Journal Entry #1

My name is Tammy Maheswaran, and I’m an independent roboticist. I’m keeping this journal to document the progress of a personal project.

My goal is to create a developmental robotics platform that will serve as a tool in the study of machine learning and emergent behavior.

Let’s get started, I guess?

Journal Entry #2

It just occurred to me that whoever ends up reading this journal might not have a background in robotics, and therefore would have no idea what I’m talking about.

Let me clarify: My goal is to create a learning machine. A machine that, through trial and error, creates its own subjective understanding of its surroundings.

Aside from a few simple rules, she’ll begin as a blank slate. She’ll know nothing. All sensory information will seem like a senseless jumble.

Her only asset will be her capability to learn.

There’s no telling where the project will go from there. My hope is that with time and guidance she’ll become something truly unique.

Journal Entry #3

I’ve spent the last few days doodling robots. I’m still trying to find the right look.

I want her to stand out from the crowd. Something really unique. People seem to have this preconceived notion of what a robot should be. I’d much rather think about what a robot could be.

So many modern robots have a bleak, sterile look to them. Machined aluminum. Smooth plastic. Even when they manage to look appealing, they have an air of inaccessibility about them. They don’t invite your touch.

When you stop to think, a robot could be made of just about anything. Unglazed porcelain. Driftwood. Bamboo. There are so many ways one could make a robot seem warmer, more textural, more inviting.

Perhaps what I’m looking for is something more along the lines of a plush toy. Soft. Safe.

I’ll take some time to think on it.

Journal Entry #4

Memory foam. I’m going to use memory foam.

I’ll have to find a solution for various heat retention problems, but it’ll be worth the trouble. Lightweight. Soft. It’s perfect.

She’ll likely be incredibly clumsy for a while after booting up for the first time. There’s no telling how long that stage will last. A body constructed from memory foam will help prevent her from damaging herself or her surroundings.

She needs to be allowed to make mistakes. It’s a vital part of learning. The least I can do is make it safe for her to do so.

Still, memory foam is a delicate material. She’ll need some sort of outer layer to protect against abrasions.

I’ll compile a list of possible materials in the morning.

Journal Entry #5

I’ve assembled a small collection of fabric swatches in my attempt to find a suitable skin for the project. Some interesting ideas, but none of them feel quite right.

Denim seems like a good candidate in the practical sense. Strong. Durable. Sturdy. Aesthetically however, I don’t think it’s a good fit. I want her to have an air of elegance about her. Sleek. Smooth. Denim has none of these qualities.

Silk would be elegant of course, but it seems antiquated. I want her to feel modern. More like a piece of abstract sculpture, and less like an eighteenth-century Japanese automaton.

I flirted briefly with the idea of using a baby blue polar fleece, like a pair of footed pajamas. It’s a cute thought, but I can’t help but feel it would look a little demeaning.

I don’t know. I’ll keep looking.

Journal Entry #6

Neoprene. I can’t believe I didn’t think of it earlier. Waterproof. Elastic. Easy to clean. It’s pretty much skin already.

Neoprene can be treated almost as if it were a fabric. It can be cut. It can be glued. It can be sewn. It shouldn’t be too much trouble to fashion a snug, form-fitting outer layer to fit nearly any body shape I happen to come up with.

I have to consider thickness carefully. Too thin and the neoprene could tear. Too thick and it may begin to restrict movement. I think four millimeters is probably a good compromise.

Also, the nylon knit fabric used to laminate neoprene is available in nearly any color imaginable. I’m tempted to try something in charcoal and turquoise, personally.

Journal Entry #7

I’ve got a vague image of what I want the final product to look like forming in my mind. It still feels like something is missing, though. All that neoprene ends up looking sort of featureless. It lacks character.

There needs to be some sort of visual interest to break up the monotony. Something to keep her from looking like a walking wetsuit. I can’t quite put my finger on what, though.

Journal Entry #8

Stitches. Big, chunky stitches, like a rag doll sewn together with shoelaces. Non-functional, running along the seams, obscuring the true needlework.

I’m going to use turquoise 550lb nylon paracord, laced through four-millimeter powder-coated aluminum eyelets. Obviously the tensile strength of paracord isn’t strictly necessary for what is basically a purely aesthetic purpose, but its thickness gives it an appealing chunkiness.

