Thinking On Paper

Carissa Véliz, Oxford philosopher and author of Prophecy, Thinks On Paper about why predictions are never facts, why tech CEOs predict a future they want you to buy, and how to take back ownership of your own life.

When Sam Altman, Dario Amodei, and Jensen Huang tell us what AI will do to jobs, society, and the future, are they making predictions, or selling products? Carissa Véliz argues it's the latter. 

In this conversation we dig into self-fulfilling prophecies, the illusion of safety that algorithms sell us, what Seinfeld teaches us about predictive algorithms, and why Epicurus might be a better guide to the AI age than the Stoics. We ask if Polymarket should be banned, is democracy at risk and wonder how books can save us. 

Carissa Véliz is an Associate Professor at the Institute for Ethics in AI at the University of Oxford. Her new book Prophecy is out now.
--
📺 Watch On YouTube:
🎧 Listen to every podcast⁠
📺 Follow us on ⁠Instagram⁠
🏠 Follow us on ⁠X⁠
🏠 Follow Jeremy on ⁠LinkedIn⁠

To suggest guests or sponsor the show, please email: [email protected]

--

CHAPTERS  

(00:00) Intro
(01:00) What is the good life? 
(02:00) Why knowing yourself matters more than strategy 
(04:44) The analog world vs the digital world 
(06:45) How prophecies exploit our need for security 
(08:47) Why ancient Rome banned predicting the emperor's death 
(10:11) The illusion of safety that AI sells us 
(12:27) When predictions work, and when they don't 
(15:00) Altman, Amodei, Huang: predictions or sales pitches? 
(28:29) How to resist prophecies as a busy person 
(29:53) Prediction markets, Polymarket, and democracy 
(31:49) TikTok, algorithms, and the Molly Russell case 
(36:08) "Engagement algorithms are cocaine in food" 
(40:54) Self-fulfilling prophecies as the perfect crime 
(43:44) Why comedy is the enemy of prophecy 
(46:59) What Seinfeld teaches us about predictive algorithms 
(52:16) Karikó and the Nobel Prize we almost missed 
(53:40) Increase your serendipity 
(56:13) Why Epicurus beats the Stoics

The AI meme war between the US and Iran is evolving. A few weeks ago, AI-generated Lego propaganda videos were a curiosity. Sometimes funny, often violent, always troublesome and never diplomatic, they quickly gained millions of views across social media. The White House Twitter (X) account was responsible for the US videos.

An Iranian media company called Explosive Media, the Iranian.America - either put off by the global consensus that it was losing the war (if you want to see their earlier Lego, Grand Theft Auto, Wii Sports and Call of Duty efforts, watch this: https://www.youtube.com/watch?v=R58dlaNaGIE&t=14s) - or bored, switched their AI models to tax season.

Iran, losing the guns and missiles part of proceedings, would change tact. Explosive Media turned up the heat. And was duly banned from YouTube.Which could of unleashed the beast. Now Iranian embassies are posting them on Twitter (X) and US creators are using the same format to mock whatever's in the news cycle - and whatever is on their mind. And a new Lego-style AI propaganda video shows Pete Hegseth getting drunk as a pirate, for example.

The Iranian embassy in South Africa posted an AI-generated clip of Trump as a French Europop keyboardist singing about blockading the Strait of Hormuz. The Iranian Embassy in Zimbabwe posted a video of Trump addressing a briefing room of US soldiers. When the soldiers leave, Trump takes off his mask revealing the face of Netanyahu. All fun and games.

The AI gurus, especially those who have been telling anyone who will listen Hollywood is doomed, celebrate the democratization of the AI tools, not a bubble of irony in their LinkedIn posts... Because this is really the Fake news story with its own mask on.

And where does this go when the French, US, German, Mexican, Argentinian or Italian general elections all hit over the next two years?

What happens to AI tools and democracy when every North Korean and Russian bot farm is an army of highly trained, skilled and determined video prompters? And what happens when its not Lego figures, but hyper realistic videos of politicians doing and saying things they haven't done or said?
--
🎧 Listen to every podcast⁠
📺 Follow us on ⁠Instagram⁠
🏠 Follow us on ⁠X⁠
🏠 Follow Jeremy on ⁠LinkedIn⁠

To suggest guests or sponsor the show, please email: [email protected]

--

--
TIMESTAMPS

(00:00) Explosive Media
(00:38) US Bowling Iran
(01:52) Trump's Mask
(03:20) Blockade, Blockade
(06:28) Drunken Hegseth
(08:00) Truth

Thinking On Paper ask who owns the Moon, whether the Outer Space Treaty still holds, why 80% of space investment is defence, what the Wolf Amendment did to US-China relations in orbit, whether asteroid mining legislation was written by the companies who'd profit from it, and why the book Space to Grow ends on a warning about war, not wonder.
 
Watch On YouTube: https://youtu.be/MVBxjZCGfxY

--
The space race between the US and China has never been about exploration. 80% of current space investment is defence-related. In the final episode of the Thinking on Paper Book Club series on Space to Grow by Matthew Weinzierl and Brendan Rosseau, Mark and Jeremy trace the politics of space ownership, space law, and who owns the Moon.

From the 1967 Outer Space Treaty through Planetary Resources and asteroid mining legislation, the Wolf Amendment that banned NASA from collaborating with China, China building its own space station in response, Trump designating space as a theater of war in 2018, and the Rumsfeld Commission warning of a "space Pearl Harbor." 

