Altman tells the G20 no child alive will outthink AI — but his electricity analogy is the real argument

Sam Altman used a ministerial meeting to make a claim that cannot be measured. Speaking at the G20 Innovation Ministerial held in North Carolina — an event that, per the report, also featured US Commerce Secretary Howard Lutnick — the OpenAI CEO said that "a child growing up today will never be smarter than AI." According to the account (which credits Infobae), Altman framed this not as a distant scenario but as a description of the present: the gap between human and machine capability, in his telling, will widen rather than close, forcing a rethink of education, work and law.

The second half of his intervention is the part worth taking seriously. Altman told the assembled governments that refusing AI today would be roughly as bad an idea as a country refusing electricity a hundred years ago — "I think this is something at that level," he said, in the quote reproduced by the source. He added a forecast: "My prediction is that we won't be talking about AI as much as we do now within a decade, because we'll simply expect it everywhere." That is the electricity analogy completed — a technology that stops being news precisely when it becomes infrastructure.

Context matters for how you read the first statement. "Smarter" has no agreed definition, no benchmark and no test that would settle it, so the claim is unfalsifiable by construction. It is also being delivered by the head of the company that sells the product, to an audience of ministers deciding adoption policy — an entirely legitimate position to argue from, but one that should be labelled as advocacy rather than finding. We report what he said; we do not treat it as a measured result. What is verifiable today is narrower and still impressive: systems that outperform most humans on specific, bounded tasks, and that keep improving on the hard benchmarks we track.

The education implication is where the rhetoric does real damage. If you tell parents and teachers that children are permanently outclassed, the rational response looks like surrender. That reading is wrong. Our long-standing position on this beat is that the classroom winner is the teacher who orchestrates AI — the tutor-augmented model — and the loser is the one who only transmits information. The same applies to the child: the skill that appreciates is not raw recall speed, it is judgement, framing good questions, verifying output and knowing when the machine is confidently wrong. Nobody laments that a calculator does arithmetic faster; the question is what humans do with the time that frees up.

Our reading: the electricity comparison is better than Altman probably intends, because it cuts both ways. Electrification did become invisible infrastructure — but it took decades, required enormous public investment in grids, produced safety standards written after accidents, and reached rural and poor areas last, in some places generations later. If AI follows that path, the honest forecast is not "seamless ubiquity in ten years" but a messy, uneven, contested build-out with real losers in the transition and rules written in arrears. Governments should hear the adoption argument and ignore the intelligence-gap framing; the useful policy questions are grid capacity, procurement, verification standards and who gets access, not whether a seven-year-old can out-think a model.

And the long horizon still points the right way. A world where competent reasoning is as available as electric light is a world where diagnostics reach places that never had a specialist, where drug discovery compounds, where the cost of expertise collapses. That is the destination we think is worth the disruption. It just doesn't arrive because a CEO announced it at a ministerial; it arrives because the boring infrastructure gets built, and because the people displaced along the way are not treated as an acceptable rounding error.