Journal Entry #9

It’s slowly becoming apparent that I probably shouldn’t stitch this whole ensemble completely shut. I’m still a little fuzzy about what kind of hardware will be housed inside her plush exterior, but whatever it ends up being, it should be kept accessible.

I’m thinking I should install a hidden maintenance zipper. One that runs all the way around the central seam, allowing her neoprene skin to be split into two separate halves and removed from her endoskeleton. It shouldn’t be too difficult to make it inconspicuous. Use a narrow, fine-toothed zipper. Match the color to the surrounding fabric. Tuck in the pull tab. Nearly invisible.

Maintenance wouldn’t be a routine thing, as unzipping would require unlacing several meters of paracord, but it’ll be reassuring to have a non-destructive access method should mechanical problems eventually crop up.

On the plus side, all that paracord will help prevent any potential self-unzipping.

Journal Entry #10

Today it dawned on me that I had yet to decide how tall the final product should be. After a lot of thought, I settled on sixty-two inches. She needs to be tall enough to comfortably interact with people, but small enough to be restrained if she starts acting unpredictably. I measure in at a whopping sixty-six inches, so that seemed like a fair enough compromise.

I realize I should probably be planning this sort of thing in metric. I’m so used to hearing people’s height in feet and inches that it was just easier to visualize this way. I’ll convert her measurements to centimeters before I begin work on the actual blueprints.

Journal Entry #11

If I were to start fabrication on the project as it stands now, the final product would end up being little more than a plush toy. She still needs a skeleton.

I know I badmouthed aluminum the other day, but it really is a versatile material. Lightweight. Strong. Fairly inexpensive. The downside is it can be tricky to weld. I don’t expect I’ll be doing much welding on this project, but it’s something to keep in mind.

Stainless steel might be an option, but it’s heavy. With a machine like this, unpredictable behavior is to be expected. It’s probably best to shave off as many kilograms as possible, for safety reasons.

Titanium would be ideal, but it comes at a price. I have some money stashed away for the project, but my budget isn’t unlimited. A titanium armature could potentially duodecuple the cost of materials.

I still need to explore carbon fiber as a possibility. It’s far lighter than any of the previous options, but I’m not sure what it would cost.

I’ll report back tomorrow.

Journal Entry #12

It just dawned on me that any skeletal support will be completely hidden in the final product, and therefore it doesn’t matter what it looks like. Machined PVC. Durable. Flexible. Cheap. Problem solved.

Journal Entry #13

I’ve been thinking carefully about potential methods of animation.

The obvious solution would be the classic servomotor. Accessible. Compact. Reliable. Not a particularly groundbreaking method, I suppose, but it works.

I do have a serious concern regarding servos though. Servos are ideal for tasks that require accurate, repetitive movements. Assembly line robots, for example. They can be precisely controlled by the programmer.

In this particular case, the programmer is not going to be the one in control.

Servos are capable of potentially violent, jerky motions. My fear is that she might accidentally damage herself or those around her in her early attempts at movement. It would be like putting a toddler at the wheel of a car. I need something safer.

Pneumatic muscles are interesting devices. Gentle. Supple. Compliant. They’re a lot like human muscles, if human muscles ran on compressed air. Pneumatic muscles move with a certain fluidity. They have a little bit of give. I’m quite confident they’ll be the safer choice for the project.

Journal Entry #14

Pneumatic muscles necessitate the use of an air compressor. This will be tricky. For the purposes of this project, the compressor needs to be compact and whisper-quiet. Most off the shelf compressors tend to be bulky and loud. It’s going to have to be a custom job.

There are several types of compression mechanism to choose from, however. Reciprocating compressors work a lot like the engine of a car, but in reverse. Problem is, all those pistons create just about as much racket as a car engine. Far too noisy for my purposes.

Scroll compressors are a clever concept. A motorized spiral oscillating against a stationary spiral, working together to force air toward their centers. But that sort of high-speed oscillation is bound to create a lot of vibration. Suitable for a stationary air compression unit perhaps, not so much for a clumsy, unbalanced robot.

A rotary screw compressor is probably my best bet. No oscillation means less vibration. No pistons means quieter operation. The whirring of the motor will be audible, but that’s not nearly as hard on the ears as the incessant buzzing of a reciprocating compressor. Besides, all that memory foam is bound to have a pretty serious sound dampening effect.

In a way, the compressor will act as the heart of the machine, so it only makes sense to place it in her chest. Something about that just feels right.

