Embedded - 163: Syringes That Give You Cake

Episode Date: August 3, 2016

Nadya Peek (@nadyapeek) joined us to talk about making machines that build things.  Nadya's website is infosyncratic.nl, which includes her blog. Nadya's dissertation defense on Making Machines that Make: Object-Oriented Hardware Meets Object-Oriented Software was standing room only. MIT Center For Bits and Atoms, which studies "how to turn data into things, and things into data." Mods.cba.mit.edu Machines that Make: MTM.cba.mit.edu

Welcome to Embedded. I'm Alicia White here with Christopher White. We are pleased to welcome Nadia Peek to the show to talk about building things that build things. Hi, Nadia. Welcome to the show. Hi. Hi, Chris. Hi, Alicia. Could you tell us about yourself? Yeah, I'm Nadia Peek. I just defended my PhD at MIT in the Center for Bits and Atoms.

The title was Making Machines That Make. Object-Oriented Hardware Meets Object-Oriented Software. It was mostly about designing infrastructure for machine building. And I work a lot in the Fab Lab network. I've set up many, many Fab Labs. And as part of that, I guess, assembled many, many shop bots. And I also spend a lot of time just in the basement by myself at MIT. So the two extremes are a nice contrast. Because the Fab Labs are going to have you interact with so many people as you get them set up. Yeah, I think that's the fun part of it.

I like to think of Fab Labs kind of like libraries, like you go there and you have access to digital fabrication tools as opposed to books. And it's nice because if you go to a Fab Lab and you don't otherwise have access to digital fabrication tools at home, you can still share them. But it does mean that you have to engage with lots and lots of other people. And Fab Labs is a particular name. I mean, we have tech shops and like the supply frame design lab we talked about a few weeks ago. And other hackerspaces. And various other hackerspaces.

But Fab Labs is one particular type. Yeah, so I've set up a lot of Fab Labs. And Fab Labs share a charter for how they operate. And everyone has their own model. And so makerspaces and tech shops everyone has a model of how to sustain yourself and how to provide access um and in fabloves it's just a like a specific set of things that they all share like they're open two days a week um to the public for free uh and they generally have the same basic set of machine capabilities. We have a shared inventory across them.

But I've also set up or helped, I've also helped set up makerspaces that specifically don't identify as Fab Labs because they were unable, for example, to be open two days a week to the general public because they were in a university or something. And so they could be open to the students, but they can't be open to everyone. So yeah, I think that the more we can get of tech shops, makerspaces,

hackerspaces, fab labs, the better. So there's no need to put too fine a line on distinguishing between them. Okay. Well, we will talk a little bit more about that, but a lot more about your PhD thesis, I think. Before that, we're going to do lightning round. Have you considered adding a sound effect to that? You know, the lightning. Anyway, lightning round. Consider it done.

I can make some noises. Pew, pew. We're going to ask you some questions and we want short answers. And if we are behaving ourselves, we won't ask for explanation okay that sometimes happens this is going to be a bit of a long um one because we have a listener bruce who added some extra questions specifically for you christopher all right we'll start with the usual one. Favorite electrical component? I like the Allegra 4983 stepper motor controller a lot. That's very specific.

It is. It's nice. Much better than capacitor. What is your least favorite component? Well, I really dislike the 3482 i think the same also an allegra stepper motor driver that keeps blowing up so so maybe that's why i have extreme opinions about them both because one would blow up all the time and the other one never blows up so okay how about favorite fictional robot robot? WALL-E? Cool. Favorite type of actual robot? Atlas? What's Atlas? Atlas is Boston Dynamics

humanoid robot. Atlas is kind of scary. That's the one they beat up with the broomstick? Yeah. He can get up from the ground that's very cool yes after he's beaten yeah yeah he comes after you i guess well a lot of the robots that i want to think are really cool you know that um i don't know what this robot is called but i saw i think maybe it was um mitsubishi or toyota is making a robot that bathes elderly people.

But I think it was all concept. I don't think it was actually ready yet. So once we have a robot that bathes elderly people, that would be my favorite robot. Yeah, they do have the robot harp seal that acts as a pet and monitor for elderly in Japan. Yeah, I think we're going to need a lot more of those kinds of caretakers soon okay next question um additive or subtractive manufacturing techniques well as i will maybe elaborate on later in the show i think that automation kind of makes the question additive versus subtractive unnecessary. But if I really had to choose one, I would go with subtractive.

Yeah, that's more fun. It's also more precise. Stepper or servo? Well, depends on the load, I guess. Do you expect it to steps? I'm going to go with steppers. Servos are often overpriced. I don't like paying too much for stuff.

Science, technology, engineering, or math? Technology. If you have to choose one direction, would you design a system? The band? No. Sorry. I would definitely not choose the band. One pathway. design a system the band no sorry if you had to choose one pathway would you design a system that is more flexible for an expert user or one that is easier for a newcomer to learn

are those things mutually exclusive usually given a limited amount of time you have to choose one i would go with the expert user what would machine tool would you take with you to a deserted island what good is that going to do you no it specifies with three phase power and endless feedstock okay so it was part of the question you're on a desert island that has nothing but three phase power and endless feedstock what machine tool would you take with you? Cincinnati HydroTel. You're going to have to tell me what that is. Is that a water jet?

Oh, no, it's a mill. Oh, okay. It's a really, really old mill from the 50s. I bet it wouldn't break down on a desert island. So, I have to ask now what you're going to mill

on the desert island? Well, you said I have endless feedstock. Yeah. So, I'm just going to take it from there. Okay. Favorite physical constant. Hmm.

I don't know. No favorite physical contract. Maybe a coulomb. Can we call that a constant? Charge value? Sure. Oh, I'm lost now. That wasn't in the list.

