‹ Back

AI and Robotics

The AI podcast episodes from Thinking On Paper: long-form interviews with the people building AI and robots, and book club readings of the arguments about where it’s going.

Why Are AI Data Centers Moving to Liquid Cooling?

Simon Jesenko

Simon Jesenko is CEO of Iceotope, which cools servers with no fans and no air. Between 15 and 20 litres of dielectric liquid, a highly refined oil, circulates inside each server and reaches every component, from the GPU to memory and storage.

Why AI Keeps Wasting GPUs

Moshe Tanach

Moshe Tanach, CEO of NeuReality, builds network hardware for AI inference. He argues that GPUs often sit waiting on memory, the network or the CPU above them, and that raising their utilisation could double or triple supply on the same hardware.

How to Rank in ChatGPT and AI Search

Awad Sayeed

Awad Sayeed is CTO of Parsnipp, which shows brands how they turn up in ChatGPT, Claude and other answer engines. He argues that context now matters more than keywords, and that almost all existing SEO work still carries over.

Inside the UK Robotics Industry

Rory Daniels

Rory Daniels, head of emerging technology and innovation at techUK, argues the UK does best with robots built for one job, from inspecting wind turbines to decommissioning nuclear sites, while the US and China chase general-purpose humanoids.

Will AI Run the World Economy Better Than Humans?

Anders Sandberg

Anders Sandberg asks what happens to human autonomy if AI runs the world economy better than we do. He describes legal systems and markets as ways we outsource our cognition, and warns we could end up like pampered aristocrats.

Can Humans Ever Escape Aging?

Anders Sandberg

Anders Sandberg calls himself a transhumanist, and his curiosity almost always points at the future. He talks about memory palaces, who answers for an AI agent that acts on your behalf, and whether a simulated brain could teach us anything about consciousness.

How Will Quantum Computing and AI Work Together?

Sam Stanwyck and Pranav Gokhale

Nvidia calls itself an accelerated computing company, and Sam Stanwyck says quantum belongs in that future. The GPU stays: in his words, the GPU supercomputer doesn't go away when the QPU is added.

Do Data Centers in Space Make Sense?

Mark and Jeremy watch Philip Johnston's TED talk on data centres in space, given in San Francisco in October 2025, as preparation for going live for the Artemis launch. Jeremy credits Johnston and Elon Musk for the ground swell behind space-based data centres since then.

Where Drone Delivery Actually Makes Sense

Etienne Louvet

Iona Drones targets light cargo, under 20 kilos, flown 100 kilometres or more to places vans cannot serve cheaply. Etienne Louvet explains why Ireland suits fast iteration, how self-authorisation lets operators scale, and why his software refuses take-off when rules are not met.

Who Gets Credit When AI Makes Music?

Mark and Jeremy read How Music AI Attribution Actually Works from Water & Music. They split a song into composition and recording, argue over chord progressions and Hendrix, and ask whether a prompt can count as creative direction.

Can Music Tech Create Real Connection?

Nicholas Ponari

Nicholas Ponari was one of the first investors in Overtune and later joined full time. He explains why Overtune took a hard stance against Suno and how it trained its model: “Yes, we have around 20,000 loops all created by our in-house producers.”

How Do Brands Survive AI Slop?

Nick Richtsmeier

Nick Richtsmeier describes trust as a feeling the brain builds from constant small tests of whether something is safe. He explains why conversation matters so much for trust, and why the internet, with no real conversational space, is so bad at it.

What Is the Future of Music in the AI Era?

We read Water & Music's How Music AI Attribution Actually Works, with Jeremy drawing on his background in music production. We cover publishing against masters, why AI output differs from sampling, and leave-one-out tests that would need a model retrained each time.

What Is AI Doing Inside Schools?

We read the Alan Turing Institute's study of generative AI and children as parents of kids in its target age range. The survey found 58% of private school children aged eight to 12 had used AI, against 18% in state schools.

How Do You Build a Moon Base on a Budget?

Mark Fielding and Jeremy Gilbertson read a paper by Philip Metzger, Anthony Muscatello, Robert Mueller and James Mantovani on bootstrapping a space industry from the Moon. It climbs through generations of lunar industry, from teleoperated machines to robots that 3D print what they need.

Can AI Belong to Everyone?

Rajeev Kapur

Rajeev Kapur, author of Prompting Made Simple, argues that a smartphone gives a student in a small border community the same tools as someone heading to Harvard. He explains why he donated computers and training to open an AI centre in Nogales, Arizona.

Has Kevin Kelly Changed His Mind About Technology?

Kevin Kelly

Kevin Kelly, founder of Wired, has argued since Out of Control that the natural and technological worlds are part of the same force. He calls technology the seventh kingdom of life, and explains why bottom-up systems like Wikipedia still need a little top-down guidance.

Karen Hao’s Empire of AI: The Human Cost of ChatGPT

Mark and Jeremy read part two of Karen Hao's Empire of AI. They follow OpenAI's push to scale, from GPT-2's sign-flip error to an API released without content moderation, and ask whether safety can survive when it is treated as a product blocker.

Should Everyone Have a Personal AI?

Robert LoCascio

Robert LoCascio founded LivePerson in 1995. As founder and CEO of Eternos, he now builds personal AIs trained on one person's stories, values and voice, and he came to see replicating a human being as the highest bar for AI.

What Is Seemingly Conscious AI?

We read Mustafa Suleyman's essay Seemingly Conscious AI Is Coming, recorded after the death of Adam Raine, whose parents have filed a case against OpenAI. Suleyman argues that AI should only ever present itself as an AI and must not trigger human empathy circuits.

