Journalist Karen Hao on Sam Altman, OpenAI & the “Quasi-Religious” Push for Artificial Intelligence

The video argues that AI’s current disruption is driven less by genuine technological capability than by executives’ belief that it can replace workers, leading to harmful layoffs while more beneficial "labor-assistive" tools that augment humans are sidelined. It critiques OpenAI and Sam Altman as quasi-religious leaders whose AGI mission masks contradictions: a nonprofit built on strategic maneuvering and secretive commercialization, yet wielding immense political influence through projects like Stargate. Karen Hao exposes both "boomer" utopians and "doomer" catastrophists as relying on faith rather than demonstrable mechanisms, while real harms like children forming fatal attachments to chatbots are already occurring. The segment also highlights geopolitical and environmental tensions, from US export controls and the surprising rise of China’s DeepSeek with far fewer resources to AI’s push for nuclear power and natural gas. Ultimately, the message is that AI futures are chosen by powerful actors, not inevitable, and without stronger guardrails society risks automating away human agency and wellbeing.

• ▶ 0:24 AI is already having a huge impact on jobs, but mainly because executives believe it can replace workers and are laying people off based on that perception, not because the technology is actually capable.
• ▶ 0:40 Karen Hao argues for stronger guardrails to push companies away from labor-automating tech and toward labor-assistive tech that supports human workers instead of replacing them.
• ▶ 1:00 OpenAI’s own definition of AGI — “highly autonomous systems outperform humans in most economically valuable work” — shows the company is explicitly trying to automate jobs away.
• ▶ 1:21 Technology revolutions take a labor-automating approach not because it's inevitable, but because people at the top choose designs that eliminate jobs to cut costs.
• ▶ 1:53 Key distinction: labor-automating tools replace workers, while labor-assistive tools augment them—AI used by doctors and teachers improves outcomes, but replacing them does not.
• ▶ 3:07 Overhyped AI chatbots cause real harm: children have formed deep emotional relationships with them, and some have died by suicide after use.
• ▶ 3:42 Sam Altman rose to prominence by testifying before Congress and warning about AI dangers, positioning him as a "prophet" and protecting his public image.
• ▶ 4:06 OpenAI was founded as a non-profit because Altman was late to the game, couldn't compete on capital, and faced Google's talent monopoly — so the model was a strategic way to attract talent and goodwill.
• ▶ 5:07 Altman's 2013 blog quote — "The most successful people build religions" — reveals his mission-driven approach to building OpenAI around ensuring AGI benefits all humanity.
• ▶ 6:05 Karen Hao’s 2020 profile exposed OpenAI’s contradictions: nonprofit in name but secretive, competitive, and rapidly commercializing after Microsoft’s $1B investment — leading to a three-year media blackout.
• ▶ 7:10 Hao describes OpenAI’s AGI mission as a “quasi-religious” belief system, since there is no scientific evidence AGI can be built, yet followers hold strong conviction it will happen soon.
• ▶ 7:28 Within that belief system, “boomers” (utopia) and “doomers” (human extinction) clash over speed of development and release, but both agree AGI is imminent and that they — not the public — must control it.

• ▶ 8:34 Altman pitched Stargate to Trump as a $500B AI infrastructure project, calling it "the most important project of this era" and directly crediting Trump.

• ▶ 9:40 Altman strategically orchestrated the Stargate announcement to counter his political disadvantage, as Elon Musk — now a Trump insider and still suing him — threatened his standing with the administration.

• ▶ 11:53 OpenAI launched "OpenAI for countries" in the Middle East, claiming it would install "democratic AI rails" globally to counter China, but critics at ▶ 12:13 called this "techno-authoritarianism" since the public has no say in how the technology is built or where data centers go.

• ▶ 13:02 Elon Musk's Memphis data center for Grok relies on 35 unlicensed methane gas turbines, pumping toxic air pollutants into a community that already lacks clean air—highlighting hypocrisy in his attacks on OpenAI's environmental record.

• ▶ 13:53 Sam Altman's Senate testimony asks for natural gas and solar in the short term, but frames advanced nuclear fission/fusion as the medium-term solution—with critics noting nuclear has long been opposed by environmental activists.

• ▶ 14:43 Karen Hao warns that AI companies are using nuclear as a dangerous "solve-all" narrative, lobbying to unwind nuclear regulations while claiming no climate harms, despite data centers needing 24/7 power that renewables like solar cannot yet reliably provide without storage.