It was in the list. Which is most important to your job? Whiteboard, soldering iron, or keyboard and mouse? I use a keyboard and mouse almost. I mean, soldering iron is pretty important. Whiteboard is not very important. But keyboard and mouse use quite a bit. I don need really the mouse just the keyboard yeah yeah he added mouse i'm not so sure about that when you were young what did you want to be when you grew up um

i think i wanted to well i wanted to be a popsicle inventor. I thought you were just going to say popsicle. I want to be a popsicle. I have a very detailed notebook full of popsicles that I was going to make. I think when I got a bit older, I think I wanted to be a chemist. So that changed a lot. I don't know if I have a bit older, I think I wanted to be a chemist. So that changed a lot. I don't know if I have a good answer for that. One of the people in our graduating college classes ended up as a math professor, and he started a popsicle business.

He does. He still has it. Weird popsicle business. So there's still hope. You can still do it. Oh, yeah. A friend of mine also just started a popsicle business. She picks up kind of fruit that no longer is beautiful enough to be sold

and makes it into popsicles. Okay, bits or atoms? Atoms. Is this exactly like your PhD defense went? No. That's too bad. Is this harder? You can tell us.

We understand. This is much harder. Although I did have a really long question that was about security, trustworthiness, privacy, and ended with da-da-da vis-a-vis Pokemon Go. That was a hard question. Anything that ends with Pokemon Go. Pretty hard. Okay, I think that's enough for lightning round.

I do sometimes ask making or engineering as part of that, but it's a longer question for you. So let's take it as a longer question. What are you more interested in for your career, making or engineering? I guess I'm not, or as indicated by my past, I think I'm not totally sure what the difference is. But I'm going to go with making. I think I like the sound of that better. You did just defend your PhD thesis. Do you already have results from that or is it? Yeah, I am provisionally Dr. Peak as long as I, you know, actually submit my dissertation. I think it's all going to be swell. That's great.

Congratulations. Thank you. Do you know what's next for you? Yeah, I still work here. I am not switching desks or anything. Cool. Cool. And there is the MIT center for bits and atoms. Yep. I'm in center for bits and atoms. Um, although now that we know that you're biased for atoms, I'm not so sure they're going to keep you. Uh,

she can break, break it off as a splinter group. They, they have a, they have, they have space for both types here, but people and Adam people and bits don't take a lot, a lot of space.

Yeah. They all sit upstairs and atom people sit downstairs. Makes sense. The site says it studies how to turn data into things and things into data. Yeah, that's a pretty... I guess it's a poetic way of putting what we... I mean, I think there are many levels in which we do that. There is, if you think of programming a machine tool to make a thing, you can think about that being turning data into stuff. And if you think about going from, so one of the research projects that's pretty big here is one of digital materials so

instead of for example additively manufacturing by squeezing out hot plastic or curing um curing resin uh you assemble parts out of say tiny lego um and so there you're programming a material that is being assembled out of tiny ligo or maybe programs can be put into materials those are all things that people work on here how can you uh one of uh one of the center for bits and items students noah jackmo he's uh he's working on using using DNA as a kind of storage media. So instead of, since we can now make our own DNA, why wouldn't you just want to back up your computer in DNA? So it's really small, very portable. So yeah, there's all kinds of different ways in which you can interpret how the Center for Bits and Atoms works on going from data into things and things back into data. And for you in particular, it has to do with building

machines that can build things. Is that right? Yep. So I build machines that are computer controlled or even controlled by things that are less maybe don't entirely qualify as a computer oh i understand that i often say i write software for things that aren't computers yeah i do i get that and your your phd dissertation was making machines that make object-oriented hardware meets object-oriented software. What does object-oriented mean in those contexts? Well, I borrowed heavily from the paradigm of object-oriented programming to understand how we can make it as easy as possible to build up machines. And so if you think of a mechanical component

of a machine being an object that has state, so it has data, it has position, or it might have a method as well, like you could make it move back and forth. And you also think of a software object that maybe can control or can request new position. You can think of those as two different objects that can pass messages between each other.

And if we wanted to add a couple more objects, like more motion stages to build up a more complicated machine, then you can think of an assembly of objects that together form a machine with an end effector and some interface and then an infrastructure of objects um that makes the machine which makes the machine easier to reconfigure or maybe easier to make application specific the other thing i uh borrowed from uh the other paradigm i used heavily was um the end-to-end principles for networking, where in the early days of the internet,

some smart people, including David Clark and David Reed and Vint Cerf, argued that you shouldn't implement too much in the network itself. Because for example, if you wanted to check if your data had been corrupted in the middle of the network, it could still get corrupted again later. So all the things that are specific to the application should be implemented at the endpoints and everything in between should just be as simple as possible shared infrastructure. So you have machine objects, in my case, you have machine objects that you can then assemble together

to make a machine platform that you put an end effector on and a software interface into so that you can have an application specific machine. How does this differ from modularity? It is modularity. I could also call it modular, making modular machines that make or it's a i think that the um the reason i i used a object-oriented programming as a metaphor was because the you know i don't necessarily just want to connect everything in one um in one like connection point there the way that um the way that you know if you have like

a modular mechanical kit that you can snap together and distributed controls that you can also connect together but they don't necessarily then you have to connect the control system to the mechanical system and the mechanical system to the solver system you still end up with these kind of choke points in between the different layers of description um so i'm trying to avoid that by saying, well, you know, this control node goes specifically with this machine node. But, I mean, you could call that modular too.

It's also kind of a distributed system, right? You've distributed the controls throughout the modules rather than having a central brain. Modular, distributed. Yeah, there's many words to choose from, I guess. But yeah, I generally call them distributed network controls,

but I don't know if you said distributed mechanical system. I don't know if that would just make me seem crazy. Not to us. Okay, well, we can call it all the things. Distributed, modular, networked, object oriented assembly of things to make machines.

