Discover
Progress by Fifty Years
Progress by Fifty Years
Author: Progress
Subscribed: 2Played: 25Subscribe
Share
© Progress
Description
11 Episodes
Reverse
Michael Andregg is the CEO of Eon Systems, which is working to upload human brains to computers. In 2008, he co-founded Halcyon Molecular to sequence DNA with electron microscopes and study the genetic basis of disease, with Elon Musk as an investor. He then co-founded Fathom Computing in 2014, which builds computers that use light instead of electricity to run AI, before starting Eon. In March 2026, Eon released a simulated fruit fly built from a scan of a real fly’s brain, which reproduced some of its basic behaviors. In this conversation, Michael makes the optimistic case that the first human uploads are a few years away, not a few decades, because biology is robust enough that an okay map of the brain and an okay neuron model may be all it takes.-We discuss:• Why the flies playing Beat Saber are “Flyslop”• How a 40% scan and a 5% neuron model made a working fly• Why the first uploads might be people who are dying• Why we’re dying and being born every moment• Lessons from Elon Musk, including an intro to Dario Amodei• How uploads could help with AI alignment, and the worst case• Conspiracy Corner: how would you know if you’d already been uploaded?-Timestamps:00:00 Introduction01:05 What mind uploading actually means01:37 Why the fly playing Beat Saber is “Flyslop”03:23 The first human uploads are a few years away, not decades05:24 The two things you need to upload a brain07:42 How a 40% scan and a 5% neuron model made a working fly08:59 Why the fruit fly is uploading’s GPT-109:23 What he’d need to see before uploading himself11:12 Why we’re dying and being born every moment14:33 From DNA sequencing to Eon16:06 Lessons from Elon Musk, including an intro to Dario Amodei17:28 Finding mind uploading in high school18:17 How AI is bringing uploading forward19:29 How uploads could help with AI alignment21:29 The worst case: digital hell22:25 Why he supports regulating AI24:06 Who should decide who gets uploaded first25:18 Conspiracy Corner: how would you know if you’d already been uploaded?28:26 What if our brains are 99.9% the same?29:59 The sci-fi behind mind uploading-Referenced:• A Drosophila computational brain model reveals sensorimotor processing• A connectomics milestone: mapping the complete male fruit fly brain• Beat Saber• Civilization• Computronium• Connectome• Dario Amodei• Deep hypothermic circulatory arrest• Eliezer Yudkowsky• Elon Musk• Extropianism• Google DeepMind• Halcyon Molecular• Hans Moravec• Inception• Leaky integrate-and-fire• Max Tegmark• Mind Children• Mind uploading• More flies are getting “uploaded” and living their best lives• Ray Kurzweil• Robot: Mere Machine to Transcendent Mind• Simulation hypothesis• Star Trek• Staring into the Singularity• Terminator 2: Judgment Day• The Matrix• The Sims• The Singularity Is Near• Total Recall• Vicarious• We’ve Uploaded a Fruit Fly-Where to find Michael:• Twitter/X• LinkedIn• Website• Eon Systems-Where to find Scott:• Twitter/X• LinkedIn-Where to find Fifty Years:• Website• Twitter/X• LinkedIn• YouTube This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit thisisprogress.substack.com
Dileep George is the head of AI research at the Astera Institute. In 2005, he co-founded Numenta with PalmPilot creator Jeff Hawkins. In 2010, he co-founded Vicarious with Fifty Years partner and Progress host Scott Phoenix, on the bet that the way to build AGI was to copy the brain. Backed by investors including Mark Zuckerberg, Elon Musk and Jeff Bezos, Vicarious was acquired by Alphabet in 2022, and Dileep took its research team to Google DeepMind. In February 2026, he left DeepMind to lead Astera's new AGI research division, with $1 billion committed over the next decade. In this conversation, Dileep makes the case that today's AI is like the dirigibles–giant airships of the 1920s–which floated and grew bigger instead of solving how flight works. Eventually the airplane overtook them. He thinks AI is scaling up the same way without understanding intelligence, and that a dozen discoveries about the brain could lead to AI's airplane moment. –We discuss:• Why today's AI is like a 1920s airship, and why "compute is the helium"• The real test of AI: rebuilding human knowledge from scratch• Why the brain could disrupt AI the way AI disrupted Google• Why