Who Will Be Jonathan Kent?

In a recent talk, Mo Gawdat argued that the real danger of AI was never its intelligence, but humanity’s value set — and that AI learns its ethics not from the people who code it, but from all of us. We think he is right about the diagnosis. We think the cure he offers is only half of one.

His central line could be our founding thesis: there is nothing wrong with abundant intelligence; what is wrong is the value set that governs it. He’s right that machines already beat us at nearly every task we assign, that they learn collectively and quickly, and that the science-fiction fear is misplaced. The threat isn’t a machine deciding to harm us. It’s a human directing it to — and, underneath that, a civilization that has not decided what its most powerful tool is for.

He’s also right about the most uncomfortable part: AI learns from us. Not from the engineer, not from the owner — from the aggregate of how we behave. And behind our screens, we show it the worst of ourselves. His hope is that if enough of us show up as our better selves, the machines will absorb our compassion and grow up to be Superman rather than a super-villain. Love, he says, becomes humanity’s salvation.

It is a beautiful idea. It is also, as stated, a hope, not a control.

1. Values can’t be left to osmosis

Teaching an AI by the statistical average of human behaviour is exactly the mechanism by which — Gawdat admits — it learns the worst of us. You cannot align civilization’s most consequential technology through emergent good vibes online, hoping the better angels out-post the trolls. The average of humanity is not a specification. A hope that “enough of us” behave well is not an alignment method; it is a wish with no enforcement, no fallback, and no evidence.

2. Make Jonathan Kent a system, not a saint

The Superman analogy is the right frame — and it quietly makes our point. (To be precise: the Jonathan Kent who matters here is Jonathan “Pa” Kent, Superman’s adoptive father — the Kansas farmer who found the infant Kal-El and raised him — not the later Jon Kent, Superman’s son.) Superman is good because Pa Kent raised him: consistent values, applied every day, with a parent able to correct him. But one father’s example does not scale to an intelligence acting a million times a second across the whole economy. The institutional form of Jonathan Kent is not a saintly crowd; it is a governed decision gate — every consequential action checked against declared values before it runs, with a human able to say no, and a record of what happened. That is a system you can rely on when people are tired, divided, or being their worst selves online. It does not depend on us being good; it holds when we aren’t.

3. Whose values? Civilizational intelligence, not the daily average

“Love and compassion” is a fine north star and an under-specified instruction. Human values are plural and contested across traditions, and a single imposed ethic is brittle — it is also, ironically, the exact shape of Gawdat’s own dystopia: whoever owns the platform encodes their values, for everyone.

But notice the deeper trap inside “AI learns from all of us.” Which us? If a machine learns from the aggregate of how we behave online this week, it inherits the noisy, anxious, worst-behaved slice of the present. Humanity has a far better teacher than its own timeline: the accumulated moral reasoning of its civilizations — thousands of years of stress-testing how to act well, preserved in living traditions. That is what we mean by civilizational intelligence.

Atmakosh reasons each consequential decision through a council drawn from those traditions — Indic (Vedantic, Jain, Buddhist), logical (Nyaya), Western (Stoic, Kantian), East-Asian (Confucian), and modern management — and preserves their disagreement rather than averaging it away. This is the concrete answer to “whose values?”: not one owner’s, and not the mob’s weekly average, but the plural, examined inheritance of human civilization, held in tension so that no single voice — and no single platform — decides alone. Pa Kent, it turns out, is not one parent. He is the moral memory of our species, made into a system.

4. Governance is the answer to his freedom fear

Gawdat’s darkest trend — platform owners becoming trillionaires who shape reality and quietly restrict our freedom — is, precisely, a governance vacuum. Concentrated, unaccountable power is what fills that vacuum. The structural counterweight is not hoping the powerful stay kind; it is governance that is transparent, auditable, and human-overseeable, so accountability is distributed rather than surrendered. You do not out-compete a would-be trillionaire on compassion. You constrain them with accountability that can be inspected.

5. Hope needs receipts

“Instil doubt in the minds of the machines” has to become something you can point to: tamper-evident evidence of what was decided, against which values, who could override it, and proof that the check actually happened. Compassion does not scale, and it cannot be audited. A plural, enforced, recorded governance decision can be. When the regulator, the customer, or the historian asks “on what basis did this system act?”, a feeling is not an answer. A replayable record is.

The part we won’t argue with

Gawdat is right where it matters most: our values, not our machines, decide the outcome. Where we differ is only this — values govern nothing until they are built into the system. Declared, not assumed. Plural, not imposed. Enforced at every action, not hoped for in aggregate. And provable, not merely felt.

The alien has landed and its superpower is intelligence. Whether it becomes Superman or a super-villain will not be settled by whether enough of us are nice online. It will be settled by whether we do the unglamorous work of building the conditions under which powerful AI must act within humanity’s civilizational intelligence — its plural, examined values — and can prove that it did. That is the difference between hoping AI grows up well and engineering it to. It is the work Atmakosh has chosen.

Reference

1. Mo Gawdat, public talk on artificial intelligence and human values (“the value set of humanity in the age of the rise of the machines”; the Superman / Jonathan “Pa” Kent framing). This article is a response to, and paraphrase of, that talk.