One of the difficulties of modular hardware is that it tends to be more expensive. I mean, the connectors alone aren't cheap. Yeah, that's true. So I mean, the connectors alone aren't cheap. Yeah, yeah, that's true. It's a, so one thing that, especially in machine building, you know, MIT has a very large mechanical engineering department.

And one of the ways that I explain to them why I'm breaking up machines into all these different parts where connections have opportunity to fail. It's never going to be as good as something that's purpose built. And so sometimes I try to bring up the analogy of packet switching networks. You know, packet switching networks were sort of ridiculed when they first came out because people were like, this is a really strange way to try to transmit data from point A to point B. It would be much better just to have a dedicated line. And at the time, no one imagined that there was going to be the Internet and there were going to be millions and millions of connections. It was just like, so you want to connect Stanford to MIT? Why don't we just put a wire?

And so I think that the unanticipated applications that you can build using modular components make it worthwhile to, it's definitely a trade-off, right? Like, yes, it's going to be less optimal and the connections might fail. But instead of you having to design a machine or in this fail. But instead of you having to design a machine, or in this case, machine, instead of you having to design a machine and spend maybe,

like say you were pretty fast at it, like say in a week you can build a custom machine to do something, it's still not going to compete with like in a few hours or even maybe minutes, you can put together a bunch of modules and start your automation process and so i think it's more of a how do you go from machine no machine to your first machine rather than going from um your first machine to a better machine maybe by then you should remove if you wanted to make the same machine over and over again, you want to reduce costs by removing connectors, reduce,

maybe make your structural loops smaller. Maybe, you know, there's a lot of ways that you could optimize something. But if the problem is that you haven't automated it at all yet, or you don't have access to that computer controlled precision yet, then having the drawback of having connections which might fail seems worthwhile. It definitely worked out in the case of TCP IP. So maybe we could take advantage of that. Well, like object-oriented modularity is something we often recognize as good in software.

And in hardware, I would think it is also very good. I certainly appreciate all of the breakout boards that exist and the eval boards and the dev kits so that I don't have to design a board for everything. Yeah, definitely. That makes a lot of sense. But there is a downside to it. And it's good to hear that sometimes you make a different decision. Have you been very successful in putting out modular systems that are buildable by other people? Yes, I would say so.

So one of the machines that I built a long time ago, or a long time ago, it was maybe 2012, 2011, together with Jonathan Ward, who now works at shop on, uh, was a plastic, a small plastic milling machine. It was like made out of cutting boards and it was for milling circuit boards. Um, and, uh, uh, and it was, uh, it was, it was, it was like a small, cute mill. And, you know, the, the cutting boards, the HGPE cutting boards that we, uh, that we bought come in all these different colors for restaurants. And it was like cute.

And we thought, and we made it open source so anyone could make it. And we thought, oh, everyone is going to make our, everyone's going to make our milling machine. And, you know, almost no one made it. It was really, you needed access to a big mill to cut out the parts. And then you needed to follow our assembly instructions, which were written for people, or like it was hard for us to write assembly instruction, I think, without watching. We should have watched someone who had never put together a machine and written instruction based on that or something.

I don't know. I don't know why the documentation was not clear enough. And, you know, it never ended up working well. And so that's one of the reasons I think that I started working on this much more modular approach. And so I built a bunch of modular machine components. aluminum folded uh folded aluminum um stages and uh milled aluminum rotary stages and uh distributed um control systems to go along with that but those were not necessarily that accessible because you needed those parts to build them um and so james coleman and i made a different version of it that was made out of cardboard uh and the cardboard machine kit

or yeah the cardboard machine kit was really pretty successful because it turned out everyone was comfortable cutting even people that didn't have laser cutters and didn't use the laser cut files were still able to make the cardboard stages um and uh it was very easy for, it was very easy for, um, it was very easy for people to build kind of a minimal viable machine and then, and then optimize after that. And so even though the cardboard machine and the cardboard machine kit and the parts that went into the MTM snap, had a bill of materials of about 350 dollars and if you had if you want to use say four linear cardboard axes that would probably be around 250 or something like that um together with the electronics and the uh uh together with the electronics and all of the hardware and stuff that you would need um it was just much easier for people to make the cardboard machine and then also to iterate on it. Once they made machine one, they were like, okay,

actually we want not to have a pen plotter, but we want a milling machine, so we're going to attach a spindle now. Or the modifiability was very successful. So far, I think about 200 of those cardboard machines were made. And as far as I know, only maybe 10 MTM snaps were made. So, we're making progress. And cardboard has a pretty low barrier to entry. I don't feel bad if I destroy cardboard.

I was going to say, yeah. But if I had to throw away even plastic, even... Or aluminum. Yeah, throwing away aluminum would be very hard but even the cutting board plastic i still would feel odd yeah yeah but cardboard sure that comes in and out my life so fast i don't care yeah everyone is used to it as a strong lightweight packing material right so it's readily available we use the cardboard or specifically for this machine design we use the cardboard that you buy for science fairs,

you know, like the tri-fold poster board, and it comes in different colors. I'm obviously very focused on things coming in different colors. And it fits, so one linear stage fits exactly on one tri-fold poster board. And so then it makes it easy for people to find the same cardboard but even if they didn't it wouldn't matter because the cut file for it is parametric so you can you can adjust it for different thicknesses of cardboard or sizes of stage so many of these tools start with a stage i mean if you talk about a laser cutter you are talking about a laser on

two stages maybe three and if you are talking about a 3d printer again you're a couple of stages yep um what people made cardboard laser cutters that was terrifying talk about chopping off your own foot there yeah yeah. There's all kinds of things. I was kind of like, yeah, you guys, good work, great work, but maybe think about this. They're like, it works great. We don't know. The cardboard isn't underneath the laser now, yeah? And I'm like, okay. And milling machines, CNC milling machines are, of course, stages.

