Before touch screens and sleek aluminum casings became the visual language of modern computing, there was an object so striking it would influence Steve Jobs and find its way into the Museum of Modern Art. In 1985, Tamiko Thiel—now recognized among the world's foremost digital artists—solved an impossible design problem: how to give physical form to artificial intelligence.
As lead product designer for Thinking Machines Corporation, Thiel was responsible for the form and visual design of the Connection Machine. Her solution was revolutionary—a perfect black cube housing smaller cubes, with transparent panels revealing the hypnotic blinking of thousands of processors working in unison. This wasn't merely aesthetic innovation; it was functional storytelling. The machine's exterior communicated its internal workings—65,536 simple processors connected in what physicist Richard Feynman described as a "twelve-dimensional Boolean hypercube." Thiel’s vision made this abstract concept tangible, mysterious, and compelling.
The Connection Machine's distinctive form challenged how we visualize computation itself, transforming what might have been yet another anonymous box into an object of almost sublime technological beauty. When potential clients encountered this enigmatic black cube with its pulsing red lights, they were experiencing the future of computing made manifest. This design sensibility would later echo through Jobs' work at NeXT and ultimately Apple, where products increasingly became objects that transcended their functionality.
Tamiko Theil was Lawrence Peryer’s guest on the Spotlight On podcast for an extended two-part conversation. In part one, Tamiko discusses her journey from mechanical engineering to product design, her collaboration with visionaries like Feynman and AI pioneer Danny Hillis, and how a T-shirt logo became the blueprint for a machine that would change computing forever. You can listen to the entire episode in the Spotlight On player below. The transcript has been edited for length, clarity, and flow.
Some Small Box
Lawrence Peryer: Were you drawn specifically to human-machine interfaces?
Tamiko Thiel: Yeah, that was an interest while I was at Stanford Product Design. Also, when I see a product for the first time, what does it say to me? If I need to use this, how can I design a product so it’s clear how it should be used, and what it should mean, to me as a tool or as someone it's supposed to help with the function?
It was clear to me that all these design books have helpful rules of thumb and things like that. And one is, looking at the American population, the span for adults from the smallest and the tallest.
And I was in a design group at Hewlett-Packard where I was definitely in the lowest fifth percentile in size. And another one of the product designers was definitely in the ninety-fifth and upwards percentile in terms of height and size. So we joked that we were the perfect team members because between the two of us, we could test it out and see how one product could be used by the whole extremes and everyone in between.
Lawrence: There are a few things that occur to me, and one is just how long things like artificial intelligence and virtual reality, and to an extent AR, have been with us and have been in the labs, with artists, and with product designers. I will ask you to tell the story of the Connection Machine, but before I do that, how did the public unveiling of the Macintosh land for you? Was it a paradigm shift?
Tamiko: Oh yeah. It was. The Macintosh came out in '84. I had been in the Xerox PARC social crowd since at least '79 because my boyfriend was working there, so I visited him sometimes after hours and saw the Windows system and the mouse. This would be the most incredible tool for a designer like me, but it was only on these Altos in Xerox PARC. My boyfriend said, "We have Windows and mice, and you can't have them.” The idea that, okay, we're going to keep this technology to ourselves, was infuriating to me.
So when the Macintosh came out in '84, it didn't have separate windows but a bitmap display. So the bitmap display, the mouse, and all of the sorts of programs I had been able to use because I had been working at computer companies that provided access to these tools. And now all of it in some small box. It was a revolution.
It was the advent of the Macintosh that made it possible to do a lot of things using computer technology, whether it was editing or creating images, where you were using the computer to do things, but you didn't have to learn to program to do them. It's quite possible that, in some funny ways, the Macintosh was the beginning of the demise of women's participation in computer science departments. Because they were not interested in developing new technology, they wanted to be able to do things with technology and tools like the Macintosh enabled them to do so without having to build the technology themselves.
That's frankly also what ended up happening with me when I decided that I wanted to start creating media art. I've developed hardware and software, and it takes a long time. As an artist, I’ll do it if I need to. But if I can find a way to use existing hardware and software tools to create artworks, I prefer to put my time into creating the artwork rather than creating the tool. So perhaps that's what happened with me around the mid-eighties. I stopped pursuing engineering and started creating art with technical tools.
Lawrence: That is a fascinating insight and reflection on that evolution. So with that said, do you mind relating the tale of the Connection Machine?
