Tim FerrissLegendary Inventor Danny Hillis — Solving the Impossible (Plus Kevin Kelly)
CHAPTERS
- 0:00 – 0:57
Feynman’s “kooky idea” summer job: recruiting the unexpected
The episode opens with Danny Hillis recounting how Richard Feynman ended up effectively hiring himself for a summer job at Thinking Machines. The story illustrates Hillis’s attraction to ambitious problems and his habit of learning by working directly with extraordinary people.
- •Feynman initially dismisses Hillis’s parallel-computing idea as “kooky,” then volunteers himself as the one person crazy enough to help
- •Feynman arrives on day one and asks what he should do, prompting an improvised first assignment
- •A mundane startup need (no supplies yet) turns Feynman into the company’s first “quartermaster”
- •Feynman later initiates the first quantum computing project at Thinking Machines
- 0:57 – 3:09
How Danny Hillis and Kevin Kelly connected through Stewart Brand and the Long Now
Tim Ferriss asks Kevin Kelly to describe how he and Hillis first met and began collaborating. Kelly traces it back to Stewart Brand, WIRED, and the early proposal for a 10,000-year clock that became central to long-term thinking work.
- •Stewart Brand brings Hillis into Kevin Kelly’s orbit during Brand’s MIT-related work
- •Kelly publishes Hillis’s proposal about a 10,000-year clock in WIRED
- •A nonprofit forms around trying to actually build the clock, with Kelly involved early
- •Their relationship deepens into decades of collaboration around encouraging long-term thinking
- 3:09 – 6:09
The briefcase of weird objects: hiring for curiosity (and a Mogen clamp)
Tim asks about a frightening-looking device and a silver briefcase used in Hillis’s interviews at Applied Invention. Hillis explains the “box of weird stuff” as a way to identify curiosity and engagement rather than test for credentials.
- •Applied Invention uses unusual objects in interviews to see how candidates react without instructions
- •Strong candidates immediately handle objects, ask questions, and engage—signaling curiosity
- •The device Tim mentions turns out to be used in circumcision, chosen partly for its strangeness
- •The interview method emphasizes learnability and active exploration over preexisting knowledge
- 6:09 – 7:43
Robin Williams at Disney: the path from startup life to Imagineering
Hillis explains why Robin Williams was inspecting the object box and how that connects to Hillis’s time at The Walt Disney Company. He describes Disney as the only traditional paycheck job he ever had, taken after Thinking Machines ended and as family circumstances changed.
- •Robin Williams is identified as the funniest person to open the interview object box
- •Hillis met Robin Williams at Disney, where Hillis worked as Disney Fellow and Vice President of Imagineering
- •Hillis contrasts receiving a paycheck and benefits with his prior experience paying others
- •After Thinking Machines closed, he sought a period of stability while raising newborn twins and a daughter born the day the company closed
- 7:43 – 12:59
From MIT AI ambitions to founding Thinking Machines and proving parallel computing
Hillis describes studying AI under Marvin Minsky and realizing AI needed big, fast parallel computers. He recounts building a full system from chips to operating system, spinning it into a company despite MIT resistance, and achieving major technical success even as the business struggled.
- •Hillis pursued AI at MIT and concluded parallel computing was necessary for AI progress
- •He built a parallel computer end-to-end, which outgrew what a graduate student project could support
- •He started a company as a graduate student despite MIT objecting, then hired faculty (including former MIT president Jerry Wiesner)
- •Thinking Machines built pioneering parallel computers and held “fastest in the world” status for years, but business execution lagged
- •Hillis notes a later commercial path where similar chip ideas ultimately enabled video-game and GPU-style markets, mentioning Nvidia as the eventual realization of the vision
- 12:59 – 16:12
Engineering vs. artistry at Disney: storytelling as the real design tool
Hillis contrasts the engineering mindset (ground truth, correctness) with the show-business mindset (winning arguments, subjective outcomes). He shares an example of Disney Online brainstorming that exposed a disconnect between system diagrams and experiential imagery, leading him to adopt storytelling as a core design frame.