As far as compressed air storage is concerned, the best place for the canister is almost certainly the pelvic bowl, alongside her solenoid valve bank. Her hips will be the widest section of her torso by a fair margin, and more space for pressure storage means greater endurance.

I will admit that this is an unconventionally tight space to install a fully-functioning air compression system. I suspect she may need to stop and rest occasionally. Catch her breath, as it were.

Journal Entry #15

Batteries are one area where I’m willing to compromise my goal of keeping the project lightweight. One of my biggest pet peeves is seeing a robot permanently tethered to an external power source because nobody bothered to take the weight of the batteries into account.

Lithium-ion batteries seem to be the standard in modern electronics. They have a high storage capacity. They’re reusable. The downside is that they take a long time to charge. Hours. More downtime means less learning. Less learning means slower cognitive development. That’s unacceptable for a machine that has to start from scratch.

Supercapacitors might be the better choice. They’re not true batteries, as they don’t store energy chemically, but they can be used to fulfill a similar purpose. They do have downsides. They tend to have a lower storage capacity. They’re heavier. Bulkier. But they can charge in a matter of minutes. Seconds even. With a high capacity-array of supercapacitors installed, any downtime will be negligible.

Journal Entry #16

I’d like to have an easy way to read her charge level. Something that can be read at a glance. It would make day to day interaction a lot more straightforward.

It should probably use a universal symbol. Most charge indicator symbols are battery-shaped. Superconductors aren’t true batteries, I know. But it would be immediately understood by anyone who has even a passing familiarity with electronics. Kind of like how software interfaces use a floppy disk to represent the act of saving a file, despite the fact that floppies have been obsolete for nearly two decades now.

Find this and other great novels on the author's preferred platform. Support original creators!

That still leaves the problem of how to display said symbol. Her neoprene skin doesn’t leave a lot of real estate for electronic displays. I suppose I could cut a small hole in the neoprene to allow a display to peek through.

Actually, now that I’ve given it some thought, most modern displays have some degree of flexibility. I could probably glue one directly to the surface of the neoprene. The left side of her chest seems like an appropriate location. Like a patch or badge.

The display itself should probably be electrophoretic, like an e-reader. They only consume energy when refreshing the display. In an application like this, that means energy consumption would be almost nonexistent.

Also, if I were to coat the display with a matte-finish transparent turquoise laminate, white sections of the image would show up turquoise, keeping it consistent with her turquoise and charcoal color scheme. Finally, segmented electrophoretic displays can be trimmed to nearly any shape imaginable, as long as the circuitry itself isn’t severed. Aesthetically, it’s just a good choice.

Journal Entry #17

I thought the best charging option would be a retractable power cord on a spool, like you might find on a vacuum cleaner. The further along I get, however, the more I realize that space is at a premium. There’s simply not any room for a heavy spool of copper wire.

The obvious solution would be a fully detachable power cord. Plug one end into the robot and the other into the wall. However, it would be ideal if she could remain operational while charging, and I’m a little nervous about the idea of leaving her unattended near a live wall socket.

Now I’m thinking about inductive charging. No wires. No risk of electrocution. Just a pair of copper coils in the soles of her feet, and a matching pair inside a charging pad on the floor. Charging would be as simple as standing on the pad and waiting.

It would be slower than wired charging, certainly. Not particularly efficient either. But the peace of mind makes it worth it. It’s simple and safe enough that I’m confident she could learn to do it on her own, as needed.

The act of balance itself requires a modicum of energy, however. That would slow charging to some degree. I could easily offset that by providing underarm support. Like a doll in a bell jar.

Journal Entry #18

It’s time to start thinking about computer hardware. I’ve engineered myself into a bit of a corner by using all that foam. That much insulation wrapped around a powerful processing rig is pretty much guaranteed to catch fire eventually.

I’ve been trying to figure out a workaround, and I think I’ve settled on something pretty clever. If I were to fashion the neoprene above her shoulders into a sort of hood, I could stitch a strip of curved aluminum right into the fabric itself and have it attach to the upper edge of her faceplate.

A setup like that would create a spacious and well-ventilated area behind her faceplate where I could stick computer components without having to worry about creating a horrifying fire hazard.

Also, it would result in a pretty nifty “invisible neck” illusion, which is the main reason I’m doing it.

Journal Entry #19

If the robot’s computer components are going to be situated in her head, they’re going to have to be lightweight. Any unnecessary top-heaviness is going to make learning to walk that much more difficult.

