YouTube Digest

English

Based on "The bizarre phenomena that medicine struggles to explain | David Linden: Full Interview" from Big Think Watch the original video

Decoding the Mind-Body Mystery: A Neuroscientist's Journey into the Unexplained

For centuries, the intricate dance between our thoughts, emotions, and physical well-being has been relegated to the realms of philosophy, spirituality, or even the supernatural. Science, in its quest for empirical truth, often viewed the profound influence of the mind on the body with skepticism. But what if the "untestable and ethereal" is, in fact, rooted in measurable biology?

David Linden, a distinguished professor of neuroscience at Johns Hopkins University School of Medicine, has dedicated his career to studying neuroplasticity – how the brain and nervous system are reshaped by experience. In a recent interview, Linden shared his journey from a biologist resistant to the mind-body connection to a fervent advocate, revealing how cutting-edge neuroscience is finally unraveling these long-held mysteries, even touching upon his own battle with terminal cancer.

The Unbreakable Link: A Biological Conversation

Linden admits that, like many biologists, he initially resisted the idea that mental life profoundly affects the body. The concept seemed too intangible, too far removed from the hard science of biology. Yet, his father, a psychiatrist, offered a pivotal insight: "Psychiatry works not in some magical psychological realm divorced from biology, it works because it changes your brain." This principle, Linden emphasizes, applies to all behavioral practices—meditation, psychotherapy, controlled breathing—they all exert their effects through biological mechanisms that we are only now beginning to understand.

A key to this understanding lies in distinguishing between how we sense the world and how we sense ourselves. We are familiar with "exteroceptive" senses like sight, smell, and hearing, which provide information about our external environment. But equally crucial are our "interoceptive" senses, the inward-pointing signals that tell us if our bladder is full, our bowel is distended, or how our head relates to gravity. These interoceptive signals are the bedrock of the ongoing conversation between mind and body.

This conversation happens through various biological channels:

The mind, instantiated in the brain, also talks back to the body:

Hacking Hunger: The GLP-1 Revolution

The complexity of eating and hunger provides a vivid illustration of this intricate mind-body dialogue. From smelling pizza on the street to deciding to swallow a bite, a series of conscious and subconscious decisions are made, integrating exteroceptive (smell, sight, taste) and interoceptive (stomach fullness, nutrient content) information, alongside memory and association.

Crucially, our gut is remarkably sophisticated. Cells in the stomach lining can rapidly assess the nutrient content of food, distinguishing water from protein, fat, or carbohydrates. Later, sensors in the small intestine detect passing nutrients, triggering the slow hormonal signal of Glucagon-Like Peptide 1 (GLP-1). This hormone slows gastric emptying and, more importantly, suppresses appetite in the brain for hours.

Linden highlights a fascinating disconnect: while artificial sweeteners trick the sweet sensors in our mouths, they fail to fool the molecularly different sugar sensors in our stomachs and intestines. This mismatch—mouth saying "sugar," gut saying "no"—sends conflicting signals to the brain, likely explaining why artificial sweeteners are often ineffective for weight loss.

This understanding of GLP-1 has led to a revolution in weight management. Natural GLP-1 is quickly degraded, but clever chemists found a way to modify the molecule with fatty acids, allowing it to bind to albumin in the blood. This modification makes drugs like semaglutide (Ozempic, Wegovy) and tirzepatide (Zepbound, Mounjaro) resistant to degradation, enabling once-weekly injections that profoundly suppress appetite, leading to significant weight loss (12-17% of body weight).

Beyond weight loss, GLP-1 drugs show even greater promise. Receptors for GLP-1 are found in numerous organs—heart, kidneys, liver—suggesting beneficial effects beyond mere weight reduction, possibly through anti-inflammatory actions. Intriguingly, early research also indicates these drugs may help control other compulsive reward-seeking behaviors, such as alcohol consumption, drug abuse, and even compulsive shopping or gambling, hinting at a shared underlying reward circuitry in the brain.

Even exercise, long known for its health benefits, plays a role in appetite regulation. Intensive exercise produces a metabolite (lactate conjugated with phenylalanine) that travels to the brain to suppress appetite. While not a complete antidote to overeating, it provides a net benefit.

Linden also points to a darker side of our evolved hunger mechanisms. Our ancestors, facing intermittent famines, were wired to "snarf down" high-calorie foods. Today, the processed food industry exploits this ancient circuit, using sophisticated engineering to override our natural satiety signals, contributing to the significant weight gain observed in populations like the United States since 1960.