- •Hillis describes show business as loosely tethered to reality compared to science and engineering
- •Movie success is interpreted socially (who gets promoted) despite unclear causal explanations
- •Disney Online brainstorming shows Hillis drawing server block diagrams while others draw castles and visual experiences
- •Hillis comes to see that emotional experience and narrative coherence often matter more than mechanics for success
- 16:12 – 22:51
Designing the 10,000-year clock as a visitor experience, not just a mechanism
Hillis explains how Disney’s storytelling approach reshaped his thinking about the 10,000-year clock. Instead of focusing on purely technical challenges, he focuses on why people would care, how they would experience it, and what meaning the visit creates.
- •Hillis shifts from mechanical questions (winding, materials) to questions of meaning and human experience
- •He reframes the clock around the narrative of a visitor’s journey and emotional arc
- •The clock shows the time and date of the last visitor, then “catches up” when wound
- •Visitors can take home a rubbing of the date as a tactile souvenir created through their own action
- 22:51 – 27:47
AI as “human intelligence on an artificial substrate”: imitation first, then beyond
Kevin Kelly asks what story about AI people are missing, and Hillis responds with his “Songs of Eden” framing. Hillis argues current systems are best seen as a substrate carrying human culture and intelligence, with imitation as an early stage of intelligence.
- •Hillis’s “Songs of Eden” story frames intelligence as co-evolution of social brains and culture-like “songs”
- •He suggests today’s systems reflect learned human culture from training data rather than a wholly separate artificial intelligence
- •He describes the present moment as a powerful imitation stage, analogous to child development
- •He notes this isn’t the only possible path to AI, but it explains much of what’s happening now
- 27:47 – 35:16
How Hillis chooses projects: excitement, sustainability, and non-redundancy
Tim presses Hillis on how he navigates a life without a rigid plan yet avoids aimless drifting. Hillis explains invention as the culmination of societal preconditions and shares Applied Invention’s project-selection criteria, including a strict “don’t do it if it’ll happen anyway” rule.
- •Hillis frames invention as enabled by society creating the “preconditions,” after which pieces can be assembled
- •He revisits parallel computing as an example of recognizing when prerequisites (chips, compilers, digital sensors) align
- •Applied Invention uses three tests: a senior partner must be genuinely excited; it must be financially sustainable; and it must be non-redundant
- •Non-redundancy is the hardest test: they avoid projects likely to happen without them
- 35:16 – 41:40
Rebuilding internet security foundations: zero-trust packet routing (ZPR)
Hillis explains why cybersecurity keeps worsening and traces it to foundational internet design choices where network security wasn’t treated as a network responsibility. He outlines an alternative approach—zero-trust packet routing—where packets carry proof of permission, and the network enforces policy.
- •Hillis argues attackers benefit because IP packets can be effectively anonymous, forcing endpoints to guess what’s safe
- •He says a better model is for the network itself to enforce delivery policy rather than deliver everything everywhere
- •He describes ZPR as giving each packet a “passport and visa” to prove permission for its destination
- •He notes the approach can begin as an overlay on the existing internet and mentions Oracle announcing a cloud product using the protocol
- 41:40 – 42:59
Learning by “hanging out” with experts: getting into MIT’s AI Lab
Tim asks how Hillis breaks into new fields, and Hillis details his approach: do enough reading to ask good questions, then learn primarily through people. He tells the story of getting access to the MIT AI Lab by identifying an unmet need in proposals and building a solution to earn entry.
- •Hillis says he reads enough to form questions worth an expert’s time, then learns mainly from conversations
- •To reach the AI Lab, he studies DARPA/NSF proposals available in a lobby library to find a gap he can fill
- •He invents a picture-based way for kids who can’t read to program, earning an interview with Seymour Papert
- •He gets a key to the building, enabling deeper immersion in the AI Lab environment
- 42:59 – 53:37
Working with Marvin Minsky and Richard Feynman: contribution through questions and fixes
Hillis describes meeting Marvin Minsky by showing up at night in the basement project space and contributing by spotting errors and fixing problems. He then returns to the Feynman story to illustrate how serious collaboration can start from improvised tasks and evolve into foundational work.