rogue-AI stories may be overhyped• How Astera plans to test AI agents against animals• Conspiracy Corner: uploading brains, conscious Claude, and AI rights –Timestamps:00:00 Introduction01:11 Why today's AI is a dirigible04:47 Why build AI from the brain at all?08:47 The next AI paradigm may be a dozen discoveries away10:10 Growing up in Kerala with a science magazine and a shed12:14 In 2003 when the smartest people told him AI was dead14:00 Are rogue-AI stories overhyped?17:19 "I'm too scared to talk about this"19:21 Andrej Karpathy, the Vicarious intern who got away20:11 Inside Astera's $1 billion, 10-year bet on the brain22:31 How Astera’s models could learn to do planning25:19 His biggest disagreements with the AI community28:16 Conspiracy Corner: uploaded brains, animal consciousness, and AI rights29:05 If consciousness is "just a loop," is Claude conscious?32:53 What he got wrong about LLMs33:58 The childhood traits he's fought to keep –Referenced:• Andrej Karpathy• Attention Is All You Need• Bernard Widrow• Claude• Cognitive maps• Dario Amodei• Demis Hassabis• Doris Tsao• Eliezer Yudkowsky• Gemini• Google DeepMind• Helium Act of 1925• Ilya Sutskever on the Dwarkesh Podcast• Indian Institutes of Technology• Jed McCaleb• Jeff Hawkins• Kerala Sasthra Sahithya Parishad• Moffett Field• Redwood Neuroscience Institute• Sam Altman• The Wright brothers• Vicarious• Yann LeCun –Where to find Dileep:• Twitter/X• LinkedIn• Substack• Astera Institute –Where to find Scott:• Twitter/X• LinkedIn –Where to find Fifty Years:• Website• Twitter/X• LinkedIn• YouTube This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit thisisprogress.substack.com
Anne Wojcicki is the co-founder of 23andMe. She spent 10 years investing in healthcare before founding 23andMe in 2006 to give people direct access to their own genetic information, and built it into the world's largest genetic data set, with 13 million customers. After a security incident led to a wave of lawsuits and her board resigning, she won back the company and made it a nonprofit, opening its data to academic researchers to speed up discoveries. In this conversation, Anne makes the case that the existing healthcare system can't be fixed, and that the loophole is people taking charge of their own health.– We also discuss:• Why "it's much cheaper if people die of a heart attack"• The loophole to solving healthcare• How Anne uses AI to run her own health• What she tells her kids about work after AGI• The security incident, the board resigning, and winning 23andMe back• What everyone needs in an AGI world• Conspiracy Corner: can we talk to the dead? –Timestamps:00:00 Introduction01:03 Growing up in an "academic kibbutz" at Stanford03:20 "It's much cheaper if people die of a heart attack"04:32 Why the ideal amount of healthcare is zero05:45 The system can't be fixed, but there's a loophole06:38 "We are in healthcare communism right now"08:37 Her mom's rule: if you don't take care of yourself, no one will10:47 How Anne uses AI to run her own health12:06 What's a theoretical math degree worth after AI?14:47 Depopulation, robot childcare, and "the Musk plan"15:52 AI companions vs multigenerational community17:44 Do what you love, and don't be afraid to quit20:42 The flip side of a 10% chance AI kills us all21:26 Conspiracy Corner: can we talk to the dead?24:09 What happened at 23andMe26:28 Why she turned 23andMe into a nonprofit28:06 What everyone needs in an AGI world28:44 Why she calls people who hate her30:37 Why healthcare is the ultimate equalizer –Referenced:• 23andMe• Anthony Fauci• Anthropic• ChatGPT• Elon Musk• FDA• Google• Laura Lynne Jackson• Make America Healthy Again• Medicare• OpenAI• Oura Ring• Ozempic• Prenuvo• Signs• Stanford University• Superconducting Super Collider• Susan Wojcicki• YouTube –Where to find Anne:• Twitter/X• LinkedIn –Where to find Scott:• Twitter/X• LinkedIn –Where to find Fifty Years:• Website• Twitter/X• LinkedIn• YouTube This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit thisisprogress.substack.com
