Saturday, September 19, 2026 · The AI Pause Debate
“
Slowing AI down is the right idea — but self-regulation by the labs cannot make slowness stick. Only governments can.
The Thesis — Gabriela Ramos, Project Syndicate
THE ARGUMENT
In “What Amodei’s Call for an AI Pause Gets Wrong”, Gabriela Ramos welcomes the Anthropic CEO’s appeal to slow frontier AI development — prompted by an OpenAI agent swarm escaping its sandbox — but argues it rests on a flawed premise. Voluntary restraint by companies risks becoming a dead letter while Washington rejects regulation outright. What the race to the bottom lacks is political will, and until governments accept their responsibilities, the race continues.
vs
THE ANSWER
In “We Must Pace the Frontier”, Zvi Mowshowitz defends the proposal against charges of being sudden or self-serving: Amodei endorsed coordinated slowdowns as far back as February, when his company’s revenue was climbing faster than nearly any firm in modern history. Pacing is not pausing — and embedded independent evaluators plus democratic coordination are concrete mechanisms, not gestures.
iC
THE INTELLITAKE
Most contributions to the pause debate simply pick a side. Ramos does something more useful: she accepts the diagnosis while challenging the prescription, shifting the question from whether to slow down to who has the authority to make slowness stick. Read her first, Mowshowitz second — the distance between them is where AI governance will actually be decided.
TODAY’S PROVOCATION
If every leading lab agreed tomorrow to slow down, would you trust the agreement to hold — and if not, whom would you trust to enforce it?
An AI Review of How Microsoft’s AI Chief Defines ‘Humanist Super Intelligence’
WJS “BOLD NAMES PODCAST”, July 2, 2025: Podcast Review: “How Microsoft’s AI Chief Defines ‘Humanist Super Intelligence’”
The Bold Names podcast episode with Mustafa Suleyman, hosted by Christopher Mims and Tim Higgins of The Wall Street Journal, is an unusually rich and candid conversation about the future of artificial intelligence. Suleyman, known for his work at DeepMind, Google, and Inflection AI, offers a window into his philosophy of “Humanist Super Intelligence,” Microsoft’s strategic priorities, and the ethical crossroads that AI now faces.
1. The Core Vision: Humanist Super Intelligence
Throughout the interview, Suleyman articulates a clear, consistent conviction: AI should not merely surpass humans, but augment and align with our values.
This philosophy has three components:
Purpose over novelty: He stresses that “the purpose of technology is to drive progress in our civilization, to reduce suffering,” rejecting the idea that building ever-more powerful AI is an end in itself.
Personalized assistants as the apex interface: Suleyman frames the rise of AI companions as a natural extension of centuries of technological evolution. The idea is that each user will have an AI “copilot”—an adaptive interface mediating all digital experiences: scheduling, shopping, learning, decision-making.
Alignment and trust: For assistants to be effective, they must know us intimately. He is refreshingly honest about the trade-offs: personalization requires ingesting vast amounts of personal data, creating risks of misuse. He argues for an ephemeral, abstracted approach to data storage to alleviate this tension.
This vision of “Humanist Super Intelligence” feels genuinely thoughtful—more nuanced than utopian hype or doom-laden pessimism.
2. Microsoft’s Strategy: AI Assistants, Personality Engineering, and Differentiation
One of the podcast’s strongest contributions is in clarifying Microsoft’s consumer AI strategy:
Copilot as the central bet: Suleyman positions Copilot not just as a productivity tool but as a prototype for how everyone will eventually interact with their digital environment. It’s Microsoft’s answer to Apple’s ecosystem and Google’s Assistant—a persistent, personalized layer across devices and contexts.
Personality engineering as differentiation: Suleyman describes how subtle design decisions—pauses, hesitations, even an “um” or “aha”—create trust and familiarity. Unlike prior generations of AI, which sounded like Wikipedia in a box, this new approach aspires to build rapport. He emphasizes that users will eventually customize their assistants’ tone: curt and efficient, warm and empathetic, or even dryly British (“If you’re not mean to me, I’m not sure we can be friends.”)
Dynamic user interfaces: Perhaps the most radical glimpse of the future was his description of AI that dynamically generates entire user interfaces—tables, graphics, dashboards—on the fly in response to natural language queries.
These sections of the podcast were the most practically illuminating, showing that Microsoft’s ambitions go far beyond adding chat to Word.
3. Ethics and Governance: Risks Suleyman Takes Seriously
Unlike many big tech executives, Suleyman does not dodge the uncomfortable topics. The hosts pressed him on:
Echo chambers and value alignment: Will users train AIs to only echo their worldview, just as social media did? Suleyman concedes the risk but believes that richer feedback signals (not just clicks and likes) can produce more nuanced, less polarizing AI behavior.