The episode covers John Locke's labour theory of property and how it applies to mining asteroids worth 100,000 times global GDP, the prisoner's dilemma between the US and China, why there would be no space industry without national security, and Neil deGrasse Tyson on the absurdity of claiming ownership over atoms forged in dying stars. 59 active conflicts currently on Earth as of the 2025 Global Peace Index. 

The book's final chapter asks whether humanity can reach space without turning it into another battlefield. 

--
Chapters

(00:00) Global Conflict and Space Resources
(02:04) Human Nature and Space Exploration
(03:28) The Economics of Asteroid Mining
(05:53) Legal Frameworks for Space Mining
(11:05) The Space Resource Exploration Act
(13:01) International Reactions to Space Mining Legislation
(17:19) Philosophical Perspectives on Space Ownership
(20:14) The Role of National Security in Space
(20:40) The Role of Government in Space Innovation
(21:34) National Security and the Space Industry
(23:10) Weaponization of Space: A New Era
(24:47) The Prisoner's Dilemma in Space Cooperation
(26:40) Humanity's Moral Compass in Space Exploration
(27:03) The Future of Humanity in Space

Space technology investment is surging into 2026. Confidence is high. The ten largest funding rounds total over $3.7 billion. And it's only April.

Defence and national security contracts are driving much of the momentum, with companies like Stoke Space, Sierra Space and Cesium Astro attracting hundreds of millions on the strength of government partnerships.

Infrastructure remains the dominant investment thesis — from encrypted GPS alternatives and space-based weather platforms to satellite communications and reusable launch vehicles — reflecting a market that is still building the foundational layer needed for commercial space to scale.

The biggest surprise sits at the top of the list: Beijing-based iSpace China claimed the single largest raise at $729 million, confirming the US V China space race is very much happening.  

Meanwhile, three companies have crossed the unicorn threshold — StarCloud, Tomorrow.io and Sierra Space, the latter commanding a confirmed $8 billion valuation. 

Human spaceflight and space station ambition round out the upper tier, with Vast Space and Axiom Space collectively raising $850 million to build the commercial space stations and crew infrastructure that will replace the ISS when it retires later this decade.

The Top 10 In Full

iSpace China — $729M
Sierra Space — $550M (Series C)
Vast Space — $500M (Series A)
Cesium Astro — $470M
Axiom Space — $350M
Stoke Space — $350M
PLD Space — €210M (Series C)
Tomorrow.io — $175M
Xona Space — $170M (Series C)
StarCloud — $170M

--

🎧 Listen to every podcast⁠
📺 Follow us on ⁠Instagram⁠
🏠 Follow us on ⁠X⁠
🏠 Follow Jeremy on ⁠LinkedIn⁠

To suggest guests or sponsor the show, please email: [email protected]

Chapters
(00:00) Starcloud
(00:52) Xona Space
(03:27) Tomorrow IO
(06:01) PLD Space
(08:00) Stoke Space
(10:18) Axiom Space
(12:29) Cesium Astro
(14:50) VAST Space
(19:02) Sierra Space
(21:47) I-Space (Beijing Interstellar Glory Space Technology Ltd.)

The Martens Clause, a legal principle drafted by Russian-Imperial diplomat Fyodor Martens during the first Hague Peace Conference of 1899, established that even in the absence of specific written law, nations and individuals remain bound by "the laws of humanity and the requirements of public conscience." 

Originally conceived as a compromise to prevent the collapse of early international humanitarian law negotiations - when smaller nations like Belgium objected to how occupying powers classified resistance fighters - the clause became a foundational backstop in international law. 

It was subsequently invoked in some of the most consequential legal proceedings of the twentieth century, including the Nuremberg Trials of 1945-46 to counter arguments that prosecuting Nazi war crimes constituted retroactive legislation, the 1949 Corfu Channel case where Albania was held responsible for failing to warn shipping of mines in its territorial waters, and the 1986 ICJ ruling against the United States for mining Nicaraguan harbors and supporting the Contra insurgency.

Mark & Jeremy from Thinking On Paper are now asking whether this 127-year-old principle could serve as what some are calling a "minimum viable architecture" for governing emerging technologies — particularly artificial intelligence, commercial space operations, and quantum computing — where the pace of innovation vastly outstrips the speed of regulation. 

Jeremy argues that the clause's core logic — that something not being explicitly prohibited does not make it automatically permitted — could provide a much-needed ethical and legal floor beneath industries currently operating in regulatory grey zones, from AI training on copyrighted data to autonomous weapons systems and asteroid mining rights. 

Mark counters that the clause has historically only been applied retroactively to clear moral atrocities, and that its deliberately vague language, while effective at building diplomatic consensus, lacks the specificity needed to adjudicate the morally ambiguous questions at the frontier of technology, such as algorithmic bias, AI decision-making opacity, and the concentration of technical power among a small number of corporations and nation-states.

Please enjoy the show.

--

🎧 Listen to every podcast⁠
📺 Follow us on ⁠Instagram⁠
🏠 Follow us on ⁠X⁠
🏠 Follow Jeremy on ⁠LinkedIn⁠

To suggest guests or sponsor the show, please email: [email protected]

--

(00:00) The First Peace Conference: A Historical Perspective
(07:37) The Martin's Clause: Implications for Modern Governance
(10:05) Space Tech and the Outer Space Treaty
(13:58) AI and the Need for Ethical Frameworks
(17:21) Accountability in Technology Deployment
(22:56) The Future of Humanity: Collaboration vs. Competition