CDFAM NYC 2023 · New York · 14–15 June 2023

Roadmap to Replicators

Listen as a podcast

Abstract

The closing keynote address at the CDFAM Computational Design (+DfAM) Symposium, held in New York City in 2023 by Prof. Neil Gershenfeld, Director of MIT’s Center for Bits and Atoms.

Prof. Neil Gershenfeld, a leader in the integration of digital and physical domains, provides a closing keynote address that delves into the cutting-edge advancements shaping the confluence of the digital and physical worlds.

Prof. Gershenfeld’s unique laboratory at MIT’s Center for Bits and Atoms has been at the forefront of innovation, from pioneering quantum computing to digital fabrication and the Internet of Things.

In his keynote, Prof. Gershenfeld outlines a roadmap to the development of replicators that seamlessly combine digital and physical fabrication processes. He offers insights into the trajectory of these technologies and how they are poised to revolutionize manufacturing, material science, and more.

As the founder of a global network of over two thousand fab labs in 125 countries, and as the chair of the Fab Foundation and leader of the Fab Academy, Prof. Gershenfeld brings a wealth of knowledge and a global perspective on the transformative potential of digital fabrication and interconnected technologies.

This keynote is directed at professionals, researchers, and academics in the fields of computational design, additive manufacturing, digital fabrication, and material science.

Transcript

From YouTube’s automatic captions, lightly cleaned; expect some errors. Each timestamp opens the video at that moment.

Read the full transcript · 5,614 words

Foreign. Thank you, I’m delighted to join you it’s been a long day, so if you need to wave your arms or jump up and down or walk around, feel free. So I wanted and end the day by looking ahead to what what is my version of a road map to a replicator. And in doing that, I want to peel apart a bunch of assumptions that I’ve heard through the day that sound obvious, but you really have to revisit as we go through those stages.

0:37 So I never understood the difference between computer science and physical science. I’m in a TED Talk, I said computer science was one of the worst things to happen to computers, or to science, because it’s unphysical. And so CBA was created to work at that boundary, and it led to things like the first quantum computations. We were part of creating the first synthetic life, part of creating internet of things. And to look at the road map, let’s start with 1952.

1:06 There’s one Earth, there was one Whirlwind computer at MIT, and there was the first NC mill. So this was the birth of computer controlled machining at MIT for extra credit, if you want to read about it, MIT actually stole it. It would it’s an interesting and a very messy history I can tell you about later, but that was the birth of computer-controlled machining. My lab is this, you can think of it as descendant of that. I we wrote an NSF proposal to get one of every tool to make anything of any size, and that’s roughly what they funded. So nano, micro, meso, macro, input and output, and that’s what we use to do the kind of research I showed you.

1:45 Scaling ahead, there’s about a thousand cities. The Whirlwind computer at MIT was transistorized at Lincoln Labs, it was commercialized as deck, that was used to create the internet. So this is Ken Thompson and Dennis Ritchie at Bell Labs inventing Unix, and so if you have Android or iOS in your pocket, you’re using what they’re doing in that picture you’re using the operating system from that picture. And the analogy to that is Fab labs. And the way I began thinking about that is I had all these tools, it would take a lifetime of classes to learn to use them. So with colleagues I started this class, modestly called how to make almost anything, that was aimed at just a few research students, and I wasn’t prepared for hundreds of students to show up begging to get in, saying, all my life I’ve been waiting for this, saying, this seems too useful, are you allowed to teach it at MIT?

2:35 And then what was surprising was what they did it was just to teach them the tools, but they did projects to integrate them. So Kelly was a sculptor with no engineering background, she made a device to save up screams and play him back later when it’s convenient. This is a web browser: Pirates have the cognitive ability of a young child, that lets the parrots surf the Internet. This is an alarm clock you wrestle with and prove you’re awake. This is a dress with spines, instrumented so if creepy guys show up, it protects your personal space. And these happened year after year, and I realized that I was asking how to do digital fabrication, but not why. The students were answering: the killer app of digital fabrication is personal fabrication, not mass manufacturing it’s manufacturing for a market of one person.