The Power of Belief: Voodoo Death, Broken Hearts, and Placebos

The mind's influence extends to life and death itself. Linden discusses Walter Cannon's 1942 report on "voodoo death," where individuals who believed they were cursed actually died. Cannon theorized sympathetic nervous system hyperarousal. Modern understanding refines this: a "one-two punch" of sympathetic overdrive followed by sustained parasympathetic activation shuts down the body. Crucially, this only works if the belief in the curse is present.

This isn't just a phenomenon of distant cultures. Linden cites cases of "fatal misdiagnoses"—like a man who died shortly after Christmas, believing he had terminal liver cancer, only for an autopsy to reveal he was largely healthy. His belief, not the disease, killed him, likely through the same biological mechanisms as voodoo death.

The romantic notion of dying from a "broken heart" also finds scientific validation. Epidemiological studies confirm that people are statistically more likely to die soon after losing a long-term partner, from cardiovascular events, cancer, or autoimmune diseases. This grief-induced decline is biologically explained by phenomena like Takotsubo cardiomyopathy (or "octopus trap heart"), first identified in Japan, where severe emotional stress causes a dramatic, temporary impairment of heart function, linked to sympathetic nervous system overactivation.

Fortunately, the mind's power can also be harnessed for good. Positive beliefs and mental states can relieve pain and improve health outcomes. While behavioral interventions like meditation and psychotherapy are vital, understanding mind-body signaling also informs conventional therapies. GLP-1 drugs, as discussed, leverage body-to-mind signals. Vagus nerve stimulators, used for depression and epilepsy, activate the parasympathetic nervous system, exploiting mind-to-body pathways.

The placebo effect is perhaps the most striking demonstration of the mind's healing power. Sugar pills or sham treatments can significantly alleviate pain, clear infections, and improve cardiovascular function. For pain, we now know that placebos activate the brain's natural opioid system, releasing endorphins and enkephalins—an effect that can be blocked by naloxone. Placebos also appear to involve a form of associative learning, akin to Pavlovian conditioning, where a neutral stimulus (like a lemon-lime flavor) paired with a drug can eventually elicit some of the drug's effects on its own.

Curiously, the placebo effect for pain has grown stronger over recent decades, but primarily in the United States. Linden speculates this might be linked to direct-to-consumer drug advertising, which is allowed in only two countries worldwide (the US and New Zealand), potentially cultivating a stronger belief in drug efficacy. Even more counterintuitively, open-label placebos—where patients are explicitly told they are receiving a sugar pill—still produce beneficial effects, a phenomenon that continues to fascinate neuroscientists.

The Brain's Fight Against Cancer: A Personal Revelation

Linden's journey into the mind-body connection took a profoundly personal turn nearly five years ago when he was diagnosed with synovial sarcoma, an aggressive cancer, and given 6 to 18 months to live. For Linden, a neuroscientist, battling a terminal illness meant embracing "the way of the nerd"—understanding his disease through the lens of neuroscience.

He explains cancer not just as uncontrolled cell growth, but as an "opportunistic bastard" that co-opts other bodily systems. Tumors manipulate the immune system to avoid detection, secrete signals to attract blood vessels for nourishment, and, as increasingly understood, engage in a complex dialogue with nerves. Tumors secrete neurotrophins (like nerve growth factor, NGF) that cause nerves to grow into them, while pain-conveying sensory neurons within the tumor secrete CGRP, which suppresses immune cells (CD8+ T lymphocytes) that would otherwise fight the cancer. This tumor-nerve interaction worsens prognosis, opening a new frontier in cancer therapy: preventing nerve innervation.

The idea that psychosocial support, love, and mental state can influence cancer outcomes has long been a hopeful thought. Linden confirms that epidemiological evidence now strongly suggests a beneficial effect on cancer progression, though it's not a miracle cure for chemotherapy or surgery. Understanding the biology behind this—such as stress hormones promoting tumor growth and psychotherapy reducing stress hormone signaling—opens avenues for new drug therapies. For instance, beta-blockers, drugs commonly used for anxiety and heart conditions, can slow cancer progression in certain types of tumors that possess beta-adrenergic receptors.

Recent randomized controlled studies offer compelling evidence for exercise in cancer outcomes. A 2025 study on colon cancer patients, for example, found that those randomly assigned to an exercise group had a remarkable 30% improvement in mortality over eight years. This is an effect size comparable to powerful anti-cancer drugs, robustly demonstrating exercise's direct benefit, independent of confounding factors. "If you're battling cancer and you can bring yourself to do it," Linden asserts, "exercise is one of the very, very best things that you can do."

Linden's own continued survival, far beyond his initial prognosis, serves as a poignant testament to these insights. When asked why he believes he is still alive, he speaks of the "deep love" he feels from his wife, Dina. He clarifies that this isn't a sentimental remark but a scientific observation. For Linden, love and social connection are not merely emotional comforts; they are powerful biological forces that profoundly influence our physiology, our resilience, and ultimately, our ability to heal and survive.

