Neil Gershenfeld: Self-Replicating Robots and the Future of Fabrication | Lex Fridman Podcast #380

Lex FridmanLex Fridman
Science & Technology4 min read128 min video
May 28, 2023|1,208,216 views|14,465|1,372
Save to Pod

Key Moments

TL;DR

Neil Gershenfeld on self-replicating robots, digital fabrication, and the future of creation.

Key Insights

1

Traditional computer architecture has a physics mistake, as bits are not truly separate from atoms.

2

The Center for Bits and Atoms (CBA) at MIT bridges digital and physical realms, enabling complex fabrication.

3

Digital fabrication, inspired by natural processes like ribosomes, allows for self-replication and scaling of creation.

4

Fab Labs democratize fabrication, enabling personal creation and fostering innovation globally.

5

The future involves 'assemblers' and 'disassemblers' that can create and break down complex systems, moving beyond 3D printing.

6

Life's ability to locally violate thermodynamics through molecular intelligence is a key insight into creation and evolution.

7

The greatest untapped resource is human ingenuity, and Fab Labs provide the tools for it to flourish.

RETHINKING COMPUTATION AND FABRICATION AS EMBODIED

Neil Gershenfeld critiques the historical separation of computer science and physical science, citing a fundamental error in Turing's machine where the head is distinct from the tape. This separation, propagated through Von Neumann's architecture, leads to inefficiencies in modern computing where moving data is a significant effort. He argues that computation is inherently physical, requiring space, state, and time, and that all models beyond this are fictions. This perspective is crucial for understanding the embodied nature of computation and how physical processes are fundamental to creating and processing information. This understanding underpins the work at MIT's Center for Bits and Atoms (CBA).

THE ORIGINS AND VISION OF THE CENTER FOR BITS AND ATOMS

Gershenfeld's journey, from wanting to attend vocational school to working at Bell Labs, highlighted an artificial divide between theoretical and practical work. The creation of CBA at MIT was a response to this, born from a desire to integrate disciplines that didn't fit traditional departments. The 'Things That Think' consortium and NSF funding for CBA provided the infrastructure to explore the literal translation of digital to physical. The vision was to assemble one of every tool to make anything of any size, bridging scales from nanometers to meters, and integrating input/output processes to understand the full spectrum of digital-to-physical transformation.

DIGITAL FABRICATION INSPIRED BY BIOLOGY'S RIBOSOME

The core idea of digital fabrication, Gershenfeld explains, is inspired by the ribosome, nature's molecular factory. Like a ribosome assembling proteins from amino acids based on genetic code, digital fabrication aims to construct complex objects from discrete, reversible parts. This approach moves beyond analog methods (like 3D printing) by embodying digital information directly into the construction process. The ribosome's ability to detect and correct errors, ensuring high fidelity builds, is a model for creating robust, scalable manufacturing processes that can potentially create structures as large as those found throughout nature.

FAB LABS: DEMOCRATIZING CREATION AND FOSTERING INNOVATION

The 'how to make almost anything' class at MIT led to the accidental creation of the Fab Lab network, now spanning thousands of locations globally. These labs provide access to digital fabrication tools, democratizing the ability for individuals to create. Gershenfeld draws parallels between the impact of personal computers in the mini-computer era and the potential of personal fabrication. The Fab Lab network serves as a platform for personal expression, education, and local manufacturing, fostering innovation in diverse communities and reducing reliance on long, complex global supply chains. This network is transitioning towards machines that can make machines, further accelerating creation.

SELF-REPLICATING AUTOMATA AND THE FUTURE OF ASSEMBLY

Central to Gershenfeld's vision are self-replicating automata, a concept explored by Von Neumann. These are machines capable of creating copies of themselves or building more complex versions. This principle is key to scaling manufacturing capacity, mirroring how ribosomes build the organisms they are part of. The evolution of fabrication is moving from additive (3D printing) and subtractive (milling) to assembly and disassembly. This shift aims to create systems where discrete parts can be joined and unjoined, leading to 'assemblers' that can build complex technological systems and potentially even new forms of life or complex structures in space.