I’ve been toying with the idea of using smartphone components in order to minimize weight. The obvious downside to this is that mobile processors are considerably less powerful than their desktop counterparts.

However, If I can assemble a cluster of several top-of-the-line octa-core mobile microprocessors, I’d be getting a pretty incredible speed to weight ratio.

In fact, using smartphone components and a little bit of puzzle-work, I could probably fit the entire motherboard into the faceplate itself, leaving the rest of the hood empty and well ventilated.

Journal Entry #20

Most estimates place the storage capacity of the human brain somewhere between ten and one hundred terabytes. I could easily get my hands on a ten terabyte hard disk drive, but the idea makes me nervous.

With all those high-speed moving parts, HDDs are prone to catastrophic failure. I think almost everyone has had a hard drive fail at some point. The bumps and bruises one would have to endure inside a machine that’s teaching itself to walk would make failure all the more likely.

A solid state drive is the only real option. No moving parts means a greatly reduced chance of failure. Not a zero percent chance of course, but pretty close.

Unfortunately, solid state drives are a far younger technology than hard disk drives, meaning a lower storage capacity at a higher price. There are currently a few four terabyte SSDs available on the consumer market. Installing a pair of them would make for an eight terabyte total storage capacity, which is certainly nothing to sniff at.

Also, it’s not like she’ll need to permanently store every kilobyte of data she processes. I’ll teach her how to forget.

Journal Entry #21

If I were to start fabrication on the project as it stands now, the end result would be nothing more than an animatronic puppet. A machine does not qualify as a robot until it has the ability to perceive its environment.

Vision seems as good a place to start as any.

Whatever type of camera I choose, she’ll need at least two of them. Depth perception will be a necessity for learning how to interact with her surroundings. Obstacle avoidance, object manipulation, all of these tasks require a sense of depth in order to be performed with accuracy.

What’s left to choose is what she’ll see through those cameras. Near-infrared light would be a unique and beautiful way to view the world. Pale blue skies and snow white leaves.

Ultraviolet would be gorgeous as well. She would see hidden secrets that no one else could. Flowers with patterns meant just for bees. Invisible freckles on people’s faces.

But those types of imaging rely on bright natural sunlight. She’d be rendered effectively blind indoors. Seeing as that’s where she’ll be spending most of her time, they’re not really viable options.

Thermal imaging would allow for clear vision regardless of light conditions, but she’d miss out on so much that the world has to offer. Shadows. Colors. Translucency. None of these things are reproduced in a thermal image.

If I’m being honest with myself, full color perception will probably be more important to her development than I’m giving it credit for. It’s an easy visual shorthand for the boundaries of any given object. For a machine that will have to teach itself to see, that’s important.

Obviously the only way to go about that is with an ordinary, run-of-the-mill visible spectrum camera. It makes my job easier, at least. There are plenty of compact stereoscopic camera modules on the market.

Journal Entry #22

Hearing should be fairly straightforward. A simple binaural microphone array will allow for satisfactory directional hearing.

They don’t have to be conspicuous. I should be able to fit a pair of small directional microphones directly inside her faceplate without much trouble. In fact, I can probably rig up the array using a couple of hearing aids.

Internal noise interference can sometimes be a concern with machines like this, but seeing as the faceplate will have no moving parts and will be situated away from the rest of the body, I don’t foresee any problems cropping up.

Journal Entry #23

Touch will be a bit more complicated. It will likely end up requiring hundreds of sensors. Luckily, simple and flexible pressure sensors can be made for next to nothing.

It’s as simple as sandwiching a piece of pressure-sensitive conductive foil between two layers of conductive fabric. When pressure is applied, the electrical resistance of the foil is reduced, allowing for a simple numerical readout of the force being applied. Then I just sew each sensor into a pocket of ordinary non-conductive fabric to keep the whole unit from falling apart. The only tools it requires are basic sewing supplies.

After that, I can glue each sensor to the inside of her neoprene skin, stencil the necessary circuitry directly onto the rubber using a conductive metallic paint, then laminate the entire sheet with a layer of nylon knit fabric to prevent friction wear.

I’ll have to plan the sensor pattern carefully though. For instance, her shoulders might only require a few sensors each. The palms of her hands however, could require a dozen or more. The sensor resolution will be dependent on the dexterity and fragility of the corresponding body part.