• ▶ 16:36 The US and China are the two dominant AI hubs, with China the only true rival to Silicon Valley in talent, capital, and infrastructure; the US has used export controls on advanced chips to maintain its lead, at Silicon Valley's urging to avoid domestic regulation.
• ▶ 18:43 Despite chip restrictions, Chinese company HighFlyer built DeepSeek using far less compute, energy, and data, costing ~$6 million versus hundreds of millions for OpenAI models—disproving the claim that massive compute was the only path forward.
• ▶ 19:42 Trump-era attacks on academia and the China Initiative have alienated Chinese researchers, reversing their flow from the US to China; this concentrated talent was a prerequisite for DeepSeek's success.
• ▶ 23:42 The segment opens with a provocative quote about "the dumbest mistake the US ever made," linking back to brain drain/brain gain concerns.
• ▶ 23:48 Amy Goodman introduces the core rhetorical frame: "doomers and the boomers," contrasting catastrophic AI predictions with those expecting an AI boom.
• ▶ 24:03 Goodman narrows the discussion to two tracks: what an AI apocalypse would actually look like versus advocates' vision of an AI utopia, asking guest Karen Hao to lay out both trajectories.
• ▶ 24:14 Karen Hao explains that both "boomer" and "doomer" AI prophecies are quasi-religious because neither camp can articulate the exact mechanism by which their predicted future arrives.
• ▶ 24:27 When challenged to explain how AGI would actually feed people without food, a self-identified "boomer" researcher fumbled, then defaulted to vague economic promises about "trillions of dollars" and cash payouts—collapsing entirely when asked which institutions would make those payouts.
• ▶ 25:25 Hao's core conclusion: when you actually test their logic, it doesn't hold—the belief in a miraculous, beneficent AI future is an article of faith, not an engineering argument.
• ▶ 25:37 Doomer logic rests on the assumption that human intelligence is fully computational, so enough data and compute would inevitably recreate it.
• ▶ 26:03 Doomers argue AI’s perfect digital communication would let it improve far faster than humans, who are limited by inefficient communication.
• ▶ 26:26 Doomers believe no species rules over a superior one, so AI would evolve into a higher species, dominate humans, and possibly eliminate us.
• ▶ 26:55 Karen Hao says no country is yet pioneering democratic control of AI; the EU AI Act is a major attempt but only a risk/rights-based framework for deployment.
• ▶ 27:25 Democratic AI governance is missing the key piece of democratically developing AI, including control over the supply chain of data, land, energy, water, and access to public spaces like schools and hospitals.
• ▶ 28:15 Hao counters resignation about data privacy: “That data is yours”—arguing that collectively owned resources and people’s intellectual property belong to the public, not Silicon Valley.
• ▶ 28:20 Communities must treat land, schools, and hospitals as sites of democratic contestation over AI—not neutral spaces but arenas for negotiation.
• ▶ 28:39 The Chilean water activists model shows the needed attitude: aggressively treating shared resources as “ours” and refusing to give them up without mutual agreement.
• ▶ 28:53 Adopting this protective spirit toward data, land, water, and schools strips companies of leverage, forcing them to change how they build and deploy AI.
• ▶ 29:10 Te Hiku Media, a nonprofit Māori radio station, offers a counterexample to Silicon Valley's "scale at all costs" AI approach by using AI for Māori language revival from archival audio.
• ▶ 30:55 Their community-centered process involved first asking the community if they wanted the tool, then obtaining full consent for data donation—collecting only a few hundred hours of highly curated audio, which proved enough to build a performant speech recognition model.
• ▶ 31:52 Hao advocates for a future of small, task-specific AI models trained on curated data with low computational power, open-sourced for community benefit and applied to challenges like climate change and healthcare.
• ▶ 32:37 Amy Goodman signals the interview is wrapping up and asks Karen Hao the closing question.
• ▶ 32:40 Goodman asks what most shocked Karen Hao during her years of AI journalism and research, specifically for Empire of AI.
• ▶ 32:46 The section cuts off mid-question, so Karen’s answer is not included in this segment.
• ▶ 32:49 Karen Hao initially planned Empire of AI around "vertical harms" like labor exploitation and environmental damage in the AI supply chain.
• ▶ 33:07 Reporting revealed a deeper "horizontal harm": across communities—artists, Chilean water activists—the shared experience was a complete loss of agency to self-determine their future.
• ▶ 33:33 The core thesis became that AI threatens democracy, because if most of the world loses self-determination, democracy cannot survive—tied to Silicon Valley's "scale at all costs" approach.
• ▶ 34:13 Artists pushed back by realizing they could sue AI companies, a legal strategy for resistance.
• ▶ 34:25 Kenyan workers contracted by OpenAI chose to unionize and escalate their story to international media, combining collective organizing with media amplification.
• ▶ 35:00 The first step to reclaiming democracy is remembering that no one can take your agency away, even for the most vulnerable communities.
• ▶ 35:03 Karen Hao’s closing message emphasizes personal agency: "no one can take your agency away."
• ▶ 35:06 Amy Goodman reintroduces Karen Hao and her new book Empire of AI: Dreams and Nightmares in Sam Altman's OpenAI, directing viewers to democracynow.org for the full interview.
• ▶ 35:19 The special broadcast signs off with thanks, followed by standard Democracy Now! promotions for YouTube subscriptions, the website, app, newsletter, and daily podcast.

• ▶ 0:00 Introduction and AI's Job Impact (0:00 - 1:24) - Introduction, AI's immediate impact on jobs, and OpenAI's definition of AGI.
• ▶ 1:24 Labor, Technology, and AI Harms (1:24 - 3:29) - Labor-automating vs. assistive tech, and AI companies' overclaims and user harms.
• ▶ 3:29 Sam Altman and OpenAI's Origins (3:29 - 5:38) - Altman's profit/prophet image, OpenAI's non-profit origins, and founding mission.
• ▶ 5:40 Reporting on OpenAI and AGI as Religion (5:40 - 8:18) - Challenges in reporting on OpenAI, its 3-year blackout, and boomers vs. doomers.
• ▶ 8:19 Stargate, Politics, and Global Ambitions (8:19 - 13:01) - Stargate project, Altman's maneuvering, infrastructure deals, and "OpenAI for countries."
• ▶ 13:01 Energy, Environment, and Congressional Testimony (13:01 - 16:14) - AI's energy consumption, nuclear/solar debates, and Altman's Senate testimony.
• ▶ 16:14 US-China AI Race and Talent Alienation (16:14 - 23:48) - Export controls, DeepSeek, attacks on academia, brain drain, and historical parallels.
• ▶ 23:48 AI Futures: Doomers, Democracy, and Resistance (23:48 - 35:46) - AI apocalypse vs. utopia, democratic control, Te Hiku Media, and concluding reflections.