So you have those and the plotters, again, stages. Now, many of those need some sort of spindle thing, mostly for milling, I guess, is where the spindle comes in. What other modules have you made and put out in the world? Well, so I'm working currently on like an end effector kit. I've built a bunch of different end effectors that can, so the cardboard stage has these four little nubby things on it, kind of like Lego,

so that you can stick them onto each other into different configurations. So say you wanted two axes, you just stick them on top of each other and they can move around. But the nubby thing also accepts for example an end effector so James and I built a plotter end effector 3D printer end effector

syringe end effector microscope for gigapan well syringe would be really useful for chemistry if you do little tiny micro stages and you can do a whole bunch of experiments at once sorry he was wincing. I'm imagining the robot in Star Wars with the syringe. Never mind.

Go on. Yeah. James used a version of the syringe robot to decorate a birthday cake for me once. So that, you know, you can think about scary syringes or syringes that give you cake.

Very precisely decorated cakes, good for everyone. This could really work back into popsicles. Yeah, yeah. See, like all this buildup and in the end, I'm going to be a popsicle inventor. Have I put other end effectors on these machines? Oh, yeah. Knives for, you know know swindle knives um swivel knives I mean uh I don't know what other end effectors I mean sometimes um what you want

to do is uh polishing but that wasn't really a different end effector we just put a polishing sponge on the end of our spindle because we're questionable engineers but yeah i think that making it easier for people to put different end effectors on the cardboard stages would be a great kind of next addition to the kit um because uh right now a lot of people have made hot wire cutters with the stages because you don't really need to do any kind of end effector control for that um but uh but so i use the distributed network controls um that use a library called pygashult um that uh was originally invented by elon moyer uh and i contribute boards to that or different boards to control different things um library uh and so i think if it were easier for people um to get different boards to control other things not just

um you know extruders or stepper motors but you know can can they attach different kinds of feedback loops um then that would be an easy way to get people to be able to more easily make more machines. And so that actually comes to the next question I had for you, which was about software. It is all well and good to make a stage and make some motors go back and forth and put an end effector on so that they do different things as they go back and forth. But how do you control all this? Great question.

So I think of the software interface as often one of the endpoints of the network. And there are a lot of different ways that you can control things. Like people who listen to podcasts about embedded hardware might have heard me complain about g-code extensively on a different podcast and that is definitely one way that you can control a machine you can use you can take a design file and run it through a processor that will be able to generate a tool path based on the object that you want to be fabricating. And that toolpath consists of coordinates or places that the machine needs to visit. And then you stream that file to the machine.

And that works really well for lots of machines, like milling machines or laser cutters. They can all operate well on just moving this following pattern. But I'm also interested in other kinds of software, hardware, feedback loops. So, for example, one that I use is putting a video camera on the machine so that it will behave differently based on what it sees. So if, for example, you're pipetting something into a Petri dish and you want to pipet until the Petri dish colors from blue to pink, then how do you write a control system for that process?

And so to be able to do that with my advisor, Neil Gershenfeld, I've been developing kind of a workflow composition tool that runs in browser called mods for modules. See, we like modularity. And mods, so mods, you can, it's all JavaScript. Took me a while to get used to writing javascript because it's all high level one crazy but uh uh you um you can you have different modules that you can

connect together um like like you might do in simulink or labview or grasshopper um and you can write modules that you can also put in so So, for example, if you wanted to do like a standard 3D print toolpath, you could import an STL and then slice that into a toolpath, generate coordinates, and then export it to a machine. And to be able to do that part, we're using a virtual machine to talk to the distributed control network and this is kind of a confusing use of the word virtual machine it's not like a actual virtual machine but it's the machine it's the it's a description that explains which of the distributed control nodes do which part of the work in the machine. So if you had two stages, X and Y, then there would be one control node that's moving the X motor and one control node that's moving the Y motor. But if you use, for example, a HBOT configuration, which is now in a lot of

3D printers, then moving the A motor moves the stage in both x and y directions so you need to move the a and the b motor just to be able to move x and the a and the b motor just to move y and so the thing that translates from how does the machine need to move to what are the motor commands that need to be executed is the virtual machine in this case. And so from the mods environment, which runs in the browser, there is a WebSocket connection to a Node.js server that then communicates with the machine over. Most of them run um rs485 uh that's a little bit that's

a part of the system that i don't like as much because i would prefer uh not one choke point of connection but i'm working on that so for now you could for example put through a 3d print tool path that way or um we also have, you can import an image and then do a threshold on that image, then vectorize that or like do edge detection and vectorize that image and then take those vectors as a toolpath for, for example, a milling machine. And also has a video, a video input so that you can compose all of these different programs out of modules or write other modules, like GPhoto2, the photo library that allows you to control a lot of different kinds of cameras remotely. If you wanted to remote image capture as one of the modules.

So it's kind of like a software composition framework for writing, for automating actions on your machines. And that hasn't been as widely used yet because I think it is not as well documented as it needs to be. And I'm also working on that. But if you want to test it, you can test it at mods.cba.mit.edu. mods.cba.mit.edu. Put that in the show notes. Cool. I'm using it right now.

That is not a friendly landing page, but we'll talk about that later. Yeah, see, these are problems you have when you don't interact with enough people. I know that MIT is also doing that thing, Beetle Blocks, which is sort of a CAD program. Are you familiar with that? Never heard of it. I want to look it up after this. Sounds awesome.