23:11 And so that led to this accident. NSF told us we had gotten a lot of money, we have to tell people what we’re doing. We thought it was more interesting to use the tools than talk about it. So with a grand old man of community activism in Boston, who recently passed away, Mel King, we set up a Fab Lab. He pioneered community cable, tail division, community computing, community internet, and inner city Boston. And so it was a natural step to go from providing access technology to providing the means to create it. So instead of millions of dollars of machines, these were the tools you need to do the how to make class the 3D printer, but the nine other tools you need to make complete functional systems. So large format machining, precision machining, surface mount rework, embedded programming, composites, molding, casting, and with that you can make all of this.

And we set up just one of them, we thought we’d go back to the lab and go to work, and what happened was there’s now 2500 and 125 countries they’ve, every year and a half, they’ve gone virally all around the world. We didn’t plan that. You can’t wake up in Cambridge and decide rural India needs these tools. This is one with at-risk youth in Detroit. This is Bhutan’s prime minister, where the country is based on gross national happiness, but by crap, bias, crap trucked in from India. This is Holland, Palestine Israeli boundary in Israel, cooking, the tribal council, amazing cultural traditions in Alaska. NATO’s terrible alcoholism, suicide, unemployment, kids sit around playing video games, it connects digital technology with traditional craft, a Protestant Catholic boundary in Northern Ireland. All, we didn’t plan that, but every time we opened, when somebody else wanted one, providing access to the means to make, in all of these communities.

Interestingly, we helped a number of companies set up networks of them Airbus, de so, GE, Chevron. They did, they thought they were doing it as outreach, but more knowledge was coming from the kids back into their engineers. That was then, was going from the kids, from the engineers out to the kids.

So then I started, we had Brighton event of kids in African shanty towns, in rural villages, who would be doing the work of my MIT classes and then fall off a cliff and be told, you’re smart, you have to go away. So we started using this whole network as an educational platform, where instead of teaching a hundred students, we teach a hundred labs all around the world, spanning from elite universities to these, you know, overdeveloped, underdeveloped communities.

And that led to things like this was a lab we ran at the White House in the Obama Administration. This is a Barcelona’s mayor, a fabulous design sense in Barcelona, over 40, over 50 youth unemployment, a whole generation can’t work effectively. And so my colleague Vicente gayart, who is Gowdy’s successor, started investing in these labs as urban infrastructure. They described the city as a product to trash conversion device, is how it currently functions, they instead want data to go in, data to go, the bits to come and go, but the atoms to stay. So they started investing in that as urban infrastructure, that’s grown into a Fab city commitment, of Mexico City, Paris, cities all around the world signing up, not just to be smart, but having means to make, as urban infrastructure. So instead of jobs equals work equals money equals consumption, it’s not utopia, it’s not free, but you change the notion of power and economy functions by providing access to these tools.

And this is really interesting legislative in the House and Senate for universal access to digital fabrication as a new kind of right, that’s really unusually bipartisan, it spans red, blue, north, south, sort of east-west, all of those axes. And for time I’ll skip those. So that’s where we are today.

These Fab La, everything I just told you was an accident, we didn’t plan any of that. But in retrospect, the Fab Lab is about two tons, it’s about a hundred thousand dollars, it fills a big room, it’s the cost and complexity of the mini computer. And like the mini computer, it’s being used to figure out the equivalent of the internet: if anybody can make almost anything anywhere, how do you live, learn, work, and play, how do you we thought the technology was hard, in many ways that’s easy. What’s hard is how it breaks traditional boundaries in education, in industry, in aid, in entertainment, how you, how you catch up to the organizational structure for that.

So then from the PDP came the Altair, this prompted this was the first really personal computer. This prompted the Homebrew Computer Club to meet, that became Apple. It prompted the first software from micro Dash soft, was written for this, they eventually took out the dash and became Microsoft. That was the PC era and the birth of it. And we’re at the scale of millions of towns, and here the transition is, for the Fab Labs, we started developing faster, better, cheaper tools, so companies like form Labs, or shaper tools, or Fab light, or Bantam, or handy bot with shopbot, were all things that were spun out from making faster, better tools for Fab Labs, but these weren’t the point. This was the point: the real point was not rapid prototyping, it was rapid prototyping of rapid prototyping.