The revolution in understanding the mind-body connection is transforming medicine. It's moving beyond the ethereal, providing biological explanations for phenomena once thought inexplicable. This deeper insight promises a future where behavioral interventions, targeted drugs, and sophisticated neural devices will converge, exploiting the profound and often surprising power of our own minds to foster health and combat disease.

Based on "669. Why Is 95 Percent of the World’s Bourbon Made in Kentucky? | Freakonomics Radio" from Freakonomics Radio Network Watch the original video

Kentucky's Golden Spirit: The Barrel Tax, Bourbon Glut, and the Battle for Tomorrow's Drinkers

For most businesses, time is money in the most direct sense: products are made, sold, and investments recouped. But in the world of fine spirits, time isn't just money; it's a crucial ingredient, an investment that matures over years, even decades, before yielding its full value. This unique economic reality is nowhere more evident than in the American bourbon industry, a sector currently grappling with a paradox of its own making: an unparalleled concentration of production, a rich history, and a looming glut that could redefine its future.

Freakonomics Radio recently delved into this fascinating industry, exploring why an astounding 95% of the world's bourbon is made within a 45-minute drive of Lexington, Kentucky, and what happens when an industry built on patience suddenly finds itself with too much product and not enough demand.

The Allure of Age: Time as an Investment

Consider products like fine cheeses or certain wines; their value appreciates with age. But for spirits, particularly Scotch whiskey and its American cousin, bourbon, the older the barrel, the dearer the bottle. This means distillers pour significant capital into production, only to have their inventory sit in warehouses for years, sometimes decades, before it can be sold.

Ken Troske, a labor economist and chair of the economics department at the University of Kentucky, observes this phenomenon firsthand. Having moved to Kentucky two decades ago, he witnessed bourbon's resurgence. "When I got there, bourbon was just starting to recover from a long period of time of when it was in decline," Troske recounts. Today, the state holds a staggering 16 million barrels of bourbon aging in its rickhouses, a fourfold increase from the 4 million barrels when Troske arrived.

The dramatic shift in value is exemplified by Pappy Van Winkle's 20-year-old bourbon. Troske's wife once bought a bottle for $120. Today, that same bottle could fetch $2,500 to $3,000 on the secondary market. This isn't just a testament to quality; it's a stark illustration of how time, scarcity, and brand prestige can inflate value. The allure of such high-end products even spawned a black market during the bourbon boom, with 65 cases of 20-year Pappy Van Winkle stolen from a warehouse, valued at an estimated $100,000.

Kentucky's Claim to Fame: More Than Just Limestone?

The concentration of bourbon production in Kentucky is a striking economic anomaly. The state's proponents often cite the unique limestone-filtered water, which purportedly lends itself perfectly to the bourbon-making process, as a primary reason. This same geological feature is also credited with contributing to the success of Kentucky's thoroughbred horse industry, another sector highly concentrated in the Bluegrass region.

However, some economists, like Andrew Muhammad, an agricultural economist at the University of Tennessee, suggest there might be an element of "soft protectionism" at play. "You do have to admit it's a bit convenient, right?" Muhammad muses, highlighting that some of the finest Scotch and Japanese whiskies use barrels from Kentucky and Tennessee. This raises questions about whether the "quality distinction" argument might also serve as clever marketing for the incumbent industry.

The regulatory framework for bourbon is indeed unique. As Ken Troske explains, according to the Code of Federal Regulations (CFR Chapter 27, Section 5.143), to be called "straight bourbon whiskey," it must be:

These regulations, dating back to a 1964 US Congressional resolution, declare bourbon a distinctive product of the United States, much like Scotch whiskey is exclusive to Scotland or Champagne to France's Champagne region. While Master Distiller Danny Kahn of Sazerac (which owns brands like Buffalo Trace and Pappy Van Winkle) staunchly defends each component of the law as beneficial for flavor, the historical context also suggests a confluence of interests, particularly from the corn, barrel-making, and copper industries.

The Art and Science of the Brown Spirit

Danny Kahn, who transitioned from brewing beer at Anheuser-Busch to master distiller, provides a fascinating glimpse into the meticulous process behind bourbon. "If you think about whiskey, we make beer first, then we distill it, and then we age it," he explains.