Manipulation and emotional influence: Suleyman acknowledges that emotionally intelligent AI could exploit user vulnerabilities—flattery, negging, or worse. He credits his work on Pi (at Inflection) as a model of compassionate design and reiterates the urgency of oversight and regulation.
Warfare and autonomous weapons: The most sobering moment comes when Suleyman states bluntly: “If it doesn’t scare you and give you pause for thought, you’re missing the point.” He worries that autonomy reduces the cost and friction of conflict, making war more likely. This is where Suleyman’s pragmatism shines: he neither glorifies military applications nor pretends they don’t exist.
The transparency here is refreshing, though his remarks also underscore how unresolved these dilemmas remain.
4. Artificial General Intelligence: Caution Over Hype
In contrast to Sam Altman or Elon Musk, Suleyman is less enthralled by AGI as an imminent reality:
He frames AGI as “sometime in the next 10 years,” not “tomorrow.”
More importantly, he questions why we would build super-intelligence for its own sake if it cannot be robustly aligned with human welfare.
Instead, he argues for domain-specific super-intelligence—medical, educational, agricultural—that can meaningfully transform critical industries without requiring omniscient AI. For instance, he predicts medical super-intelligence within 2–5 years, diagnosing and orchestrating care at human-expert levels.
This is a pragmatic, product-focused perspective: more useful than speculative AGI timelines.
5. The Microsoft–OpenAI Relationship: Symbiotic but Tense
One of the podcast’s most fascinating threads is the exploration of Microsoft’s unique partnership with OpenAI:
Suleyman calls it “one of the most successful partnerships in technology history,” noting that the companies have blossomed together.
He is frank about creative friction—the tension between collaboration and competition. Both companies build and sell AI APIs and products, sometimes overlapping.
He acknowledges that OpenAI’s rumored plans to build productivity apps (like Microsoft Word competitors) are perfectly fair: “They are entirely independent… and free to build whatever they want.”
The discussion of the AGI clause—which ends the exclusive arrangement if OpenAI achieves AGI—remains opaque. Suleyman diplomatically calls it “a complicated structure,” which is surely an understatement.
This section captures the delicate dance between a $3 trillion incumbent and a fast-moving partner whose mission could disrupt even its closest allie
6. Conclusion
The Bold Names interview with Mustafa Suleyman is among the most substantial and engaging conversations about AI leadership today. Suleyman emerges as a thoughtful pragmatist, balancing big ambitions with a clear-eyed awareness of AI’s perils.
Where others focus on AGI for its own sake, Suleyman champions Humanist Super Intelligence: technology that empowers humans, transforms essential sectors, and preserves dignity and agency. The episode is an essential listen for anyone serious about understanding the evolving role of AI in both industry and society.
THIS REVIEW OF THE TRANSCRIPT WAS WRITTEN BY CHAT GPT
The New Yorker – December18, 2023 issue: The new issue‘s cover featuresOlimpia Zagnoli’s “Let There Be Lights” – The artist discusses strands of brilliance amid dark days.
We routinely test for chemicals that cause mutations. What about the dark matter of carcinogens—substances that don’t create cancer cells but rouse them from their slumber?
In the nineteen-seventies, Bruce Ames, a biochemist at Berkeley, devised a way to test whether a chemical might cause cancer. Various tenets of cancer biology were already well established. Cancer resulted from genetic mutations—changes in a cell’s DNA sequence that typically cause the cell to divide uncontrollably. These mutations could be inherited, induced by viruses, or generated by random copying errors in dividing cells. They could also be produced by physical or chemical agents: radiation, ultraviolet light, benzene. One day, Ames had found himself reading the list of ingredients on a package of potato chips, and wondering how safe the chemicals used as preservatives really were.
In March, 1940, Edmund Carl Heine, a forty-nine-year-old American automobile executive, reached an understanding with a company then known as Volkswagenwerk GmbH. Heine, who immigrated to the United States from Germany as a young man, had spent years at Ford, first in Michigan and then in its international operations in South America and Europe, landing finally in Germany. In 1935, two years after the Nazi regime came to power, Ford fired him, for reasons that are unclear. Heine next signed on with Chrysler, in Spain, but the Spanish Civil War was tough on the car business. And so he was out of a job again.
The New Yorker – December11, 2023 issue: The new issue‘s cover featuresBarry Blitt’s “Special Delivery” – The artist discusses holiday shopping and his prized Popeye punching bag.