So these are all Fab Lab tools made in Fab labs, this is yens in Norway making large format tools, this is Danielle in Hamburg, this is his laser cutter, this is a beautiful precision machine from south of India. So these, you don’t go to the lab to get access to the tools, you go to the lab to make the tools. And to do that we’ve had to revisit the boundaries between CAD, cam, machine control, and motion control. So one of the things we do is a G-Code, and anything like that is the enemy, because it enshrines a historical division of labor, and if you want to change anything in the machine, you need to change the machine, this configuration.

So instead what we do is we take the the leaf nodes in the network, every motor, every actuator, we connect them in real-time networks, and then we we build virtual data flow controllers over these real-time networks. So the machine is stateless, it’s similar to end-to-end principle in Internet architecture, and then the control, the intelligence of the machine, is distributed over it. And so this community is worked mightily on what comes after STL, that needs to be done, but I would argue what comes after that is, is nothing, what comes after it is networks that we don’t want stayed in the machine. You don’t want assumptions about what the machine is, you just want to expose the degrees of freedom and allow geometry to talk to those degrees of freedom.

Within that, then, lets you do is rapidly change the configuration of the machine, and it lets you control loops at many levels of description. And so I view better file formats as transitional, what comes after that is just the machines are stateless, you expose the underlying degrees of freedom, and you build virtual controllers over that. And so this paper is an example, and there’s theses through CBA site on building those architecture.

So at this stage in the evolution, then we’re doing things like this project, we set up a number of these Fab labs in Bhutan. A Fab Lab can make a machine, but it can’t make most of the parts that go into the machine, so we’re deploying super Fab Labs that have tools closer to what I have at MIT, that can make not just the machine but the parts of the machine. So this is bhutan’s prince doing his first official function opening the lab, but now there’s things like this, on flatbed cutter, and the edms, and the kind of tools that let you make the parts of the machines.

One fun thing that spun off from that is a haystack Mountain School of craft in Maine is a, or the, premier craft retreat in America, where the leading glass blowers and printmakers and blacksmiths and fiber artists retreat, and a few years ago we set up a Fab Lab there. It was initially very controversial, I describe it, when Dylan went electric at Newport, it was described as half of the audience was electrified, the other half was electrocuted, and that was like that at Haystack, half of the crafters were completely horrified we were bringing in technology, and the other half were horrified at the first half, not recognizing craft is full of technology, and there’s good and bad technology. Then by the next year everybody forgot about that and just fought about time on the laser cutter, but the machines don’t match well the needs of the artist.

So now what we do every year is we take over the campus with artists and scientists and engineers, and instead of teaching them to use the machines, we design machines for art, we build a range of new sort of machines to create. And earlier today there’s a mention of the nist materials genome initiative, we’ve been working very closely with that, that’s the national initiative on advanced materials, but there’s a problem, there’s a big outer loop from developing a material to testing the material to modeling the material to designing the material to learning to process the material to finally using the material.

So we started applying the rapid machine building to material science with them, building open material science instrumentation, and in particular, once the machines are stateless you can then start closing high-level control loops. And so for example, making printers where we measure the forces and flows, and then rather than doing reality offline, to expand use of recycled and renewable materials, the printer builds computational materials as it’s printing, and it learns reality online. We’ve started to describe this for nist as, as computational metrology. So instead of measuring a parameter, you measure computational models by closing loops in these machines.

And so to do all of that, and for time I’m just skimming, in this tour with DARPA, Jan vandenbrun, who ran design at simulation modeling at Boeing, came to the conclusion those who were broke, and he went to DARPA to reinvent design. And we worked with them on morphogenesis, on propping, propagating back into the design tools the implications of these sort of machines. What we did in that is, it sort of tautological, the design represents a thing, but your body doesn’t do that. Nowhere in your body is the number of five for your fingers, what your body stores in one of the oldest parts of the genome is a developmental program, at the end of which you have five fingers, and it does it for a very fundamental AI reason, which is, most changes in the genome are fatal or inconsequential, but not interesting, in these Hox genes, these developmental programs are a basis where search is interesting.