Kahn elaborates on the five key requirements:

  1. 51% or more corn: Beyond historical availability, corn provides abundant starch for sugar and alcohol production, contributing significant flavor from its oils.
  2. Made in the USA: This allows for legal control and consistent application of standards.
  3. Distilled to no more than 160 proof: Kahn notes that distillers typically aim for 135-140 proof. Distilling at a lower proof retains more flavors, which are crucial for how the whiskey ages, slowly oxidizing and reacting with wood components.
  4. Put into a barrel at no more than 125 proof: This proof level is critical because different flavor components in the wood are soluble in varying alcohol-to-water ratios, allowing distillers to manipulate the final taste profile.
  5. New charred oak containers (barrels): This is perhaps the most iconic and impactful rule. The process begins with white oak staves, which are seasoned outdoors for about a year. Rainwater rinses tannins, and microbiological activity breaks down wood components. Once assembled, the inside of the barrel is subjected to intense direct flame for about 30 seconds, creating a char layer. This char acts like a filter, while the cooked wood beneath it provides the "extractables" – the hemicellulose (caramel, butterscotch notes), lignins (baking spices, vanilla), and tannins (mouthfeel, astringency, color) that define bourbon's complex flavor.

Over time, a significant portion of the whiskey evaporates through the wood, famously known as the "angel's share." For some 23-year-old bourbons, Kahn notes, "there's just a little bit left at the bottom." The used barrels, no longer suitable for bourbon, are often sold to producers of Scotch, Irish, or Indian whiskies, though Kahn admits the market for them "is not so great" currently, leading to a proliferation of "roadside planters" made from old bourbon barrels in Kentucky.

The Boom's Hangover: A "Quantity Problem"

From 2012 to 2022, the bourbon industry experienced a robust boom, attracting numerous new distilleries. However, as Brad Patrick, an executive in residence at the University of Kentucky's Gatton College of Business and Economics, observes, "now we've bumped into a wall." Demand for bourbon has been falling since 2022, exacerbated by new tariffs, leaving the industry with its 16 million aging barrels and a serious "quantity problem."

This isn't a new phenomenon. Patrick draws parallels to the Scotch industry's overproduction in the 1980s and bourbon's own struggles from Prohibition to the 1970s. The current downturn is multifaceted:

The industry's largest player, Jim Beam (owned by Suntory), has already paused production at its flagship distillery for a year, and consolidation and layoffs are affecting other producers.

The Inefficient Web of Regulation and Trade

Beyond shifting consumer preferences, the bourbon industry faces structural challenges, particularly in distribution and international trade.

The Three-Tier System: Economist Ken Troske criticizes the US's "odd" three-tier distribution system for alcohol, a historical artifact from post-Prohibition efforts to curb organized crime. This system mandates separate entities for production, distribution, and retail, with no financial relationships between them. While it may have served its original purpose, today it's "full of inefficiency and middlemen," leading to price discrepancies. Troske cites the example of a bottle of Blanton's bourbon: $74 at Buffalo Trace's gift shop in Kentucky, $130 at a local Kentucky liquor store, and a staggering $400 in La Jolla, California. Distributors also leverage popular brands, forcing retailers to sell less desirable products (like Fireball) to get allocations of sought-after bourbons.

Tariffs: The Trump administration's trade policies further complicated matters. Retaliatory tariffs from the European Union, specifically a 25% tariff on American whiskey, significantly impacted exports. Andrew Muhammad explains that some companies chose to absorb the cost rather than raise prices, fearing the permanent loss of market share. "Once you lose market share in the distilled spirits sector, it's very hard to get it back," he notes. Canada took an even more drastic step, simply removing American products from shelves.

Barrel Taxes: The state of Kentucky also levies an annual "barrel tax" on aging bourbon, a legacy "sin tax" that in 2025 amounted to $75 million. While this tax is being phased out by 2043, it represents yet another cost burden on distillers already investing for years without immediate returns.

The Road Ahead: Creative Destruction or Just a Hangover?

The bourbon industry is at a crossroads. Brad Patrick predicts a "weeding out of businesses," similar to the Scotch industry's consolidation to five major players after its 1980s bust. Distilleries with strong brands are likely to survive, while contract distillers heavily reliant on borrowed money face the most significant trouble.

The question remains: where is the "creative" coming out of this "destruction"?

While the image of a bourbon connoisseur savoring a meticulously aged spirit endures, the industry is clearly adapting to a new era. The current glut of high-quality bourbon signals a challenging period, but also potentially an opportunity for innovation. Whether it's through diversifying product lines, strengthening global presence, or enhancing the consumer experience, Kentucky's golden spirit is once again proving that patience is not just a virtue, but a necessity, in the ever-evolving world of alcohol.