Whatever the specifics, the end result will be a machine that can sense the slightest touch on nearly every square inch of her body. Not many robots can make that claim.

Journal Entry #24

External temperature sensors are not a huge priority in a machine of this type. However, for self-preservation purposes, she should be able to monitor internal temperatures.

For starters, she’ll need temperature sensors in her faceplate to prevent her CPU from overheating. Luckily, off-the-shelf processors have temperature monitoring capabilities built in, which saves me from having to worry about it too much.

More urgently, she’ll need a sensor monitoring the temperature of her air compressor. I failed to mention this before, but air compressors get hot. Hot enough to potentially catch fire. She’ll need to monitor compressor temperature in order to avoid overexerting herself. If the temperature gets too high, the compressor slows. This will inevitably lead to occasional shortness of breath, but it’s better than inevitably leading to a toxic foam-rubber bonfire.

Journal Entry #25

If she’s going to navigate her environment efficiently, she’s going to need to know which way is up. That’s where an accelerometer and gyroscope come in handy. The adjustments needed to keep a bipedal body upright are subtle and constant. These two sensors working in tandem will provide an X, Y, and Z axis that will be vital for learning to balance.

Modern smartphones have both of these sensors integrated into a single minuscule hybrid chip. It’s actually pretty incredible the level of miniaturization they’ve achieved. Finding room in the faceplate won’t be an issue. Besides, she’ll only need one. I’m pretty sure the fact that people have two sets of semicircular canals is a byproduct of our symmetry, rather than a matter of necessity.

Journal Entry #26

The accelerometer and gyroscope will only be enough to tell her where in space her head is located. She’ll need to deduce the rest from there.

A sense of proprioception can be achieved via a digital encoder in each joint. Two or three in multi-axis joints like shoulders, wrists, and hips. These sensors will make her aware of the rotational angle of each joint at all times, allowing her to calculate the position of her torso and extremities in relation to her head. In addition, they’ll allow a programmed rotation limit that will prevent her from overextending her joints.

Proprioception is a sense that people tend to take for granted, to the point where most people aren’t even aware they have it. However, it’s so vital to basic motor control that disorders of the proprioceptive system are serious enough to be considered disabilities. When building sensory systems from the ground up, you have to keep things like this in mind.

Journal Entry #27

I want to avoid giving her overtly human facial features. It’s far too easy to fall into the uncanny valley, even when using a somewhat stylized design. Just a simple vacuformed white plastic mask is enough to give some people the creeps.

I think I’d like to go more abstract than that. Smooth. White. Featureless. Almost shield-like. I’d still like her to be able to express herself, but perhaps via simple animated iconography rather than mechanical means.

Don’t ask me how one would go about displaying animated iconography on a convex faceplate. I haven’t thought that far ahead yet.

Journal Entry #28

I’ve got it. A convex electrophoretic display. Just like the one she’ll use for her charge indicator, but fitted perfectly to the contours of her faceplate. Her entire face would be a display surface.

Changing expressions would be a simple matter of refreshing the display. The fact that electrophoretic displays require no backlight means reduced eyestrain during long teaching sessions. The low power consumption means longer running times. On top of all that, it would look absolutely gorgeous.

There is one significant stumbling block, however. I’ve never actually seen a convex electrophoretic display before. Curved displays are everywhere nowadays, but convex? I’m not entirely sure it’s possible.

I guess if I’m going to be fabricating a robot from scratch, I might as well start with the uncertainties.

Journal Entry #29

Hypothetically, let’s assume a convex electrophoretic display is possible. What kind of face should be displayed on it? I’ve been toying with the idea of eyespots, like on the wings of a moth. Non-functional visually speaking, but enough to give the impression that someone is home.

Her eyespots wouldn’t be quite as striking as a moth’s. The intent isn’t to frighten, but engage. I like the idea of two simple rings. Light gray on stark white, one larger, one smaller, situated one above the other on the right side of her face. Perhaps that’s straying a little too far from what’s recognizable as human, but I feel it has a sort of haunting beauty.

Besides, I’ve done some rough drawings, and even with a vertical eyespot arrangement, her expressions would be immediately recognizable. Imagine each ring is made up of two circles. A gray outer circle, with a white inner circle layered on top. Layer that on top of a white background and it appears as a ring. Now, split each circle into fifteen degree radial segments. By removing segments, readable facial expressions can be displayed.

