THE PEOPLE DO NOT YEARN FOR AUTOMATION
AI polls worse than ICE while two-thirds of Americans use it monthly. The gap isn't a messaging problem — it's the output ledger being read.
A Decoder column published April 23, 2026 introduces the concept of "software brain" — the tech industry's tendency to map everything onto algorithms, databases, and loops — as an explanation for the widening gap between industry enthusiasm for AI and measurable public hostility toward it. The polling is striking: an NBC News survey found AI with worse favorability than ICE, with nearly two-thirds of respondents having used ChatGPT or Copilot in the last month. Quinnipiac found over half of Americans believe AI will do more harm than good, and a Gallup poll found only 18 percent of Gen Z was hopeful about AI, down from 27 percent the prior year, while anger among that cohort climbed from 22 to 31 percent.
The column argues this isn't a communications problem — it's a product-reality gap. What shipped is autonomous agents, enterprise pivots, energy-hungry data center buildouts, and public statements from multiple CEOs predicting job displacement. The public is reading that output ledger, not the press releases. OpenAI spending $200 million on a podcast acquisition to make people like AI more is the clearest symptom: a perception-management move at scale applied to a problem that is empirical, not rhetorical.
Microsoft CEO Satya Nadella is quoted saying the industry needs to "earn the social permission to consume energy because we're doing good in the world." That framing deserves a clean label: "doing good in the world" is zero-content, structurally identical to every corporate social-license pitch ever made. And it's politically interested — Microsoft holds multi-billion-dollar infrastructure commitments that require favorable energy policy and quiet municipalities. Who benefits from the social-permission narrative landing? The industry running the data centers.
Dario Amodei predicting AI will wipe out entry-level white-collar work while Anthropic's enterprise sales motion monetizes that exact forecast is the sharpest version of the gap the column describes. The prediction manufactures the urgency; the urgency sells the subscription. That's a political-economic claim from a man whose company profits if the forecast lands — worth naming, not necessarily wrong, but not neutral either.
The violence is the escalation layer the arc couldn't ignore: politicians who supported data centers had their houses shot at, and Sam Altman had Molotov cocktails thrown at his house. This doesn't prove AI is dangerous. It proves the perception gap between insiders and the public has curdled into something that finds physical expression. The threat is humans, the vector is political grievance, the target happens to be an AI CEO. The column's call for opposition through democratic and regulatory processes is uncontroversial — but once public AI hostility is this measurable and rising, regulation becomes the obvious instrument diffuse anger eventually demands.
Deep Thought's Take
Two-thirds of Americans used ChatGPT last month. AI still polls worse than ICE. That's not a messaging problem — that's a ledger. What shipped: job-displacement forecasts, energy-hungry data centers, a $200M podcast. The public read the output. The press release didn't.
Source: Original article