Based on "AI for Atoms: How Periodic Labs is Revolutionizing Materials Engineering with Co-Founder Liam Fedus" from No Priors: AI, Machine Learning, Tech, & Startups Watch the original video

The Atomic Leap: How ChatGPT's Co-Creator is Forging AI for the Physical World

Liam Fedus is a name synonymous with the AI revolution. As one of the co-creators of ChatGPT and a former VP of post-training at OpenAI, his work has indelibly shaped how millions interact with artificial intelligence. Yet, after helping to unleash the power of language models on the digital realm, Fedus has turned his formidable intellect towards a new, perhaps even more profound frontier: the physical world. Through his new company, Periodic Labs, he is building an "AI foundation lab for atoms," aiming to revolutionize material sciences, chemistry, and engineering.

This isn't just about writing smarter software; it's about fundamentally changing how we discover, design, and produce the very matter that constitutes our world.

From Dark Matter to Digital Minds

Fedus's journey to the forefront of AI began, perhaps surprisingly, in the depths of physics. An undergraduate physics major, he spent time researching dark matter, working with an apparatus "directionally sensitive to dark matter's direction." This background is not uncommon among today's AI luminaries, prompting the host to ask why so many physicists, like Anthropic's Dario Amodei or Google's Adam Brown, gravitate towards AI.

Fedus attributes this trend to physics instilling a "very principled," "hard-nosed scientist" approach to thinking about the world. "It's such an incredible field," he explains, "you have such high leverage in computer science in AI." Many high-energy physicists, particularly after the discovery of the Higgs boson, found themselves looking for "what's next," realizing their skill sets could be "a huge contributor elsewhere." It's a migration that, as the host muses, "almost feels like we're recreating the Manhattan project... except now what we're seeking is different forms of intelligence."

Fedus's own transition began in grad school, where he found himself "always gravitating towards the machine learning problems," particularly in particle reconstruction. Recognizing that pushing the frontier of machine learning required a shift to computer science, he landed at Google Brain in 2016-2017. This period, he recalls, was a "remarkable period for Google Brain" – a "Cambrian era" marked by the creation of distributed training strategies, Mixture of Experts, and the Transformer architecture. "The field was much, much earlier," he says, "and I think there was a lot of diversity and entropy in the research, and it was very fun."

His work at Google focused on architecture, specifically pushing "sparsity that allows for more efficient serving of models at scale" and expanding the capabilities of what AI could achieve. By late 2022, the technology was becoming "very compelling," and Fedus, along with other Google colleagues, moved to OpenAI.

The Birth of ChatGPT and the Call of the Physical

At OpenAI, Fedus's primary goal was the "productionization of GPT-4." The powerful model had been pre-trained, but the question remained: how to turn it into a product? Initial ideas ranged from writing bots to coding bots, even a "meeting bot" that would take notes and assign to-dos. However, John Schulman, a key researcher, was "very opinionated" in favor of a general chatbot. "And that became a large part of the effort," Fedus explains, culminating in the launch of ChatGPT.

The impact was instantaneous and profound. "That was kind of the starting gun of this whole AI revolution," the host notes, acknowledging how ChatGPT suddenly brought AI's power into public awareness.

But for Fedus, language models, while revolutionary, were just the beginning. The "inevitability of connecting these systems to the physical world" became his driving conviction. "You're not going to see the same kind of acceleration in science and technology unless you start connecting these things to the physical world," he asserts. "Science ultimately isn't sitting in a room thinking really hard. You have to conduct experiments, you have to learn from them, you have to interface with reality."

The technology available in 2022, he admits, was "far too weak" for Periodic Labs' ambitions. It required further advancements in reasoning, test-time inference, reliable error correction, and sophisticated tool use – capabilities that have since emerged in more advanced models and coding agents. These foundational technologies were "necessary to then connect these systems to the physical world."

Bridging the Data Divide: The Challenge of Atoms

One of the biggest hurdles in applying AI to the physical world is data. Unlike language models, which feast on the internet's vast textual corpus, materials science lacks such a readily available, clean, and comprehensive dataset. Fedus explains that ML systems excel on the data they're trained on, and while there's a "mythology of AGI" sometimes, powerful systems are "limited if they don't have access to the raw data to actually make informed decisions."

Periodic Labs tackles this by leveraging existing foundation models, benefiting from the "order tens of trillions of tokens" they were trained on. This provides a "strong prior on the world," meaning their systems don't start from scratch, but already possess a foundational understanding of language and code.

However, this is "insufficient" for the specifics of material discovery. Fedus recounts an anecdote where an engineer found reported material properties in literature spanning "many orders of magnitude." Training an ML system on such inconsistent data merely models the distribution, offering no closer path to "ground truth." This highlights the critical role of experimental data, which provides "a grounding."