You mentioned, I think, Eric Rosenbaum, and he was one of the people on that. They were calling it Scratch, similar. For CAD, that's awesome. Eric Rosenbaum is very cool. Also for uh the makey makey uh yes uh okay so i'm not going to ask you about beetle blocks let's just uh cross all those questions off and all of these things that you're talking about building quickly and building simply and making modules, modular, making, yeah, all these things. You're talking about building tools. You're not talking about building things. You're not talking about my next TV or my next cell phone you're building the tools and not the end products yeah so I think so for

I think it is hard to anticipate what the infrastructure is you might need to make a thing so for example a 3d printer is considered to be like a universal thing right like a 3d printer can make almost anything except in reality like a 3d printer can considered to be like a universal thing right like a 3d printer can make almost anything except in reality like a 3d printer can't make very many things 3d printer can make kind of smallish things mostly plastic um and so what if i wanted like a coil what if i needed to wind a coil for a solenoid and it need to be like of specific resistance um you know if i had a bunch of different modules that i could assemble to be a coil winder um then i could i could have a kind of precise computer controlled process um and after i made the coils i needed i could reuse those parts in other places or um if I wanted to make uh if I wanted to mill a part um that was very long and skinny

uh then I might not have access to um a laser cutter of that particular size but if it was easy for me to assemble different modules with that work volume, I could do that. James, who I built a lot of the cardboard machine kit with, James Coleman, he works at Zaner Architectural Metals now. And they do a lot of the sheet metal fabrication for buildings like the Walt Disney Concert Hall or the Status Center at MIT, Gary Building, both Gary Buildings, or the Peterson Automotive Museum and so they make a hundred thousand parts for any given building and and ten thousand of them

are unique or ten of ten thousand unique styles or ten thousand unique designs and so for them you know they have to constantly either have to have like the biggest water jets or that they can, but even then, you know, it's hard to, it's hard to make, it's hard to make things that are that unique and still have an unreconfigurable shop space. think that the things were the application like you were mentioning like what why am i not talking about the things that you can make with machines but i have since then started to couple the machine and the thing together in the application so i think that to make a thing you need a machine you don't always need a machine you know maybe it's a very simple thing and you don't need computer control but for a lot of things to be able to make things that are complex or precise you want to have some level of precision in its control system so there then I couple the machine with the thing and so I'm thinking a lot about how to make making the machine part easier because right now I think

making the machine is kind of hard machines have gotten so much more accessible my my shop in college years ago was wood and then metals and then aluminum and I loved the shop and it was amazing but I always knew that I couldn't do it at home and then it wasn't that much longer before I started to see these tools where I could start to do some of them at home not just a saw and not even just a table saw that was home ready. More, I mean, we just got a 3D printer. We have a laser cutter on order. And we have all these neat tools that I couldn't imagine. And I know places I can go to get milling machines. And I look at the other mill and I drool every single time

to think about all the things you can build. Other mill, semi-based on the MTM snap. So these tools have gotten a lot more accessible as well as the idea of these tool co-ops like the fab labs and the maker spaces. Do you see this continuing? Do you see everybody having their own set of tools? Well, I think that's one way that you can think of how these things might progress. Lucy Suchman is an anthropologist who used to work at Xerox PARC. And she worked at Xerox PARC at the time when they first decided that copy machines were no longer going to be something that required dedicated technical staff but was going to be something that offices could have

and any user could go up and use them and so they had all these this is like 1985 so they had all this advertising Xerox did at the time where you're like it's just a big green button and go press the big green button but the experience that a lot of people had when they walked up to a photocopier was one of sheer rage and upsetness. And so Lucy studied that experience. Like, what is the experience that people have when they go up to a copier and then conclude that up in her dissertation and later turned into a book called plans and situated actions and so a plan is uh is like you go up to a copier and you use it for its intended purpose so you're you photocopy one-sided you photocopy two-sided and maybe a machine has a bunch of different plans like photocopying one side or photocopying

two-sided but then there's also all kinds of contextual information that you as a person have like what is all of what how is your action situated within a context and a lot of machines do very poorly with that especially the xerox machine at the time apparently did very poorly with adapting to situated knowledge of humans. And her work was of great influence to me because she kind of, or Lucy Suchman is credited with having invented human-computer interaction by studying people interacting with photocopiers and then stating that user, doing an ethnography of user experience is something that would probably be pretty useful for companies that were making machines. And that totally changed how companies like Xerox or other companies like Nokia, they regularly now employ anthropologists and ethnographers to figure out how to make something easier to use. But plans, I think, are a good way. Plans are a good kind of crutch, or it's a good way. It's very easy for me to walk up to a 3D printer

and 3D print something off of Thingiverse, for example. It's a very well-scripted user experience. Or if I want to go up to a laser cutter and cut something out there but um each one of these user experiences is very well tailored like if i walk up to a form one and i press its big glowing button you know it's all you know people think about what you're doing there but uh i do often feel like it's not all of the ways in which you veer off script, all of the ways in which you leave the plan. Those are still not totally well developed. So if you wanted to use your shop bot for assembling giant Lego instead of using it for milling, you might run into a lot of hiccups, even though, you know, like we discussed earlier, they're just a bunch of stages.

And so I think that yes, as market demand increases, we're going to see way better user experiences and lots of different kinds of digital fabrication machines, bringing them into homes or of computer control for and turn them into products. I think that people are more inventive and need more flexible tools than that. Well, beyond the tools, I think the other barrier is ease of design. Because right now, for things like 3d objects just a pretty big learning curve for okay how do i learn to do cad well you go here's a week-long tutorial you go to debra and you get a certificate whereas you know certain other things are not not so impossible like if you get a laser cutter okay and you're doing outlines people anybody can draw yeah um

so that's that's another axis i think that that's the needs addressing too is, okay, we have these machines, they're cheap, anybody can use them. But if you're just downloading something from Thingiverse. Kind of the amazing thing about digital fabrication is also that, you know, software is very easy to run or to write and then run and then figure out what's wrong with it. And then write another iteration of, but in machine tools the software seems to get like the least amount of attention in terms of upgrades that you can get in a machine. We have a SODIC wire EDM and the software is absolutely incredible.