And so in this project we looked at how to represent design through developmental programs, not as explicit representation of the geometry, that in turn was limited by the physics engines. In this DARPA program we filled in the space between molecular dynamics and pde solvers with universal particle systems, to make mesh-free modeling of complex geometry, and papers in all of those.

So to keep scaling, there’s billions of people on Earth, billions is now the smartphone era, and to look at that, this whole meeting you have, without really thinking about it, has, has been analog. Nothing has changed from what I’ve heard so far, and it’s true almost everywhere, from the 1952 NC mill, where the computer controls the milling machine, what’s changed is you might deposit versus cut, but the information is in the computer, not in the material, it’s fundamentally analog.

Now consider a child playing with Lego bricks. When you play with Lego bricks the tower is more accurate than the motor control of the child, because you can detect in errors in the construction rather than accumulating them. You don’t need a ruler to play with Lego, because the geometry comes from the parts, you can join Lego bricks made out of dissimilar materials, and there’s no trash in a Lego playroom, you can disassemble because the parts contain enough information to not just assemble them but disassemble them, so that’s fundamentally a digital process.

And that idea isn’t new. Shannon showed how to communicate reliably with an unreliable communication channel, Von Neumann applied that to computing reliably with an unreliable computing system, and there’s very interesting history for how that came about, and that’s all based on threshold theorems. Digital isn’t one and zero, digital is a scaling property, if the noise is above a threshold you’re doomed, if the noise is below the threshold, for a linear increase in physical resources, the error goes down exponentially, and there’s very few exponentials in engineering, so that’s digital communication and computation.

And that led me to be interested in this idea of digital materials, that are digital not in the design but in the construction, so discrete sets of parts, reversibly joined with local geometry, giving global constraints. So one of the first places we did that, I was approached by spirit arrow, Boeing’s factories, about the problem of how you glue together big composite parts, which are vulnerable, vulnerability in the composites, and they wanted a better way to glue them together, and what it led us to realize is, rather than eliminating joints we should make joints everywhere.

We showed by linking loops of carbon fiber, you make an effective metamaterial that increased the modulus and the ultralight by an order of magnitude, with making macroscopic, not tiny, research samples, in a purely reversible process. Then here we showed, starting from the same base material, that’s rigid, that’s flexural, that’s oxidic, that’s chiral, we can design very anisotropic material properties just by composing those couple building blocks.

We use this with NASA, this is Langley’s biggest wind tunnel, to realize a long-standing goal, a purely morphing arrow structures, not by building a mechanism into it, but by effectively turning it into a mechanism. This was a high efficiency race car with Toyota, where instead of a year of tooling, it was done with a direct rate assembly process. MIT years ago made robotuna, this amazing swimming robot that had thousands of parts, this was in effect, not quite, but almost, an afternoon process to make a high efficiency swimming robot out of these very anisotropic structures. This is load bearing in one direction, soft in another direction, by the code of how they’re assembled.

24:35 And then the discretization of these parts propagates in a very interesting way to how you build with them. These are little robots that locomote on the structure to navigate, they error correct on the structure, and then by doing local operations they can build globally on the structure, and so it’s really a kind of a hybrid material robot system, you don’t need a gantry to make a big thing, you need these little robots to make a big thing. And what we’re about to write something up on, to reach the material strength limits, we’ve been doing this in metal plate lattices, and so this is an offshoot, an art installation from the design in the metal plate lattices, this, this ball is about 10 feet tall, and a fun detail is, was about five hours from sheet stock to being finished in this kind of assembly process. So that’s all promising.

So then there are trillions of things on earth, in the Internet of Things era, the Nest thermostat, not metaphorically but literally, has the power of the PDP computer, and here there’s a scaling issue. So you’ve heard a lot about AI today, if you look at the numbers, as of about a, a year or two ago, a supercomputer could do 10 to the 17 ops per second, and a brain could do 10 to the 17 ops per second, and so the real thing that’s happened isn’t a breakthrough in the algorithms, it’s really just the scaling, we’ve caught up to brains, and we’d be sort of derelict if it couldn’t roughly do what a brain does.