EMBODIED AI, MOLECULAR INTELLIGENCE, AND UNIVERSAL COMPUTATION

Gershenfeld posits that the next frontier in AI is embodied intelligence and molecular intelligence, moving beyond current computational paradigms. He connects this to fundamental physics, suggesting that information and computation are the root resources of the universe. Life's ability to locally violate thermodynamics, akin to Maxwell's demon, is enabled by this molecular intelligence. This intelligence operates through complex developmental programs, not just simple mutations. The concept of computational universality, where simple physical systems can perform complex computations, extends to fabrication, suggesting that complexity can emerge from simple rules and building blocks, mirroring biological evolution.

ADDRESSING RISKS AND UNLOCKING HUMAN POTENTIAL

While the potential for misuse exists, Gershenfeld argues that the openness of Fab Labs fosters transparency and collaboration, acting as a countermeasure against nefarious uses. He believes the greatest natural resource is human ingenuity, and the Fab Lab network provides the means for this capacity to flourish globally, transcending geographical and societal divides. The challenge is not just in technological advancement but in societal re-engineering: how do we live, learn, and work when anyone can make almost anything, anywhere? This democratized creation holds the potential for profound positive impact, sustainability, and a deeper sense of meaning through shaping our environment.

Common Questions

Gershenfeld claims Von Neumann and Turing made a fundamental mistake by separating the 'head' (processor) from the 'tape' (memory) in computing. This separation leads to inefficiencies and a fictional view of computation, where bits are not constrained by atoms. Both, however, ended their lives studying the embodiment of computation, suggesting they realized this limitation. (Timestamp: 105)

Topics

Mentioned in this video

personDavid Borden

A legendary electronic musician who was the first to create electronic music, inspiring Neil Gershenfeld's thinking about the computational capacity of musical instruments.

organizationMedia Arts and Sciences

A department at MIT, essentially a 'department of none of the above,' where students in quantum computing and synthetic life could earn degrees.

personAmira and Miana

Students of Neil Gershenfeld who published a Nature Communication paper on robots made from the parts they are manufacturing.

mediaStar Trek replicator

A futuristic device mentioned as the ultimate goal of digital fabrication, where a digital description becomes a physical object.

personSherry Lasseter

Named after 'Lasseter's Law,' which describes the exponential growth of Fab Labs; runs the Fab Foundation.

personKelly Dobson

A sculptor and former student in the 'How to Make' class, who created a device that saves and releases screams.

productPDP (minicomputers)

Mini-computers spun off from MIT, part of the era that created the internet, on a workgroup scale.

personDavid DiVincenzo

A colleague who thought about early quantum computing and articulated the properties needed to compute.

conceptMicrofluidic bubble logic

A method of universal computation using bubbles in a fluidic channel, stemming from a failed project, demonstrating switching, memory, and logic gates.

mediaThe Game of Life

A cellular automaton game where simple rules lead to complex phenomena, referenced to illustrate computational universality.

personTodd Machover

A colleague at the MIT Media Lab who initiated a collaboration with Yo-Yo Ma to instrument his cello, leading to new insights into musical interface.

companyEllisis

A company that spun off from MIT, developing auto safety sensors based on electric field tomography, becoming a $100 million-a-year business.

personJoe Paradiso

A colleague at MIT who collaborated with Gershenfeld on the 'Things That Think' consortium and later on CBA.

toolNASA Langley's wind tunnel

A large wind tunnel at NASA used for testing aircraft designs.

locationCarol (India)

A location in the South of India where SuperFab Labs have been created.

conceptHox genes

Morphogenes in the genome that store developmental programs, not body plans, illustrating how complexity arises from simple rules and evolution.

bookEndless Frontier

An influential report by Vannevar Bush that advocated for public funding of research, but Gershenfeld argues its linear model of innovation was a mistake.

organizationLincoln Labs

A research center collaborating with a student on superconducting computers.

conceptGenerative design

A design approach where a computer is told what to do rather than how to design it, proving effective for topology optimization in structures.

personSteve Reich

An American composer, mentioned in relation to David Borden's legendary status in electronic music.

personPhil Ritmuller

Worked with Honda and NEC, recognized the potential of the sensor technology for auto safety, specifically for airbag control.

companyData General

A company in the mini-computer industry.