Like you can only give it file names um with uppercase alpha that wasn't a great incredible yeah no it's a terrible it's like it's incredibly bad there's the file names can't be longer than six characters but it doesn't say that anywhere so if you have a file name that's longer than six characters just kind of silently fails um but they did upgrade like they upgraded the the panel with which you can look at their software that looks like it was designed in like 1987 um it's like a touchscreen and you can like you know squeeze pinch it uh and there's different parts of the touchscreen that you can touch and there's also a keypad and there's all these other parts but but like the software itself is incredibly incredibly bad um and in terms of like ways in which they as a machine tool company could

improve the user experience like writing better software seems like a pretty cheap and easy way um to to do that on already existing machines but um it just it's just not one of the uh it's just not one of the the the iterations that seems to happen a lot um and i think it's partially because the system integration going from the software interface into all of the different parts of the machine um is complex enough that you can't just kind of write a script for it so uh uh so so instead you have i have like a better example uh epilogue laser cutters i hope they don't get mad at me for saying this i don't think they will i was about to pick on them so go ahead so epilogue laser cutters so um for the tools in the fab lab includes laser cutters and milling machines we have this software that

we write called the fab modules different than mods but also modular and the fab modules is a is like an easy tool pathing program for lots of different machines you can also do some rudimentary design in it but mostly it's for tool pathing so you you start from a PNG and the PNG then can be exported to a milling tool path or a laser cutting tool path or a water jet cutting tool path. And so the idea is that you don't have to learn lots of different kinds of CAM packages to be able to use the different tools in the Fab Lab. And so the epilogue file format we needed that to be able to talk from the fab modules to the epilogue um and uh and so you know we uh print it so when uh when you're sending it from a windows machine an epilogue file we uh we printed that to file instead of exporting it out over um the usb cable

and looked at it and it's the end of the file has 10,000 spaces in it and uh and then we're like hmm well maybe that was an accident so we removed the 10,000 spaces and tried to send it from the fab modules and didn't work it needs the 10,000 spaces and then after having run epilog and so we just added them in like okay so we put in 10,000 spaces that's what it means and at some point we actually talked to an epilogue engineer and we were like what's up with these 10,000 spaces you know like what's going on with this file format we don't really understand I mean like we were able to use it but but we don't really understand it and uh and the epilogue engineer was like yeah the guy who wrote that software retired like a

long time ago and no one really knows how to modify it so we just put the 10,000 spaces into that's incredible this is completely ridiculous but it's just like that's the way that people are used to people are used to kind of having to have this technician attitude when dealing with machine tools and so this kind of crazy craziness at the software level seems to be unfortunately not uncommon well you know it could become a point of pride oh i know how to use this machine it's really complicated no i think i think it definitely is totally pointless i mean you don't actually know how to do anything you're just you're just interacting with the idiosyncrasies of something that's badly designed yeah you've

been uh you've been programmed by the machine or as lucy suchman would say you're exactly following their plan or you've conformed to their plan and it doesn't accept your situated action um yeah i think that that attitude i think is something that i don't know exactly how to address that in makerspaces or hackerspaces or fab labs but this attitude of like leetness or like being good at something a lot of it has to do with this invisible knowledge. Like how do you transform, you know, like how do you go from RML to G code or like how do you export something from Illustrator so it can be seen by this laser cutter? You know, all of this kind of invisible expertise. And once you know it, it's like then you become good at using a tool and that makes you feel good. But it would be better if once you know it, you could fix it. So it's easier for everyone else. I don't know. This is something I think about a lot. How do we make it so that it is not cool to know things that other people don't know? Instead, make it cool to make it really easy for everybody else not to have to know those things. Well, there's a process here.

You find out that you need 10,000 spaces. You push back on the manufacturer. The manufacturer refuses to do anything. You document it so other people know it. And then you build a tool that eats the manufacturer's lunch. Because clearly they were too stupid to survive. Yeah, yeah. We're working on it i guess and and we are i mean there are so many lower cost tools than there used to be and many of them are

recognizing that the software needs to be more accessible more modifiable it doesn't have to be baby software but having it be unnecessarily complex or technical. With your new 3D printer that was super cheap, you just picked up PID tunings off the web. We talked about tuning it ourselves, but in the end you just loaded stuff that worked a lot better than the manufacturer had. Right.

And that sort of sharing of knowledge is important for everybody. Or it could just work right out of the box well that would be optimal but that's never going to happen for everybody right having it work out of the box in a situation where we have to ship things is really hard sadly and you never really know exactly what context things are going to be in right like uh one thing that i think is kind of funny is if you ride the subway in new york and there's no cell reception you have all these people who use apple phones who get frustrated because their phone a lot of the phone functionality

relies on persistent network connection connectivity um you think that they would consider new york city as a major use case of apple iphones uh but uh but it's kind of like yeah it's it's just not a situation that was imagined by the designers and and and then the end users have very little control over how to modify um the tools that they're using and so i think it would be better to build in more end user control or more access for end user control rather than trying to anticipate in advance all the possible things that your machine might need to cater to.