The, you know, the exascale computer has 10 to the 15 transistors, the petabyte storage is 10 to the 16, which is about the order of the number of synapses, so the real threshold is just the computational complexity matches the brain. But if you look at the fab making the supercomputer, it’s doing about 10 to the 10 parts per second, you, as you sit there digesting your lunch, you’re doing 10 to the 18 parts per second, so there’s eight orders of magnitude in fabricational complexity, in sort of molecular intelligence.

So the insight there is a little earlier, it’s four billion years ago old, that’s the evolutionary age of the ribosome, and the key, inside, one of the key insights of the ribosome is, ribosomes make ribosomes. So a ribosome is slow, it adds one amino acid a second, but you can make an elephant one amino acid a second, because a cell can have trillions of ribosomes, and then that adds up to this enormous throughput, this, this 10 to the 18 parts per second comes from the recursion of ribosomes making ribosomes.

The other insight is digikey stocks hundreds of thousands of electronic part types, you’re based on just 20, the amino acids, and what’s interesting is they’re not interesting, that they’re just typical, they’re hydrophobic or hydrophilic, or basic or acidic, but that’s enough to compose them to make the molecular machinery that makes you possible, so you can design with them. These are projects we did designing with that basis set, but you’re limited to the material properties of molecular biology.

So in molecular biology the coding sequence becomes geometrical motifs, that become building blocks, like electron donors or acceptors, that then become molecular machines like the ribosome, so that’s primary, secondary, tertiary, quaternary structure. And that led me to think about why can’t we do that with engineered materials, so why can’t we, in effect, reproduce life from non-living materials, but instead of having a code describe a thing, reproduce this idea of the code becoming the thing.

And so this was a paper where we took the robotic assemblers I showed you, and then I’ll just show you the design tool, we showed how the robot can make the robot, so making, by making the robot out of the parts it’s assembling, the robot can locomote on the structure, but it can take a feedstock of the parts and assemble another robot, and then that robot wakes up and can build another robot, and it leads to a really interesting design problem, of a robot can build a thing, it can build another robot, or it can build a bigger robot, and so in, in this paper, in this paper, we talk about that design problem.

So then from there, this is a micro assembler, and what the micro assembler is doing is it’s building electromechanical systems, so it’s placing rigid, flexural, magnetic, inductive parts, and one part is rigid, two is parallel transport, four of them becomes a four bar linkage, one magnetic bit is one bit, two becomes cyclical motion, then it becomes micro motors, and then you can make locomoting micro robots, and so that builds the electromechanical systems.

And then these are machines we’re making that are across between a pick and place and a 3D printer, and they assemble the construction of the computation of the intelligence in the systems, so, so they construct, there’s a traditional division between software and hardware, but here we’re compiling the computation into its construction by building these computational building blocks.

And so that leads to this hierarchy, from these base, just fundamentally discrete materials, and I’d say the, the, the, the, the, the two message, first message from today is, kill off file formats and replace them with networks, and the second one is, the really interesting transition isn’t analog, is an additive versus subtractive, the really interesting one hasn’t happened yet, it’s, it’s assembly and disassembly, it’s, it’s, it’s the discretization, the digitization, and that’s every bit as profound in construction as it is in communication and computation.

It’s a, as a footnote, I was at Bell labs, and I overlapped at the very end of people who had been around when Shannon was there, and they explained, when Shannon came and explained why the phone system should be digital, there’s a lot of fighting over analog versus digital in the phone system, and from Bob lucky I learned one of the greatest management lessons, he explained the battle was won by death, that, that the analog managers died, and then the phone system became digital.

24:41 But so there’s this hierarchy of the discrete parts, make the micro mechanical systems, that build the micro robots, that build parts, that build the bigger robots, on through the stack, and so I’ve shown you examples of work at each of that scales.