Tamiko: Not at all. When I was at MIT as a grad student, the people at Xerox PARC working in artificial intelligence said, "Oh, you're going to MIT, you have to meet these friends of ours, Marvin Minsky and his PhD student, Danny Hillis." And so I went to MIT in 1981 as a mechanical engineering student, but with social introductions to several people in the MIT AI Lab.
So I graduated, there was a ceremony, and then two or three days after, Danny called up and said, "Can we meet? I want to tell you about this company I'm starting." And I knew he'd been working on this crazy concept using Marvin Minsky's sort of society of mind ideas, that intelligence probably doesn't arise from having really good processors, because here we've got eighty billion stupid processors. Somehow, out of these eighty billion stupid processors, human intelligence arises.
That spark set Danny in motion to say, “Okay, I don't think we're going to solve all of the multitudinous problems of artificial intelligence by creating very powerful processors. I think it's something about the connections.” And that's why his concept was called the Connection Machine. He said, "We've got funding to start a company, but we don't even know if it's physically possible to build this machine. If it is, will you join us and design the packaging for the Connection Machine?"
Then he said in the following sentence, with a twinkle in his eye, "And if you do come, you get to work with Richard Feynman," because he knew I had started as a high-energy physics student at Stanford, and we all revered Feynman. And indeed, when he said that, my eyes opened wide and I said, "Where do I sign? Where do I sign?"
Lawrence: Yeah, why not? (laughter)
Tamiko: There’s a funny story that Danny tells. Richard's son, Carl, was studying AI along with Danny and Marvin, and was one of the people who helped start Thinking Machines. At some point, Danny was hanging out with Carl and his family and told Richard, a professor at Caltech in high-energy physics, about the idea for the Connection Machine. Richard said, "Oh, that sounds like a crazy idea. Can I come out and play?" And so Richard came out and spent his summer vacations hanging out at Thinking Machines.
When he first showed up. Richard said, "Richard Feynman, reporting for duty! What should I do?" And they were in the middle of setting up, and they said, "Well, we don't have anything for you to do. Can you go to the stationery store and buy some paper and other office supplies?" And Richard said, "Yes, sir." And Richard Feynman went off to buy office supplies.
At some point, they figured out one of the big problems was starting with 64K processors. That's 65,536 processors. They will be simple, one-bit processors, connected in a grid pattern with wires. But that means that when you have 65,000-plus processors, it's really hard to communicate between them, and it's important that these communications go quickly. So why not figure out some way to connect them much faster than just talking to your nearest neighbors? And Richard came back and said, "Why don't you connect them in a twelve-dimensional Boolean hypercube?" (laughter)
And Richard also said, "I've been solving all these partial differential equations, and I figured out how we can route the messages between all the processors." And it's like no computer scientist had ever used partial differential equations to solve a computer architecture problem. But Richard was a physicist, and that was one of the tools he used.
So Richard designed the router that passed messages between these 65,000-plus processors. And part of my job, of figuring out whether it was possible to build this machine, was understanding how everything fit together because I had to do all of the drawings that would explain to the fabricators what to do.
So I'm going like, "Danny, what is this?" And he says, "Oh, it was Richard Feynman's idea. Go ask Richard." I went to Richard and said, "Richard, what's a twelve-dimensional Boolean hypercube?" And he says, "Oh, that's easy." And if you've ever read any of Richard Feynman's books, you'll recognize this immediately: he grabs a piece of paper and a pencil and starts drawing diagrams.
So I'm playing around with paper and pencil, and I have to say, this is 1983, we're designing the world's first commercial artificial intelligence supercomputer on paper and pencil. Well, they were also using symbolic Lisp machines. But all the work on the packaging design was done on paper and pencil.
The Swag is Important
Lawrence: What was the Connection Machine meant to do?
Tamiko: The machine was meant to duplicate human intelligence. There are a lot of different components to artificial intelligence, but Danny's intuition was that the first thing we need is massive parallelism, because that's what we've got in our brains. We don't know why, but somehow out of those connections that change, when you practice something, whether it's a new language or whether it's a new physical skill or learning to ride a surfboard for the first time, all of your neurons are rewriting their connections and that's what we call learning.
And so, of course, there were neural nets already at that time, and there was a theory that neural nets might be the secret behind this, but we didn't have the hardware to implement them. And even with a Connection Machine, they could only do neural nets with a couple of layers. And it wasn't until 2012 or so that we started having the computing power to have really deep levels of neural nets.