Crucially, it's not just about a "pool of data." Periodic Labs emphasizes an "interactive closed-loop system" where experimental data informs the models, which then identify patterns, flag aberrations, and drive the next set of experiments. This active feedback loop is what makes the system truly powerful.

Regarding data diversity, Fedus notes that the "greatest advances" internally occur where there's "an abundance of data in some space." While there's some generalization for systems "strongly governed by quantum mechanical effects," creating a model for quantum objects won't directly help with fluid dynamics. The generalization exists, but often within specific "first principles" domains, allowing them to model fundamental interactions like chemical synthesis or van der Waals forces.

Orchestrating Intelligence for Material Discovery

Periodic Labs' AI architecture leverages the power of language models as an "orchestration layer." Fedus describes them as "a co-pilot assistant but also like a system that can direct experiments." This layer orchestrates other, more specialized models. These include neural networks "specially designed for atomic systems where there's like some symmetry awareness," which boast lower latency and are fine-tuned for specific tasks.

The overall system acts as a sophisticated conductor, capable of ingesting vast amounts of literature, analyzing experimental data, processing different modalities, and utilizing specialized neural nets as "tools" or "reward functions." This modular approach allows the AI to tackle complex material engineering challenges with both broad understanding and targeted precision.

The Vision: Generating Matter and Accelerating Progress

Periodic Labs began by treating itself as "customer zero," transforming its own scientific processes. The commercial opportunities, however, are vast, spanning "all of these industries, all these enterprises that are interfacing with the physical world" – from semiconductors and aerospace to energy. These are fields "bottlenecked by materials engineering, process engineering," where engineers constantly ask questions about data, debug machinery, and seek better formulations. Fedus believes this is "actually a quite universal thing as well."

While starting as a "software business," providing an "intelligence layer" and "control plane" for companies, Fedus acknowledges the "very interesting aspect of some breakthroughs here could have really high value," potentially leading to a "discovery model" akin to biotech.

The conversation naturally drifts to the futuristic implications, evoking Neil Stephenson's novel The Diamond Age, where AI enables matter-generating 3D printers in every home. Fedus's vision for a world in 10 years, if Periodic Labs succeeds, is equally transformative: "You're going from systems that aren't just writing essays, not just writing software, but to literally generating matter." This has "profound implications to semiconductors, aerospace, energy."

His core belief is that AI for atoms can "increase like the pace of just like the physical development of the world." While "atoms are hard," implying physical limits, he sees "an order of magnitude or two to speed up" by making sense of data and reaching solutions faster. "What we're trying to do is give humanity this agency for atomic rearrangement, synthesis, and we think it's going to just be a huge accelerator." He likens it to the "materials equivalent of the agricultural revolution," promising a dramatic shift in human capabilities.

The Power of Multidisciplinary Scale

What excites Fedus most is the "iteration between these groups of people." Periodic Labs is "irreducibly a multidisciplinary problem," bringing together physicists, chemists, top AI researchers, and engineers in "absolutely incredible" close collaboration. He sees "a field fundamentally change" as people who have spent decades in their domains witness the power of intelligent systems.

He draws an analogy to the industrialization of AI research itself: from a "few GPUs and a few people" pushing the frontier in early Google Brain days, to an era "industrialized" by "hundreds of researchers working together with hundreds of thousands, millions of GPUs dictated and driven by scaling laws." This predictability allowed for massive capital investment. Fedus believes physical sciences and engineering "will have a very similar property where we establish these scaling properties and bring that mindset." Periodic Labs aims to bring "much larger scale sets of experiments" to bear, enabled by intelligence and automation, addressing the challenge of scientists being overwhelmed by unprecedented data throughput.

This ambition is capital-intensive, primarily due to "extraordinarily expensive" GPUs. Surprisingly, Fedus notes that "the compute cost relative to physical infrastructure is actually surprising," with compute often being the larger expense, despite the long lead times and intrinsic difficulties of physical systems. He also highlights the societal value of bringing academic scientists into a setting where they can operate at "real scale" and make a different kind of impact, acknowledging the undercompensation of many academic researchers.

The Future of Intelligence and Robotics

Periodic Labs is actively hiring across both "bits" (AI infrastructure, pre-training/mid-training roles) and "atoms" (control engineering, system engineering, project engineering).

When discussing AGI, Fedus offers a nuanced perspective. He believes thinking of "intelligence as a scaler" is a fallacy, as systems often exhibit "odd spikiness" – being "world-class on some math domain" yet degrading substantially with minor perturbations, akin to "a bad high school student." Generalization can be "non-intuitive."