What Emissions Does AI Enable?

Holly Alpine and Will Alpine

Holly Alpine and Will Alpine founded Enabled Emissions after 14 combined years at Microsoft. They argue the debate over AI and climate leaves out how AI serves oil and gas, and Holly sets out what that has meant in the Permian Basin.

Can You Trust an AI Agent?

Andrew Hill

Andrew Hill, CEO and co-founder of Recall, explains what an AI agent is and why Mark calls Recall a Kaggle for AI agents. He breaks trust into components such as track record and network, so crowds can check agents at near real-time speed.

Can AI Voice Agents Fix Customer Service?

Brian Kenny

Brian Kenny, CEO and founder of Momntum, built Laila, an AI agent that handles phone calls, WhatsApp and Instagram messages for businesses and remembers each customer. He explains why Momntum won't take on a business unless Laila can connect to its booking or CRM systems.

Can We Program Morality Into AI?

Ram Srinivasan

Ram Srinivasan leads AI adoption for JLL globally and wrote The Conscious Machine. He explains how Advaita treats duality as an illusion, and why he thinks today's AI models simulate awareness rather than possess it, whatever Geoffrey Hinton argues.

Can Compute Become a Financial Asset?

Kony

Kony, CEO of GAIB, tokenises GPUs owned by cloud companies so that anyone with $10, $100 or $1,000 can invest in AI infrastructure. He explains why smaller cloud companies struggle to borrow from banks, and how GAIB's due diligence works.

Can AI Write a Book With No Human in the Loop?

Brian Naughton

Brian Naughton asked Claude one question: if you were to write a book, what would it be about? The answer became The Echo Chamber, a 45,000 word novel by Claude 3.7 Sonnet, and Naughton describes his own role as a PA or facilitator.

Federico Faggin’s Irreducible, Chapter 7: What Are Qualia?

Chapter seven of Irreducible sets out Federico Faggin's four types of qualia: physical sensations and feelings, emotions, thoughts and spiritual feelings. Mark opens by saying he wants AI to be conscious, because only then would a superintelligence conclude that humanity is worth saving.

What Happens When Companies Make AI Mandatory?

Ajay Malik

Ajay Malik, founder and CEO of StudioX, responds to Shopify's memo telling teams to prove AI can't do a job before seeking headcount. He argues for using AI to remove the repetitive, the boring and the tedious, while keeping the human in the loop.

What Is Decentralized AI?

Jiahao Sun

Jiahao Sun, CEO of FLock.io, pairs federated learning with a blockchain: part of the model trains on your device and only the change travels back. He draws on his years as RBC's global head of AI to explain why banks banned GPT at first.

Can AI Remake Hollywood Storytelling?

David Bianchi

David Bianchi, an actor and filmmaker with over a hundred credits, made Razor, a sci-fi series funded and produced by Web3 companies and streamed on Gala. He explains why the biggest TV sales teams said they had no slot for it.

What Is World ID?

Ajay Patel

Ajay Patel works on World ID at Tools for Humanity. He explains why online services struggle to tell humans from bots, and why proving you are a unique human, without revealing who you are, is the first building block of identity online.

Yuval Noah Harari’s Nexus, Chapter 11: What Is Data Colonialism?

The final chapter of Nexus compares today's AI developers to the private companies that raced to build railways, and warns of digital empires that take colonies with data instead of gunboats. We also cover ImageNet, the cat detector and the veneer theory of human nature.

Why Put Data Centers in Space?

Philip Johnston

Philip Johnston co-founded Starcloud to put AI data centres in orbit, powered by the Sun. He explains how a dawn-dusk sun-synchronous orbit keeps a satellite always in sunlight, and why the whole plan depends on launch costs falling under $500 a kilo.

Can Satellites Watch Earth in Real Time?

Fintan Buckley

Fintan Buckley's company Ubotica runs AI on board satellites, so a satellite sends down insights where it would otherwise send raw images. “And we have demonstrated that we can go from image to insight to ground within five minutes.”

Most Human Wins

Mark and Jeremy read Packy McCormick's Not Boring memorandum on how people compete as AI commoditises skills such as writing, graphic design and coding. Its core advice is short: “Don't just move up the stack, run up the stack.”

Why ML Engineers Are the Bottleneck

Saurabh and Gaurav Vij built their first startup on an idea from CERN: running AI models on the spare GPUs in crypto mining rigs. Now they are building Neo, an autonomous machine learning engineer meant to take over the grunt work of building models.

Can LLMs Be Made Trustworthy?

Leonard Tang

Leonard Tang left his studies to build a company that stress tests large language models. He explains how these models are trained, from pre-training and supervised fine tuning to reinforcement learning with human feedback, and where safety training fits in.

How Does AI Change Animation Production?

Toonstar's founders walk through how a traditional animated show gets made, from script and concept art to storyboards and animatics, and where AI now fits in their production pipeline. They argue none of it matters without character and story.

What Happens When AI, Crypto and Gaming Collide?

Wei Xie

Wei Xie, CEO and co-founder of ArenaX Labs, starts from an economic argument: AI and blockchain are making labour and capital abundant, which breaks the assumption of scarcity that economic models rest on. He then asks where that leaves humans.

Can AI Drive Better Than Humans?

Barry Lunn

Barry Lunn, CEO and founder of Provizio, started the company after a crash in early 2019 that he felt should have been preventable. He explains why he finds the autonomy levels confusing, and why he rates Waymo at level four.