That's what suddenly allowed artificial intelligence to start working as it is now. The computing power is catching up to the ideas that were already there on neural networks, and how somehow the connections between neurons and being able to change the relative weights between these nodes were going to be storing memory or storing some sort of ability to move or to distinguish something to recognize. And we still don't know exactly how it works. And that's what's spooky about neural networks.
Lawrence: Incredible, isn't it? So we have this construct that Feynman described to you that you could finally articulate and design around. How did you come to your thinking about how to visualize computation?
Tamiko: Through the T-shirt logo. This is the only supercomputer designed after a T-shirt logo. At the beginning of any good startup company, at some point, you need a T-shirt that sort of symbolizes your vision.
Lawrence: The swag is important! (laughter)
Tamiko: The swag is really important. It fuses people and gives you something to present to the outside world. It's corporate identity, if you will.
I had been talking about this cube of cubes, and here we can visualize it in this simplified way that explains all of that structure. And so Danny approached me and said, "We need a T-shirt logo." And he said the way that he had been visualizing the machine to himself at this point was this cube of cubes, but within this rigid, unchanging structure of the twelve-dimensional hypercube, the software could make connections on its own. Thanks to Richard’s routing system, the processor would just say, "I want to send a message to that processor way over there." And those connections would conceptually be independent, abstractly, be independent of the hardware structure.
So that's the logo that I came up with. This geometric sort of Cartesian cube of cubes, but inside these pom-pom balls, whose connections were independent of that rigid hardware structure. So we made the T-shirts, and everyone was wearing these T-shirts, and Richard really liked the T-shirt. He wore it for a lot of his lectures. So you'll see him wearing the T-shirt in many videos while explaining his cool ideas.
When it became time to start building the machine, I took all the input from the electrical engineers and drew up images. I thought, “Okay, these are the components you want and the connections you have. Then, in a real physical structure, we've got a board that's two feet by two feet, only one board. And there's a whole lot of these, and if you put them together and leave room for cooling and the cabling, you've got something ten feet tall and about five feet by five feet."
And everyone said, "Oh my God, that's way too big." And so all the electrical engineers went back to their drawing boards and said, "Okay, well, we can throw out all these connections and reduce these and double up these." And so we finally got the boards down to their current size, which was still large. But at some point, I said, "Okay, we can put this in a standard nineteen-inch rack. Guess what? After you guys did all that reduction work, we could make this look like any other computer out there, the refrigerator you've got in your kitchen, or any sort of rectangular box about a person’s size.”
The problem was we were all, like Danny and I were twenty-five. Many people like Brewster Kahle, better known today as the founder of the Internet Archive, were about twenty. So it's like he and Carl Feynman, they're all twenty, twenty-one, or something like that. And we're going up against the big boys, the older men, who were probably in their forties and fifties. They were used to getting all of these million-dollar contracts from DARPA, from the Defense Advanced Research Projects Agency. We were going to bring in these people and say, for only three to five million dollars, this machine can be yours. It's got an architecture unlike anything you've ever seen before. It will do things unlike any computer you've ever seen before. But we knew we had lost if we brought in that guy with his checkbook, and if he saw it, he said, "Oh, that looks exactly like the refrigerator that I just bought for my wife.” We would have no credibility.
I felt we needed a distinctive form that looked like no other machine. But if that form could also immediately communicate to the person how the machine works, how it's so different, how it's so revolutionary, then that guy will say, "Here's the check. You've got it."
So there you see also my product design and human factors background; how does the appearance of this object we've built affect the person who sees it? Do they understand how to use it? And this is marketing, right? Does it communicate to them that with this object, they can do something they cannot do with any other object out there?
You can say it's pure marketing, but that's how you form an emotional relationship between an object, the human who uses it, buys it, or somehow has to deal with it. So that's a very large part of a product designer's job—to create this emotional bond between you and some built object. And that's when they engaged two designers from what was then the Eliot Noyes Company. Al Hawthorne and Gordon Bruce were the principals after Eliot Noyes himself had died. And so they came in as the industrial designers.
I presented this whole speech about how the machine is connected, how it functions, and what it's about. And they said from the beginning, "Oh yeah, I think we should do something with this cube of cubes idea." I said, "Yeah, probably. But I've been living and breathing it for the last year and a half. You go off and make cube of cubes drawings. I'm going to go off and run through different alternatives that have nothing to do with that, just to make sure we're not blinding ourselves, that we're checking out all the alternatives."