- •Hillis finds Minsky working nights on a personal computer project in the basement with students wire-wrapping hardware
- •He earns trust by identifying a diagram mistake; Minsky instructs him to fix it on the diagram and machine
- •Hillis becomes a regular contributor and eventually receives a formal job after already doing the work
- •In parallel, the Feynman collaboration at Thinking Machines evolves into substantive technical exploration, including early quantum computing work
- 53:37 – 1:02:30
Entering biotechnology via “last resort” problems: cancer as a verb and proteomic debugging
Tim asks how Hillis moved into biological sciences, and Hillis credits incoming “hard problems” brought to his invention company. He recounts working with oncologist David Agus to reframe cancer as an ongoing process and argues for measuring proteins over time as a way to see disease develop before symptoms appear.
- •A doctor, David Agus, approaches Hillis out of frustration with cancer diagnosis and treatment paradigms
- •Hillis reframes cancer as something the body does (a process) rather than something you simply have (a static condition)
- •He emphasizes protein expression and interaction as where the action is, beyond what genes alone can tell you
- •He describes building methods to measure many proteins from blood (and eventually cells), tracking changes over time in mice
- •He compares proteomics to runtime debugging: genes are the code, proteins are what’s happening while it runs
- 1:02:30 – 1:11:17
Systems thinking in agriculture: reshaping food production near where it’s eaten
Hillis explains how moving to a farm during COVID sparked a deeper interest in agriculture and food systems. Rather than focusing on point solutions, he looks for system-level shifts—like growing food closer to consumption—while acknowledging that doing so requires coordinated changes across labor, greenhouses, plant varieties, and more.
- •Living on a farm leads Hillis to notice food quality differences and question why supply chains produce older, less flavorful produce
- •He critiques energy-intensive shipping and breeding produce for transport durability rather than taste
- •He focuses on system redesign over point-optimizing tasks like weeding or picking tomatoes
- •He argues growing closer to cities requires changes in labor economics, greenhouse design for cold climates, and crop varieties
- •He describes needing visionary funding partners willing to take risks for system-level experiments
- 1:11:17 – 1:18:02
Superpower, homeschooling, and learning from great teachers
Kevin asks about Hillis’s superpower, prompting reflections on fearlessness in learning and mindset shifting. Tim then asks about homeschooling; Hillis describes a collaborative, tutor-supported approach shaped by his own uneven schooling and highlights the impact of teachers who stretch a student’s curiosity.
- •Hillis names not being afraid to learn new things as his “superpower,” likening it to a child’s natural trait
- •Kevin shares a child’s description of Hillis as a “mind shifter” who can adopt multiple perspectives
- •Hillis links this to childhood moves across cultures due to his father’s epidemiology work
- •On homeschooling, Hillis credits his wife, tutors, and a broader community rather than claiming solo responsibility
- •He highlights a librarian, Mrs. Wilner, who guided him from rocks to electricity and science fiction as a model of how great teachers stretch curiosity
- 1:18:02 – 1:27:05
Failure, time, and the entanglement of nature and technology
Hillis reflects on Thinking Machines as a painful failure rooted in business missteps and insufficient attention to sustainability. He then introduces “the Entanglement”: as technology becomes more complex and less fully understood, it begins to resemble nature, while nature increasingly becomes shaped by technological intervention.
- •Hillis calls Thinking Machines his first big failure and believes it didn’t have to fail with better business discipline
- •He cites being blindsided by external forces like laws affecting exports and cash-flow vulnerability from overconfidence
- •He says time now feels most precious and he’s focused on avoiding redundant work and meeting extraordinary younger people
- •He defines “the Entanglement” as the blurring of boundaries: nature becomes engineered (atmosphere, genes, bodies), while technology becomes opaque and emergent (internet, AI outputs)
- •The conversation frames the present era as a special moment where humans create systems more complicated than they can fully understand
- 1:27:05 – 1:47:19
What today’s AI is (and isn’t), plus long-term optimism amid risk
Tim asks how Hillis thinks about AI as discovery vs invention, and Hillis distinguishes “AI” from what is popularly called AI today. Hillis and Kelly discuss intelligence as a multi-dimensional space, the limits of current systems, and how pessimism in the short term can coexist with optimism over longer horizons—even amid climate concerns and existential-risk debates.