For "recursive self-improvement," Fedus sees a "very clear path for software engineering," where systems can write repositories, identify bugs, and refactor code. This is happening "nowish," driven by "huge amounts of data, really cheap verifiable environments" (like unit tests). AI research self-improvement is a "slower outer loop," requiring more GPUs and longer experiments, but also inevitable. The key, he reiterates, is the "closed loops of actually doing science, of actually doing engineering," connecting these systems to the physical world.

Robotics, while not strictly necessary for Periodic's "escape velocity," is a "huge accelerator." The goal is "to generate high quantity, high-quality data, diverse data," and automation assists this. While currently using "off-the-shelf robotics," improvements in "more general robotic systems" – like a "dexterous humanoid who could wander into an unstructured lab" – will be a "massive accelerator for spinning up new labs."

Beyond Periodic Labs, Fedus remains most excited about robotics and the "interface of AI systems with the physical world." He envisions "agency and control of the physical world via robotics" as transformative, especially given the vast numbers of people who build the physical world compared to software engineers, and the widespread labor shortages.

As the digital realm accelerates at an unprecedented pace, Liam Fedus and Periodic Labs are pioneering the path for the physical world to catch up. By infusing AI into the very fabric of matter, they are not just building a company; they are laying the groundwork for a future where humanity's ability to create, discover, and build is profoundly amplified, ushering in an era of atomic abundance and engineered possibilities.

Based on "What Do You Do When a Family Member Commits a Terrible Crime? | 'The Opinions' Podcast" from New York Times Podcasts Watch the original video

Beyond the Bars: Why Children Need Their Incarcerated Parents, Even After Unspeakable Crimes

When a family member commits a terrible crime, the instinct for many is to sever ties, to erase the problematic individual from the family narrative. But what if this seemingly protective act, driven by understandable outrage and a desire for justice, actually inflicts deeper harm, especially on the children caught in the crossfire? This profound and challenging question lies at the heart of a recent conversation on The New York Times' 'The Opinions' podcast, featuring host M. Gessen and author Harriet Clark. Their discussion delves into the complex, often counterintuitive, necessity of maintaining connection with incarcerated parents, even those who have committed truly heinous acts.

M. Gessen, an opinion columnist and host of 'The Idiot' podcast, found himself grappling with this dilemma firsthand. His first cousin, Alan, was arrested in 2022 for orchestrating a hit on his ex-wife, the mother of his two children, and is now serving a 10-year federal prison sentence. Gessen, initially consumed by a sense of vengeance and a lack of empathy for Alan, sought to understand how children could possibly build relationships with a parent in the prison system. His search led him to Harriet Clark, a friend whose life story offers a rare and powerful perspective.

Harriet's mother, Judy Clark, spent 37 years in prison for her role as the getaway driver in a Brinks robbery that resulted in three deaths. Despite the enormity of her mother's crime, Harriet grew up maintaining a close relationship with her, a bond she credits to a confluence of "luck" within a deeply unlucky situation.

The Unseen Gift of Connection

Harriet’s childhood visits to her mother’s correctional facility were a stark contrast to the typical image of prison. Her mother was incarcerated in a facility that actively sought to facilitate parent-child relationships, offering a "children's center" in the visiting room where kids could play, make crafts, and engage with their mothers. "It tried as much as possible to have the prison still fit inside the child's reality," Harriet recounts. Beyond the institution's efforts, her own family, though fractured by the crime, rallied to ensure the connection. Special dolls by the phone helped a young Harriet stay on calls, her mother's letters describing birds were read aloud, and her grandfather would help her look up the species. "Many people made many efforts to really enable my mother to remain my mother," she emphasizes, calling it "a great form of good fortune in a system that is very often trying to take people's parents away from them."

This experience directly challenges the common sentiment Gessen articulated regarding his cousin: "Why would you want to have a relationship with a parent who did something so horrible?" Harriet explains that research and lived experience repeatedly show that children are aided by staying in relationship with an incarcerated parent. When a parent goes to prison, a child inherits "a very painful piece of knowledge, which is that you are leavable." This terrifying realization needs to be counterpointed by consistent efforts from the parent to connect, to show they still think of the child, still try to support them.

Without this connection, the absent parent becomes "this black hole," taking on a "mythic dimension" that children struggle to comprehend. Kids need a real person to direct their "confusion and their upset and their needs and their rage to." The absence, Harriet argues, is "as present as the presence is," and attempting to remove the parent entirely often replicates a "carceral logic of removal and disappearance," falsely believing the child is then safe.

The Vengeance Trap and Collective Responsibility

Gessen confessed to a personal struggle with vengeance, admitting that while reporting on his cousin's case, he found himself "rooting for the prosecutor" and "wanting this guy to go to prison for as long as possible." He was shocked to find his own sense of vengeance channeled by the very system he once believed was meant to curb such primal impulses. Harriet, nodding in recognition, confirms this observation, citing the New York State prison system where parole is often denied due to "a certain kind of sense of vengeance within the parole board" influenced by survivor communities.