So I hired a friend of mine, Tom Chytrowski. We went off, and Gordon and Al went off, and we all did a lot of drawings. At some point, I said, "Okay, let's just get this over with. I'm going to draw a cube of cubes also." And I drew it and Tom and I looked at it and said, "Well, that's it, isn't it?" As soon as it was there on paper, we knew that's what it would be.
Concept sketches for the Connection Machine by Tamiko Thiel and Tom Chytrowski, via tamikothiel.com .
Okay, so that's the hardware. That's the rigid hardware structure. The T-shirt logo was the best expression of the machine’s full functionality. So, how do you express that T-shirt in real life? (laughter)
I’d read science fiction like everyone else and dreamt of AI like everyone else, or had been doing it for the last several years. And Carl Feynman had, too, and said, "When I imagine the Connection Machine, I imagine this cloud of lights and they're all like blinking as they send their messages to each other." And so we said, "Okay, we want to make the machine’s internal workings visible to the outside world. What if we just make the outer plastic skins transparent?" Well, what do you see? You see boards and you see lots of cables, and you see chips. But that doesn't tell you anything about how the machine functions.
There’s this design idea that form follows the function of the machine. That works fine if you're dealing with a mechanical function, but if you're dealing with an inherently invisible function, like electronics, then it doesn't tell you anything because it could be any board, it could be any chip, it could be any cable. That's not unique.
We had the idea of processors glowing and sending off their light messages. Every single chip had a status light connected to it so you can tell whether it’s getting electricity, working, or malfunctioning. And usually the lights are right next to each chip.
So we just moved those lights from being next to each chip to being on the edge of the board that was on the outside, made the doors translucent, not transparent, because if it's transparent, then you're just going like, okay, there are a lot of boards in there. If it's translucent, then they're glowing mysteriously through this door.
Lawrence: Such drama. (laughter)
Tamiko: It increases the drama. You're not saying, "Oh yeah, there are boards and cables." You're saying there's some sort of mysterious lights that are blinking and glittering in there. It's thinking.
And with this cube of cubes structure, it was a metaphorical electronic brain, when you think of the brain with its convoluted structure. And it became a cult object with that very small crowd of people who were really into supercomputers in the mid-eighties. People told me they would sneak into the computer room after hours and watch the Connection Machine sit there, blinking mysteriously like an electronic brain.
Lawrence: That's beautiful. I love that.
Tamiko: It picked up on a lot of images that we all had floating around in our heads from science fiction movies and a shared culture of techno-optimism, of thinking, yes, we can build a machine that somehow improves things and extends the range of what we can do as human beings. Let's hope that it doesn't take a wrong turn.
Danny was also thinking about that and said, "Okay, let's make it not any taller than you.” I'm five foot three, so he said, "Make it as tall or smaller than you, because then for your generic white male, which is the main customer base we have to aim at here, it'll be smaller than them. It'll be beneath their eye level, and it'll feel like a member of the family rather than an overlord."
So we were also very conscious of this fear: Will AI kill us? Because even in 1984, '85, the HAL-9000 computer of 2001: A Space Odyssey, the film that came out in 1968, was still the most dominant image of an AI. And of course, it was an AI that went wrong and killed the humans because it perceived the humans as endangering the mission. And that's still the fear that we carry with us now.
Lawrence: To react that way.
Tamiko: Right. And, but that's what we're seeing happen right now. So, the follow-on machine, the CM-5, I was not involved in the design of that machine. And it ended up being taller than a human. I'm unsure how tall, at least seven feet, maybe more. It did dwarf humans.
These strips of LED grids were added on, but they were not functional. They were decorative, which my Bauhaus background would've never allowed me to do. I mean, I would've done something like make the sides of the machine translucent so you could see status lights on those boards. There were all sorts of lights happening that we could have shown through. But I was in Germany as an art student and had no contact. So I was not involved in the CM-5 design at all.
------------------
Visit Tamiko Thiel at tamikothiel.com where you'll find extensive background and photography of the Connection Machine. You can follow Tamiko Thiel on Bluesky, Instagram, and LinkedIn. As we publish this, Tamiko is in Vienna for the Radical Software: Women, Art & Computing 1960–1991 exhibit. Part two of our interview will be published next week.
Check out more like this:
Lawrence Peryer
Lawrence Peryer
Comments