24:54 Now I started the talk by complaining about computer science as being unphysical, but the canon of computer science sort of misses the origins of computer science. So John Von Neumann did not write beautifully about the Von Neumann architecture, it’s in one dreadful memo, but he wrote beautifully about, profoundly about, a number of things, and at the end of his life he studied self-reproducing automata, how a computation communicates its own construction, and he did that to get it the heart of what is life.

25:30 Turing’s machine was never meant to be a computer architecture, and in effect it turned into one, it makes a fundamental physical mistake, the head is distinct from the tape, so persistence of state is distinct from interaction, which is why this computer spends so much time moving data from memory transistors to processor transistors that have the same computational power, but he ended his life studying morphogenesis, the profound question of how genes give rise to form.

25:56 So the foundation, the founders of computing, ended their life studying computational design from fabrication, but, but they did it not as the song, the software is in the machine and the machine is analog, they did it in this much more profound sense of how you merge communication, computation, and fabrication, so the description becomes the thing, in morphogenesis and self-reproducing systems, that, that those were the ideas somewhat forgotten at the beginning of computing, we’re now at a point where you can really do them experimentally.

26:29 One way to see what that’ll look like is, I was science advisor on a Big Hero 6, and so we were doing inflatable robots, but we’re doing these micro robots, and so the micro robots in Big Hero 6 is one version of what that looks like. A more serious one is, this is a program we did for bonus material for the movie The Martian, where we’re looking at how do you actually build a civilization on Mars, and what’s called isru, in situ resource utilization, historically is assumed you need to do the whole Industrial Revolution, you need digikey with a hundred thousand resistors. You should at this point recognize the kind of pictures behind me, what’s emerging is you essentially want to create life in non-living materials, you want the analogues to amino acids, you need these 20 material properties, you can hierarchically compose.

27:16 Or if you want to look at the data, this is time, this is Gordon Moore’s initial data points that led him to forecast Moore’s Law, this is 50 years of Moore’s Law scaling, and then this is 10 years of the Fab Lab spread, as one measure of digital fabrication, this shows the exponential growth of digital fabrication, and there’s actually more data here than Gordon Moore had, and so it, it just the data, it’s rational to project 50 years of scaling of digital fabrication.

27:54 And then I’ve walked you through steps from machines making machines, to providing access to the machines, to rapid prototyping of the machines, to assemblers, to self-assemblers, and I’ll end with, when I talk about this people clap and say that sounds exciting. My younger brother led the biggest video game studio at Activision, my older brother led the national liberal relations organization, and they would both be in the back groaning, seeing that I’m being naive about how technology comes in society. So the libertarian vision for the birth of the internet didn’t anticipate spam, and fake news, and income inequality, and all the problems we’re struggling with.

28:33 And so we wrote this book, for me to describe this road map to the replicator, that’s not happening in series, it’s hap, sort of, all these steps I showed you are in the lab, but they’re coming out between now and decades in the future, but what the book is about is to anticipate barriers to access, inequality, literacies, rather than waiting for it to go wrong in 50 years, how we can really take seriously the scaling and shape this revolution today, so the skills of this community are absolutely essential.

29:02 And I want to leave you with thinking about not just file formats, but closing higher and higher control loops with machines is real-time networks, thinking about how, as this goes out into society, it’s not just products for consumption, but chasing the notion of how the economy functions, and ultimately about this transition, not from subtractive to additive, but analog to digital, in the materials themselves. It’s late, I won’t take questions, I’ll be happy to discuss at the reception, thank you.

29:43 Thank you so much, I really appreciate your insight there, and all we can look forward to, and I, can I have one question, if we have these, the ability to make self-assembling robots, how do we design them, and how do we design what they make it.

30:00 Okay, so yeah, there are a bunch of levels to it. One step in is we’re doing work echoing some things you heard today about, there’s this gulf between generative design over here and decades of work on geometry engines, so just connecting constraints with generative design is a valley, and like a number of groups were working on helping fill that in, as one step. Then the second step is what I showed very briefly about morphogenesis in design, and the particle system simulation, is to do a lot of what I showed you just today, we break, we work closely with groups like Autodesk and deso systems and SolidWorks, and we break all of their design tools, and we break all of their simulation tools, so we have to end up writing a lot of our own tools to be able to do that, and handling that complexity is what leads to things like I was describing about morphogenesis and particle system simulation, which, if you’re interested, in the reception I can tell you more about why we do those things.