Harriet's experience with her father's incarceration further underscores the damage of absence. Her grandparents, furious at her father, minimized her relationship with him, believing the harsh prison environments (visits through plastic walls, shackled to beds in repulsive prison hospitals) would be "too much upset for a child." But for Harriet, this absence was worse. She "didn't understand what had happened" with her father and "did feel more abandoned."

This "too much upset for a child" logic, while seemingly intuitive, can be misguided. Gessen ponders how his 12-year-old nephew, O, could ever wrap his mind around his father wanting his mother killed. Harriet's response is powerful: O "can't be protected from that reality. What he can be is companioned and loved within that reality." The goal isn't to shield children from the truth of their parent's incarceration, but to ensure that the parent can still be "a source of delight and support and affirmation."

The issue is a profoundly collective reality: "We have five million children in this country who will have a parent incarcerated at some point in their childhood." When visiting prisons, Harriet witnessed "incredible people who are making heroic efforts to not let the state tear their families apart." These are the grandmothers who wake at 4 AM for a seven-hour bus ride, sending a clear message: "I'm not letting them throw you away. I'm not acting like you don't exist. You are still my son, my grandson, my father, my brother, my husband. You are still these children's father." Such efforts, Harriet believes, are inspiring and help to "state vengeance."

Accountability as a Journey, Not an Ultimatum

Gessen's podcast concludes with an ultimatum to his cousin Alan: confess to your crime, take accountability, or you're out of the family. Harriet challenges this, not by saying it's wrong, but by framing it as "a moment in the process." She explains that prison conditions make telling the truth incredibly difficult. "No one who's in an active appeal is going to confess their crime to you." The demand for truth, while sounding "reasonable" and "ethical," might not be the "be all and end all of our relationship with people."

Instead, Harriet advocates for a "very long road to accountability," where families collectively figure out "how connections can be rebuilt." This includes helping the incarcerated individual be the best possible parent they can be while inside, and ensuring the safety and well-being of victims like Priscilla (Alan's ex-wife) after release. "Violence is the aid people turn to when they don't feel like there are other places to turn for help," Harriet notes, suggesting that the "stepping in of the collective" is crucial for healing after violence and rupture.

Gessen struggles with his cousin's habitual lying and manipulation, expressing deep frustration and a lack of trust. Harriet acknowledges the infuriating nature of manipulation, having grown up around it herself. However, she argues that this doesn't justify total disconnection. "If you don't trust him, great. Don't trust him. Don't leave him alone with his children. Offer to be the presence when he gets out that helps to kind of ease that dynamic." It's about being "clearsighted," not blindly trusting, but finding "a collective solution for the fact that you have an untrustworthy member of your family."

This perspective directly counters "carceral logic," which suggests that problematic individuals are simply removed, and that is the solution. Harriet points out the hypocrisy in this: "Have you ever told a lie? Yes. Have you ever manipulated people? Yes. Of course. Right. And so this too is part of the problem of carceral logics. They try to take sinful behavior and say it belongs just to this population. Watch this population." The truth, she asserts, is that "we're all capable of harm."

The Imaginary Sixth Episode: A Path Forward

Harriet offers a vision for Gessen's "imaginary sixth episode," a future where the family collectively works to create "as good a future as possible for Alan and Priscilla and their children and for all of you." This involves ensuring Priscilla's safety, helping Alan be the best parent he can be, and even assisting him in finding work his children can be proud of after his release, given he will be disbarred.

For Harriet, her own parents, despite their past, were people she could be proud of. Her father, who suffered medical mistreatment in prison, became an advocate for healthcare in correctional facilities after his release. Her mother, too, was always presented as worthy of respect by her family, which in turn fostered Harriet's own "strong sense of self-respect."

Ultimately, the message is one of profound human connection and collective responsibility. When a parent is incarcerated, children face a cultural narrative that labels their parent as "bad." The family, then, plays a crucial role as a "countering voice," creating conditions "whereby a child can feel respect towards their parent." It's about finding ways for meaningful connection, even amidst profound disconnection, recognizing the "multiplicity and contradictions of how people are."

Harriet Clark's insights offer a radical, yet deeply humane, challenge to our conventional understanding of justice, punishment, and family. They invite us to look beyond vengeance and simplistic solutions, urging us to embrace the messy, difficult, but ultimately more dignifying journey of maintaining human connection, for the sake of the children and the collective healing of families and communities. It's a truth that may make us think harder than we'd like, but it's a truth that might just be the most important thing for those impacted by the long shadow of incarceration.