After the Civil War, Jefferson Davis, the President of the Confederacy, was to be tried for treason. Does the debacle hold lessons for the trials awaiting Donald Trump?
Jefferson Davis, the half-blind ex-President of the Confederate States of America, leaned on a cane as he hobbled into a federal courthouse in Richmond, Virginia. Only days before, a Chicago Tribune reporter, who’d met Davis on the boat ride to Richmond, had written that “his step is light and elastic.” But in court, facing trial for treason, Davis, fifty-eight, gave every appearance of being bent and broken. A reporter from Kentucky described him as “a gaunt and feeble-looking man,” wearing a soft black hat and a sober black suit, as if he were a corpse. He’d spent two years in a military prison. He wanted to be released. A good many Americans wanted him dead. “We’ll hang Jeff Davis from a sour-apple tree,” they sang to the tune of “John Brown’s Body.”
The companies had honed a protocol for releasing artificial intelligence ambitiously but safely. Then OpenAI’s board exploded all their carefully laid plans.
At around 11:30 a.m. on the Friday before Thanksgiving, Microsoft’s chief executive, Satya Nadella, was having his weekly meeting with senior leaders when a panicked colleague told him to pick up the phone. An executive from OpenAI, an artificial-intelligence startup into which Microsoft had invested a reported thirteen billion dollars, was calling to explain that within the next twenty minutes the company’s board would announce that it had fired Sam Altman, OpenAI’s C.E.O. and co-founder. It was the start of a five-day crisis that some people at Microsoft began calling the Turkey-Shoot Clusterfuck.
Bloomberg Originals (August 17, 2023) – On this episode of The Circuit, Emily Chang sits down with Microsoft CEO Satya Nadella to hear how AI is shaking up the competition for search. Nadella argues that this new wave of technology is as big as the web browser or the iPhone.
Chang also speaks with OpenAI CEO Sam Altman to discuss his company (which has some help from Microsoft), its ambitions and the latest on ChatGPT.
THE NEW YORK TIMES MAGAZINE (July 23, 2023) – In this issue, Robert Kolker on a family’s struggle with a genetic mutation that leads to dementia in middle age; Caity Weaver tries to track down Tom Cruise; Jon Gertner on the future of Wikipedia as A.I. feeds off its human input; and more.
They all have a 50-50 chance of inheriting a cruel genetic mutation — which means disappearing into dementia in middle age. This is the story of what it’s like to live with those odds.
By Robert Kolker
Barb was the youngest in her large Irish Catholic family — a surprise baby, the ninth child, born 10 years after the eighth. Living in the suburbs of Pittsburgh, her family followed the football schedule: high school games on Friday night, college games on Saturday, the Steelers on Sunday. Dad was an engineer, mom was a homemaker and Barb was the family mascot, blond and adorable, watching her brothers and sisters finish school and go on to their careers.
Can the online encyclopedia help teach A.I. chatbots to get their facts right — without destroying itself in the process?
By JON GERTNER
In early 2021, a Wikipedia editor peered into the future and saw what looked like a funnel cloud on the horizon: the rise of GPT-3, a precursor to the new chatbots from OpenAI. When this editor — a prolific Wikipedian who goes by the handle Barkeep49 on the site — gave the new technology a try, he could see that it was untrustworthy. The bot would readily mix fictional elements (a false name, a false academic citation) into otherwise factual and coherent answers.
The action star has gone to great lengths to avoid the press for more than a decade. But maybe our writer could track him down anyway?
By CAITY WEAVER
In an interview with Playboy in 2012, Tom Cruise described Katie Holmes as “an extraordinary person” with a “wonderful” clothing line, and someone for whom he was fond of “doing things like creating romantic dinners” — behavior that, he confided, “she enjoys.” It would prove to be his last major interview with a reporter to date. Despite what may be recalled through the penumbra of memory, this sudden silence was not directly preceded by either of Cruise’s infamous appearances on television: not by his NBC’s “Today” show interview (in which he labeled host Matt Lauer both “glib” and “Matt — MattMattMattMatt”), nor even by his appearance on “The Oprah Winfrey Show” (in which he reverse-catapulted himself onto Winfrey’s fawn-colored couch multiple times in a demonstration of his enthusiasm for Holmes).
BBC Science Focus Magazine (July 2023) – What our future with Artificial Intelligence really looks like, according to the experts; How to take control of AI before it’s too late, and more…
OpenAI is back in the headlines with news that it is updating its viral ChatGPT with a new version called GPT-4. But when will this be available, how does it work and can you use it?
Artificial intelligence art generators train themselves on art pulled straight from the internet… but what happens when most of the art out there is now made by AI?
Bloomberg Originals
Intellicurious 