Remove the bottom half of the inner circle, and her expression becomes one of excitement. Remove the top half, and suddenly she’s displaying half-lidded eyes of boredom. Rotate those fifteen degrees either clockwise or counterclockwise, and she begins to express anger or sadness, respectively. Removing the top half of the outer circle gives the impression of being asleep, while removing the bottom half results in a bright and cheery smile. And that’s only what I’ve managed to coax out of the pattern so far. I feel as if I’ve only scratched the surface.

Of course, this is all assuming I can create a working convex display. It’s entirely possible that I’ll need to settle for a static, neutral expression. Not ideal, but what are you going to do?

Journal Entry #30

If her facial features are going to be represented by false eyespots, her true eyes will need to be hidden from view. Conspicuous camera lenses would immediately draw the eye, ruining the illusion.

I think I’d like to try obscuring the two cameras behind pinhole apertures. Miniature cameras like the ones used in mobile devices tend to have pinhole apertures anyway, so the sensor would still be receiving a complete image. The end result would appear as just a couple of inconspicuous black dots, positioned an inch or two apart in the center of her forehead. Barely even noticeable.

This would however, result in one minor behavioral oddity. Since the hidden cameras would have only a narrow viewing window, they would need to remain fixed in place. Looking around would entail swiveling her entire head, like an owl. It’s possible that some might find behavior like that off-putting.

I think it would be kind of cute, honestly.

Journal Entry #31

Still working on that convex display. I’ve got some ideas for how to make it work, but it involves some specialized and difficult to obtain materials. It’ll take a while to get everything gathered together.

Please stand by.

Journal Entry #32

I did it. It took some serious trial and error but I did it. I figured out a way to suspend electrophoretic microcapsules in a UV curing liquid resin, and airbrush it onto the inside of a frosted polycarbonate shell that had been given a graphene coating to turn it into a transparent electrode. Then I gave the whole thing a thorough once over with an ultraviolet lamp, and stenciled on the inner electrodes using a conductive metallic paint. The result: a spray-on electrophoretic display.

I’m going to refrain from revealing the precise details of how this was done. This is a potentially disruptive technology, and my patent application is currently pending.

The full complexity of a matrix display was slightly beyond my manufacturing capabilities. I had to settle for a segmented display, meaning a simplified electrode pattern that can only display a small number of predesignated facial expressions. That was the initial plan anyway though, so I’m not particularly broken up about it.

I should mention that what I’ve created is just a rough prototype, so it’s not suitable for use in the final product. It’s made from a polycarbonate cereal bowl, embarrassingly enough.

Journal Entry #33

Aaand I almost forgot to include an off switch. The goal was always to keep her operational as much as possible, but that doesn’t mean she won’t need to be shut down occasionally for maintenance.

The switch should probably be a simple button. Her plans are so clean and sleek-looking at this point, it would be a shame to muddy up the design with a conspicuous button. I could hide it beneath her neoprene skin, but that would leave her prone to accidental shutdowns.

Actually, now that I think about it, the backside of her faceplate is still completely vacant. Installing the switch back there would keep it hidden from view, and it’s unlikely to be pressed unintentionally. You’d have to reach under her chin and up inside her head to access it.

Yeah. That works.

Journal Entry #34

The blueprints are finally complete. I’m ready to begin fabrication. Adjustments in design will need to be made on the fly, of course. It’s pretty unlikely that I’ve managed to foresee every problem that will crop up during the manufacturing process. That’s just the way things work.

Still, I feel like I’ve hit a significant milestone. It feels good.

Journal Entry #35

Mariimo. Her name is Mariimo.

I’ve had that name rattling around in the back of my mind for a long time now. I was afraid to fully commit to the name at first, but now that I can fully imagine what she’ll look like, how she’ll feel, the way she’ll move... it suits her perfectly.

The name is derived from Aegagropila linnaei, a filamentous green algae that forms colonies of tumbling velvety spheres on the bottoms of cold water lakes. The Japanese call them marimo, a term that translates roughly to “ball seaweed.” I keep a small colony of my own in an aquarium in my office.

I realize it’s a bit of a cliche to give a robot a Japanese name, but I can’t help but fall in love with it. To me, marimo represent the beauty hidden in things that most would overlook. It’s my hope that when Mariimo is activated for the first time, her curiosity and unique perspective will allow her to find beauty in things that even I had never considered.

I can’t wait to finally meet her.