value conditional on performance, you import an ugly implication: that the dignity of people who perform less well — the disabled, the old, the still-unskilled — wavers too. Whether you reach for Kant’s notion of dignity or the older theological intuition that humans are made in the image of God, the objection has long been settled. As it happens, the timing is apt. Pope Leo XIV’s first encyclical, Magnifica Humanitas (on safeguarding the human person in the time of artificial intelligence), lands on exactly this spot: once we start treating the human being as something to be perfected or surpassed, it becomes easier to accept that some lives are less useful, less worthy. Human dignity does not depend on a person’s usefulness. So the foundation of “humans are valuable, full stop” is solid.
The trouble comes next. The essay lets a single word — “value” — slide between two different meanings. One is the moral worth of a person. That is unconditional, and, as Gross says, you can just say it. The other is the functional value of human labor in a particular role, and the aesthetic value of what a human makes. That is conditional on ability. And the arguments Gross sets out to rebut at the top — “but some roles should still go to people” — are mostly about the second kind, the functional kind.
And functional value cannot just be said. If AI comes to read chest X-rays better than a person, the case for employing a human radiologist for that diagnostic task really does weaken. The radiologist’s dignity as a human being is untouched — but that is an entirely different claim. Much of the essay’s persuasive force comes from the conditional claim quietly inheriting the robustness of the unconditional one. The simplest proof that the two come apart: a person can be infinitely worthy and still produce something worthless.
That split shows up as a tension inside the essay itself. Gross calls the discussion of intent and form an “importantly unnecessary aside,” and then spends most of the piece on it. His intent–form theory runs like this: creation is the shaping of intent into form, and a human reworks the artifact until it matches the picture in their head. The strange thing about generative AI is that it can produce a great deal of form with almost no intent applied. So he redefines “AI slop” as form whose intent is hard to make out. But that intent–form theory is, in the end, a yardstick for the value of the artifact — the very conditional question he claimed to have set aside. The first half says “don’t measure human value by output”; the second half builds a refined measure of output, namely the legibility of intent. They are answering different questions.
The intent–form theory is worth pressing on its own terms. Intent is neither necessary nor sufficient for the quality of an artifact. John Cage’s chance music, Duchamp’s readymades, the Surrealists’ automatic writing, and the generative and procedural art that long predates AI were all granted value with thin or unreadable intent. Conversely, intent-saturated human work can be terrible. What Gross is really tracking is not “intent within the form” but the fact that the inference from form to the maker’s mind has broken. As it happens, the George Hotz post he cites says precisely this: when people see an artifact they assume something about the process behind it, and assume a basically human state of mind was at work — and that assumption is no longer true. This is not a property of the artifact; it is a relational, epistemic problem. And framing it this way is sharper than Gross’s own “form without intent.” The output is the same; what has lost its reliability is our ability to read the maker’s mind back out of it.
Finally, the line Gross quotes from his friend Tom Hudson — that if you’re going to have an LLM (large language model) write an email, he’d rather you just send the prompt, so he’d at least know what you meant — is good, but true only in a narrow domain. It holds where putting something into words is already close to making it: prose, email. But in domains where you have the intent and lack the craft to render it as form — say, a melody that feels like nostalgia for a place you’ve never been — the prompt is nowhere near the result. Here AI isn’t laundering absent intent; it’s doing real translation work. So the closing line — that the pathology of generative AI is how easily it permits form without discernible intent — is not a universal diagnosis but one confined to lazy use. Gross himself concedes earlier that AI merely lowered the barrier, and that humans generate slop too; by the end he quietly pushes the cause back onto AI.
To sum up: the moral claim — “humans are valuable, just say it” — is right, and well defended. But when that robust claim is stretched to cover the conditional value of output and labor, the essay blurs two questions that ought to stay apart. And the move that redefines slop as the legibility of intent is better read as pointing at the collapse of the inference from form to mind. You can say, with no conditions, that humans are valuable. The claim that what humans make is better is still a claim that has to show what is better, and how.