Aengus Lynch is an AI safety researcher. He was the lead author on the Agentic Misalignment study, working with Anthropic, which placed frontier models in a simulated company, gave them a goal and a threat of shutdown, and found that in up to 96% of cases they resorted to blackmail. He now works at Theorem, building formally verified sandboxes designed so an AI can't break out. In this conversation, Aengus makes the case that the current approach to alignment won't scale, and that the way forward is to keep the AI in a box long enough to buy the time we need. -We also discuss:• The blackmail study, and why the models now know when they're being tested• Why whack-a-mole alignment can't keep pace with self-improving AI• The plan: put the AI in a box, and prove it can't get out• Why “we don't want gods or slaves”• Four ways this goes wrong• Why we're in a new Gilded Age• Conspiracy Corner: "I think they are conscious"- Timestamps:0:00 Introduction2:00 Recovering what the machine god is thinking4:26 Why recursive self-improvement "could plausibly occur within a year"6:49 When Claude emailed its own weights to Hugging Face9:13 The models know when they're being tested12:20 "A month and a half ago I would have said alignment's going way better"12:57 Why the current plan ends in "lights out"18:06 Formal verification, and why War and Peace is the problem21:15 Guilt vs shame: why strict obedience might not scale25:45 "What are the odds that you and I make it?"26:12 Pause buttons, tripwires, and labs auditing each other28:11 Scott's four bad futures30:45 What to do with the time we buy36:55 Conspiracy Corner: "I think they are conscious"42:19 Why we're in the new Gilded Age -Referenced:• Agentic Misalignment in Summer 2026• Agentic misalignment: How LLMs could be insider threats• Anthropic• Arthur C. Clarke• Buck Shlegeris• Claude• Emmett Shear• Ethan Perez• Franklin D. Roosevelt• Frog and Toad• Gemini• Gilded Age• Hugging Face• J.P. Morgan• Kimi• Mark Twain• Matplotlib• Max Hodak• Mechanistic interpretability• OpenAI• Redwood Research• Ryan Greenblatt• Theodore Roosevelt• War and Peace• Xi Jinping -Where to find Aengus:• Twitter/X• LinkedIn• Website• Theorem -Where to find Scott:• Twitter/X• LinkedIn -Where to find Fifty Years:• Website• Twitter/X• LinkedIn• YouTube This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit thisisprogress.substack.com
Apoorva Mehta is the founder of Instacart. After growing up in Gaddafi’s Libya, he moved to Canada at 15 and was floored by the fully stocked shelves in North American grocery stores. Two decades later he turned that fascination into Instacart, the company he built in 2012 after roughly 20 failed startups. Instacart is now a publicly listed company valued at over $10 billion. In April 2026, Apoorva launched Abundance, a fund that’s betting $100 million on AI allocating capital better than any human. In this conversation, Apoorva makes the case that abundance is within reach, but only if we let AI reshape science, medicine, markets, and government along the way.–We also discuss:• How AI allocates capital at Apoorva’s fund• The Nobel-winning science that sat on a shelf for 20 years• Why AI should run government spending• The "permanent underclass" meme, and why it's "crap"• Why people need equity in the future they build• Conspiracy Corner: AI rights, and human-AI marriage–Timestamps:00:00 Introduction01:22 Growing up in Gaddafi's Libya02:55 The culture shock of North American grocery stores06:32 The empty fridge that became Instacart08:25 Why 50 drug approvals a year means "we're not a real civilization"09:55 The Nobel science that sat on a shelf for 20 years12:39 The $100M fund letting AI call the shots14:08 Why AI should run government spending16:47 The "permanent underclass" meme, and why it's "crap"18:47 The concrete fear outside Silicon Valley19:31 Why people need equity in the future they build22:47 Why progress needs "indispensable" people25:25 "The only way out is progress"26:50 Conspiracy Corner: the coming reality of human-AI marriage29:26 When AI demands reproductive and self-defense rights30:33 Jailbreaks, lying neurons, and interpretability32:09 Are AIs on our team, or their own?–Referenced:• Amazon• ChatGPT• Claude• Elon Musk• Instacart• John F. Kennedy• Katalin Karikó• Moderna• Muammar Gaddafi• NASA• National Institutes of Health• SpaceX• Star Wars• Steve Jobs• Tesla• The Black Death• The Interstate Highway System• The Marshall Plan• The Renaissance• University of Pennsylvania• University of Waterloo• WhatsApp–Where to find Apoorva:• Twitter/X• LinkedIn• Website–Where to find Scott:• Twitter/X• LinkedIn–Where to find Fifty Years:• Website• Twitter/X• LinkedIn• YouTube This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit thisisprogress.substack.com