toolSuperFab Labs

More advanced Fab Labs with tools to make precision parts for machines, making machine creation even cheaper.

bookGingery books

A series of books on how to build a machine shop from scratch, used to illustrate the idea of a personal Industrial Revolution.

organizationRad Lab

A group at MIT that invented radar, credited with winning World War II, illustrating how science influenced the war effort.

conceptSchrödinger's equations

Equations in traditional physics, described as an information technology from two centuries ago, potentially not fundamental but a representation of physics.

conceptSingularity

A hypothetical future point where technological growth becomes uncontrollable and irreversible, resulting in unfathomable changes to human civilization.

mediaMetropolis

A movie referenced to illustrate the current state of computing, where software is seen as separate from hardware, similar to frolicking upstairs while laborers work in the basement.

conceptSelf-reproducing automata

A concept studied by Von Neumann, describing how a machine communicates its own construction, a fundamental idea for self-replication.

personBob Moog

Inventor of Moog synthesizers, a physics student at Cornell who invented electronic music.

personJoe Jacobson

A core colleague who helped launch CBA.

conceptDigital materials

The idea of materials composed of discrete, reversibly joined parts, where global geometry is determined by local constraints, like Lego bricks.

personKenny Chung

A former student of Neil Gershenfeld who, with Ben Jinette, made a morphing airplane the size of NASA Langley's biggest wind tunnel.

toolFab Labs

A global network of approximately 2500 digital fabrication community labs, founded accidentally by Neil Gershenfeld, enabling personal fabrication.

conceptHow to Make Something (class)

A highly oversubscribed class at MIT for students to use digital fabrication machines and combine them in projects.

productWhirlwind computer

The first real-time computer made by MIT, a precursor to mini-computers and personal computers.

conceptGray goo

A hypothetical end-of-the-world scenario involving self-replicating nanobots consuming all matter, dismissed as a non-concern in the self-replication context.

conceptGrover's search algorithm

A quantum algorithm demonstrated using nuclear spins in early quantum computing experiments.

personJosh Smith

Neil Gershenfeld's first grad student, whose thesis on tomography with electric fields led to a significant auto safety business.

companyDigital Equipment Corporation (DEC)

A company that missed the shift to personal computing, as stated by its head, Ken Olsen.

companyWang

A company in the mini-computer industry.

locationSami Village (in North Norway)

Referenced as an example of a deeply sustainable way of living, contrasting with modern consumption-driven societies.

personIke Trang

One of the people who realized that nuclear spins could be programmed to compute, leading to early quantum computing algorithms.

personTom Toffoli

Collaborated with Norm Margulis on a paper demonstrating computational universality in cellular automata modeling billiard balls.

productEDVAC

An early computer for which Von Neumann wrote a memo, applying Turing's architecture to build a machine.

conceptMorphogenesis

The biological process of how genes give rise to form, a topic Turing studied late in his life.

personJerry Wiesner

MIT's president and Kennedy's science advisor, who was frustrated by segregated knowledge and created the Media Arts and Sciences department to house interdisciplinary work.

productRFID

A technology discussed in the context of the 'Things That Think' consortium, used for identifying and tracking objects.

conceptComputer-controlled Manufacturing

Invented at MIT in 1952, a precursor to modern digital fabrication, originally for jet aircraft parts.

personBen Jinette

A former student of Neil Gershenfeld who, with Kenny Chung, made a morphing airplane.

personStan Ulam

Co-inventor of cellular automata with Von Neumann, used to simulate self-reproducing automata theoretically.

conceptLandauer limit

The theoretical minimum energy required to perform an irreversible computation, related to information loss, near which superconducting computers operate.

personMcCulloch and Pitts

Developed a model of neurons that led to perceptrons and later deep learning, though the field diverged from understanding how the brain works directly.

personChris Moore

Philosopher and computer scientist who showed that it's easy to make physical systems that are uncomputable, solving uncomputable problems.

concept3D printing
softwareDifferential Analyzer
toolPrime

More from Lex Fridman

View all 112 summaries

Found this useful? Build your knowledge library

Get AI-powered summaries of any YouTube video, podcast, or article in seconds. Save them to your personal pods and access them anytime.

Try Summify free