But then that edges back toward being an expert technician to be able to use your machine. Yeah, sometimes I just want my phone to be a phone, damn it. Yeah, I think that there is a way in which you can have um both yeah both like i think there's a way in which you can modify things like maybe they work one way that works for most people out of the box but afterwards you can still change a lot of stuff and i think that the the challenge to changing things like cell phones for example is large enough that a lot of people don't do it i suffer immensely because i run cyanogen mod it's so annoying half the time i just i'm

like i'm sorry i can't hang out with you i have to rebuild my phone's os i used to work on that oh really that's awesome thanks for your sort kind of. So, going back to personal fabrication machines, which someday we're going to start calling replicators, but not yet because they don't deserve it yet. Yeah. They don't often match the precision and the availabilities of those things you can find in a good makerspace or at a university or at a shop,

like a real shop shop, the industrial machine tools, I mean, how much of closing the gap between personal and small fabrication and getting to those great industrial machine tools is a matter of engineering? And how much of it is a matter of engineering and how much of it is a matter of science well i would disagree with that somewhat first of all i don't necessarily think that um using small modular machine tools gives you less precision than other machine tools because for example if you wanted to mill like one of the machines that one of the fab labs made is this tube cutter it cuts pvc pipes so that you can join them together at different angles and so you need to have precision milling but the uh the work envelope that they made is like very small and

tailored specifically towards pvc pipe of maybe two inches in diameter. And so you can have local precision and have it actually work much better than, you know, if you wanted to cut like a catenary curve onto a piece of PVC pipe with like standard mill, it would be complicated to program in and complicated to kind of execute on. And so I don't entirely agree with the fact or like the assertion that industrial machine tools

are better than the machine tools that you can make yourself. And so I think that maybe there are certain things that are too expensive to expect people to have access to at a personal scale so so maybe we're not going to make um 36 by 18 foot gantries for people to make things out of but uh but maybe there are other ways concrete printer i need a 3d concrete printer yeah uh but there may be other ways concrete printer i need a 3d concrete printer yeah uh but there may be other ways that you can achieve that kind of motion like instead of uh instead of

trying to make a gigantic gantry um maybe you can do it with uh with haspels and uh and do positioning um with like a three tendon kinematic structure. Instead, you just like hang a bunch of, hang a bunch of pulleys and trees and like hang your extruder at the, in the middle and go from there. And so I think that, uh, as far as I'm concerned, I hope that the tools that we use to make tools become much better so that, uh, the tools that we need to travel to aren't as important anymore. One of the things I've seen that you talk about is how being able to shift your tools around is important to a company's agility. Do you have anything particular to say about that um yeah i mean if you're a mass manufacturer and you're going to make the same object over and over again then uh then being able to move your stuff around or being able to reconfigure your machine is not that important but if you uh

but if you want to if one day you're going to be making one thing and the next day you might be making something that has a very different geometry, then being able to reconfigure is very important. And so I think that makes a lot of sense if we're going to be producing. Agility at the production level makes a lot of sense if we're going to be working in low volume. And I personally think that working more in low volume is going to

be uh is going to be much better because it will be it will enable us to make products that are more locally specific um less uh transportation and kind of distribution objects that problems so um so so you know there's an argument to be made for never reconfiguring your shop floor because you're just going to be making the same parts over and over again but i hope we're moving away from that scenario is design for manufacturing still important for low scale production uh low volume production design for manufacturing is important but i don't think it's the only answer you know like if you want people to be able to scale up or scale down then you need to also be able to have an agile manufacturing um you need agile manufacturing as well as um as well as designs that are well suited for the manufacturing

process so sometimes i like to say yes you know go from let's teach makers design for manufacturing so that they can manufacture but at the same time let's also work on changing manufacturing so that makers can more easily access it okay i. I wonder sometimes if production, like large-scale consumer-style production, is more like cloning and this low-volume prototyping and agility is more like evolution, that you can go into niches,

smaller niches, and be better adapted for them. Yeah, that's a great metaphor. I like that a lot. Feel free to use it. Thanks. The code, the bits and the bits and atoms part, the G code or whatever we're using to describe the thing we want,

allows for more prototyping. But the tools, you've convinced me that they may not lack precision, but they do often need more maintenance. I didn't want a 3D printer for a long time because really I have enough pets and they tend to want some feeding and caring for. How are we going to build modular systems that are, well, how are we going to build systems that are modular, easy to use, easy to take care of, and inexpensive? In 50 words or less. Yeah.

With a business plan, please. Yeah. I mean, I don't, I think that all of these things are possible, but they're not, you know, there's not some like, if only we did this, everything would be better answer I can give you right now. design iterations and so i hope that um lots of people will work on machine infrastructure so that it makes it easier for people to build machines out of that infrastructure when we first started with packet switching networks they also kind of sucked but you know it's a little more robust now and um i think that the uh um the way to approach it is just iterate. And if it's too hard to iterate, well, then we need to change that first. And that's what you're doing.

I'm trying. Cool. It's not just me. Lots of people with lots of... Okay, I have a totally different set of things that I wanted to ask you about. Yeah. How was the process of getting a PhD at the MIT Center for Bits and Atoms? Well, at MIT, we have this acronym, IHTFP, which stands for, I hate this place. Or, alternatively, I have truly found paradise.

And I think that both kind of apply. At MIT, you work really hard. Like there is, you work really hard, you're expected to work really hard, and you're pushed really hard. And you do an immense, you just do an immense amount of of work more than you ever imagined that you could have done before but at the end you know you're i'm surrounded by excited people who are very bright and uh motivated um and i have all these resources that i can use all all these fancy industrial machines I can use, and yet I choose cardboard. And so I would say that it's been a really great experience for me. But I also see sometimes it's still, it can be really hard.