But ultimately I’ll give you the deep answer, and then the practical answer. The deep answer is, evolution doesn’t do what most people think, that’s what I was alluding to with the Hox genes. Evolution doesn’t randomly perturb it, it, the heart of AI in part is just computational complexity, but a big part of it is, in a high dimensional space, search in a random direction is guaranteed to not find a low dimensional sub manifold, and the, the heart of AI is parameterizing a low dimensional part of a high dimensional space, and that’s what the genome does, that, that’s what the, the Hox genes, these developmental programs, do, they parameterize a search space where you can go from walking to swimming to flying, rather than randomly perturbing in the high dimensional space.

32:10 And so what we’re inching towards is learning how to build not naive evolution but real evolution in these systems, initially offline, with these high fidelity simulation engines I described, but moving that online’s in the system, so evolution, not as random search, but really evolution, deeply grounded.

33:39 But a big part of what I would say for the answer is, if you look at the, the thousand, million, billion, trillion scaling on the mini computer, you could do everything you could do today with the smartphone, but it filled the room and cost a hundred thousand dollars. In the same sense you can already make almost anything in a Fab Lab today, but you still need a global supply chain to source it, year by year it’s going to get faster, better, cheaper, easier, more automated, but to a surprising extent you can do it today. And I, I thought things like how do you do design in these was hard, what really drives it is the, the wisdom of people, and so, you know, a big part of this global scaling is not machine intelligence, it’s human intelligence, it, it, it’s, it’s providing the tools in ways that you build on the wisdom of all of these communities.

So, you know, the deep answer is in a deep way understanding how evolution works, but a big part of the practical answer is moving design is something done by specialists for everybody else to empowering everybody to do it. Thank you.

More from CDFAM NYC 2023

Design Computation Human

Design Computation Human

Onur Yüce Gün · New Balance

Mud y Robots: Keynote Presentation

Mud y Robots: Keynote Presentation

Ronald Rael · UC Berkeley / Emerging Objects

Scaling & Standardizing Patient Specific Design

Scaling & Standardizing Patient Specific Design

Adam Wentworth · Mayo Clinic

Computational Design Ops

Computational Design Ops

Ronnie Parsons · Mode Lab

Overcoming the Design Dilemma

Overcoming the Design Dilemma

Ryan Carter · Johns Hopkins APL

Generative Design & Digital Manufacturing at NASA Goddard

Generative Design & Digital Manufacturing at NASA Goddard

Ryan McClelland · NASA Goddard Space Flight Center

Digital Materials – Geometry and Data Combined

Digital Materials – Geometry and Data Combined

Marek Moffett · General Lattice

Ten Years Teaching Additive

Ten Years Teaching Additive

John Hart · MIT

10 Years of Applying DFAM

10 Years of Applying DFAM

Tim Simpson · Penn State University

MfAM Walks so DfAM Can Run!

MfAM Walks so DfAM Can Run!

Chelsea Cummings · The Barnes Global Advisors

Informatics and AI for Materials Development

Informatics and AI for Materials Development

Wes John Alder · Citrine Informatics

Predicting Fatigue Life in AM Metal Parts Using DigitalClone

Predicting Fatigue Life in AM Metal Parts Using DigitalClone

Behrooz Jalalahmadi · Sentient Science

Designing Material Properties with Advanced Toolpath Generation

Designing Material Properties with Advanced Toolpath Generation

Alex Roschli · Oak Ridge National Laboratory

Design Automation Using Machine Learning & Computational Design

Design Automation Using Machine Learning & Computational Design

Rhushik Matroja · Cognitive Design Systems

Learning to Generate Shapes

Learning to Generate Shapes

Karl Willis · Autodesk Research

Design for Artificial Intelligence

Design for Artificial Intelligence

Chris McComb · Carnegie Mellon University

Register for Updates and Discounts on CDFAM events.