- •Hillis says early AI focused on tasks humans find hard (chess, tests), but underestimated harder “easy-for-humans” capabilities like perception and intuition
- •He argues today’s systems are strong at imitation and warns people overestimate present capabilities while underestimating long-run progress
- •Kelly compares the moment to early electricity research: no solid theory yet, just primitives emerging through experimentation
- •Both suggest intelligence is high-dimensional and will be unbundled into components; Hillis adds there may be intelligences beyond human imagination
- •Hillis calls himself a short-term pessimist and long-term optimist, citing historical improvements alongside persistent new problems
- 1:47:19 – 1:59:18
From office noise to medical futures: Babble, proteomics for dementia, and MRI alternatives
The conversation turns to practical inventions and unmet needs: a “Babble” device that masks conversations, tools for neurodegenerative disease via proteomic monitoring, and a more comfortable alternative to MRI. Hillis outlines why protein measurement is foundational and sketches a vision for ultrasound systems that can produce more MRI-like 3D outputs.
- •Hillis describes “Babble,” a device that masks speech in open offices by speaking in a similar voice with nonsensical content to make eavesdropping difficult
- •He says Herman Miller acquired the technology, restaurants showed interest, but the effort stalled after a CEO’s heart attack
- •On neurodegenerative disease, Hillis argues medicine often intervenes too late because the body compensates until damage is advanced
- •He returns to proteomics as a “debugger” to spot disease trajectories early, mentioning funding and reputation challenges after Theranos
- •He says he’s working on an MRI alternative based on ultrasound, noting current ultrasound loses valuable operator context that could be captured to generate 3D images
- 1:59:18 – 2:12:24
Pinch-to-zoom, patent ambivalence, inventor heroes, and a heresy about causality
Hillis recounts building early pinch-to-zoom prototypes and how his patent ultimately invalidated Apple’s pinch-to-zoom claim, enabling broader adoption. He shares skepticism about patents’ societal value, names inventors he admires for simplifying complex reality, and offers a “heresy” rejecting cause-and-effect as fundamental reality—while still valuing it as a useful storytelling tool.
- •Hillis traces his childhood desire for “infinite zoom” in maps to building pinch-to-zoom prototypes and a touch table used in the White House Situation Room
- •He says he never earned money from pinch-to-zoom but is proud it became so intuitive it feels innate to children
- •On patents, he says they may help inventors but often harm society, consume resources through litigation, and should be narrowed
- •He praises inventors like Claude Shannon and Feynman for making messy domains simple and empowering for others
- •His heresy is disbelief in cause and effect as a feature of reality, describing causality as a story humans tell to make sense of equations and events
- 2:12:24 – 2:29:51
Quantum computing, consciousness, and the “long tail” question before investing time
The closing arc revisits quantum computing’s promise and limits and explores whether AI or quantum research can illuminate consciousness. Hillis suggests consciousness may be less central than people think—possibly a language-like hack for accessing thoughts—and ends with his core project filter: how long a difference will matter, aiming for impact beyond his lifetime.
- •Hillis says quantum computing remains theoretically promising but still not broadly useful, similar to how he described it in the early 1990s
- •He connects quantum computation to limits of cause-and-effect explanations, noting observation and state changes complicate causal intuitions
- •On consciousness, Hillis guesses it may be a small, hack-like part of intelligence tied to compressing and decompressing thoughts through internal language
- •Kelly proposes language enables access to our own thoughts and that consciousness could be one primitive among many in possible minds
- •Hillis’s time-allocation question becomes: “Will this make a difference over how much time?” with preference for long-term impact