Like MIT, like many other places, is a sexist place, you know, and they're working on it, but they haven't solved those kinds of problems yet. with dealing with a lot of issues with dealing with like the group of people that you might expect to show up in an institution like that that we don't necessarily all have the best social skills or the best um or the best ways of working together but uh but all in all i would say that what what you gain from it or what you end up with at the end like the the friendships that you make and the work that you produce um it's the best thing ever. It's so great. I'm very happy. This sounds like an ad for Real Genius.

I don't even know what that is. It's a movie. You might want to watch it. It was pretty good. It's very 80s. It's a little dated now, but... It's still worth watching for people who went to engineering schools does it have a t-shirt in it that says i heart nuclear waste it does toxic waste i heart toxic okay someone else

has told me to watch that movie i have to watch it uh was defending your thesis what you expected um yeah more or less uh generally i think like when you when you're allowed to set a date for your defense um and your committee says that they're going to come and all that you know like then by then you you know that you're doing pretty well generally if they don't think that you're doing enough work or if they don't think that you're going to pass then they won't let you schedule a defense and so getting to the part where you're giving the defense is already pretty pretty good like there's there's a small chance you can still fail it but by then you know people are already vouching for you um and so I guess I walked into that room um kind of uh and I I laid out all of the machines uh i had made out on the table so that was also

kind of fun there's a little mini exhibition and uh this pop-up briefcase fabrication machine and the snap and the and all these metal stages and all these you know and so there was like a real progression so you know just having people come to the defense and seeing those things that was uh um that was that was already pretty fun so so the defense was less uh and a lot of people who came to the one of the things that i was surprised by a lot of people came it's like the middle of summer and and it was like in the afternoon on a monday uh but it was really really busy there were like lots and lots of people standing around like they couldn't sit down anymore um and so one of the things that was just kind of exciting was to be able to kind of tell a short version of the story of all of the things that I'd produced as part of my dissertation so

that was that was fun but um but it wasn't like uh like the word defense kind of seems like there also needs to be some offense that you're protecting yourself from. And it didn't feel that way at all. details of that or or um you know how how could you evaluate um uh how students who engage with these kinds of tools would be more successful but um but nothing like wow you reject your entire premise and all of this work sucks you know so uh so all in all i would say it was a good experience and i encourage anyone who is interested in doing a PhD to try to schedule your defense as far in advance as you possibly can, because professors are impossible to get into the same room. Christopher has a good story about that. Yeah, at my master's oral exams, one of the people I scheduled didn't even show up.

So I had to scramble through the department finding somebody who would be willing to do it. Otherwise, I was going to have to do it some other time. That was my biggest fear on Monday. I can't believe you actually went through that. That sounds terrible. And the replacement had interesting questions. Oh, no.

Steve is the one that asked you how many electrons are in a chicken. Yeah. And I was like, what? Weird. Oh, no. See if it's the one that asked you how many electrons are in a chicken. First question, how many electrons are in a chicken? Yeah. Huh. And I was like, what? Weird. Oh, God. Let's assume a 10-kilogram chicken. And it's spherical.

And it's in a vacuum. Frictionless. All right. I think we have kept you for long enough. It has been wonderful to talk to you. But I hear you're getting on a plane soon. Yeah, that's true. It's only tomorrow.

Don't worry. I'm going to Shenzhen. I go there a lot to buy motors and other kinds of parts, but now I'm going there because I'm going to FAB 12, the 12th annual Fab Lab conference. It's going to be partially focused on FAB 2.0, which is using Fab Lab machines to make more Fab Lab machines. And so I guess I'm kind of in charge of that component.

But it'll be pretty fun. Yeah, that's where I'm going next. Cool. Christopher, do you have any other questions? I had something about making machines making machines, but she just said that's what she's going to Shenzhen to do. So apparently that's happening. No, no, no.

Using machines to make more machines. Right, right. Not making machines that make machines by themselves. Right. Wow, okay. Not yet. Soon though. Soon.

Nadia, do you have any last thoughts you'd like to leave us with um yeah last thoughts uh one of the best things I think about working on projects like these has been that I work with people that have very different backgrounds from me and I learn a lot from them. But you need to have kind of the space to be allowed to do that kind of interdisciplinary work. Or, you know, if we had to really efficiently finish lots of projects, then I wouldn't be able to work with someone who didn't work in the same way as me and so I guess final thought is how can you encourage or give yourself space for those kinds of interdisciplinary collaborations um because uh for me they've been really valuable but they've also been a struggle you know uh I'll get into big fights with James because he thinks

that all of my graphics are terrible. And then, you know, you focus on that instead of doing next work. So, you know, how can you create space, especially for hardware? I think this is important because in hardware, you tend to be hyper-focused on something very specific. And so how can you see a bigger picture and, and, and encourage more kind of cross disciplinary collaboration. I mean, it's something that I was, I think about. And so if people are listening to me anyway,

then, you know, maybe you're thinking about it too. Seems like a good thing for us all to think about. My guest has been Nadia Peek, research scientist at the MIT center for Bits and Atoms. Thank you so much for being here. Thank you, Alicia. Thank you, Bruce.

You guys have a great evening. Huge thank you also to Bruce for suggesting Nadia as a guest. I heard her on the Amp Hour too and wanted to talk to her then. I'm happy he reminded me. Thank you also to Christopher for producing and co-hosting, and of course, thank you for listening.

Hit the contact link or email show at Embedded if you'd like to say hello. As usual, I do have a final thought to leave you with. This one from Maya Angelou. What you're supposed to do when you don't like a thing is change it. If you can't change it, change the way you think about it. Don't complain. Embedded FM is an independently produced radio show that focuses on the many aspects of engineering. It is a production of Logical Elegance, an embedded software consulting company in California.

If there are advertisements in the show, we did not put them there and do not receive any revenue from them. At this time, our sole sponsor remains Logical Elegance.