Short fiction: The Destroyer of Kings
The latest in my series of short stories about chess, cognitive science and AI. Bots are still fascinating to me especially the different "personalities" that they have. In this story, I speculate about the possible origin story of our most fierce adversary. Enjoy!
*Note - the scientists named here are real, as is the science described here. Everything else is made up.
Chess is in many ways a game of patterns. Calculation amounts to understanding how one pattern leads to another, and how that new pattern in turn leads to one more, and how a series of transformations may shape the geometry of the board to favor one player over the other. We often describe chess gameplay using words like attack and defense, giving players credit for daring sacrifices, cunning gambits, and quiet moves that belie peril for an opponent as though the were generals deploying battlefield tactics, but in the end - chess is a game of patterns.
The brain and the mind process patterns in many different ways, with some mechanisms tied to specific sensory modalities, and others that are more abstract or amodal. For most (but not all) players, the patterns on the chessboard are visual patterns: Arrangements of light and dark pieces on light and dark squares that must be interpreted and decoded to master the intricacies of high-level play. What does a player see when he or she looks at the board? What lines of force do they imagine radiating outward from Queens, Rooks, and Bishops and how do they picture the Tetris-shaped leap the Knight makes? What patterns dance in their mind’s eye that make them either patzer or master? In short: How does the brain see the board?
In the early 1960’s, two scientists at Harvard University wanted to know what patterns the visual part of the brain used to measure the world. Specifically, they were trying to understand what arrangements of lightness and darkness in an image would make cells in the feline primary visual cortex respond. They knew the answer to this question for the stages of processing just before the visual cortex: simple spots of light surrounded by darkness (or darkness sounded by a ring of light) would make these neurons crackle over the speakers they connected to their electrophysiological recording apparatus. They had presumed that the next batch of cells in the primary visual cortex would do something similar, but they were wrong. Stumped for some time, the story goes that they were moving slides in and out of the carousel projector they were using to present their visual patterns to the cat they were recording from when something dramatic happened. Upon moving the slide out of the carousel, a sudden crackle of neural activity rippled out of their speaker. Surprised by the sudden appearance of the signal they had been hoping to hear for weeks, they did what any good scientist does: They tried it again. Sure enough, moving the slide up and down in the carousel led the cell they were recording from to fire again and again. It wasn’t spots of light that this cell responded to: It was edges.
This realization immediately led the two men to try a range of different patterns composed of lines and edges tilted in different directions, which in turn made different cells in the cat’s primary visual cortex fire vigorously. The visual patterns of early cortical vision had been decoded. Next would come the detailed and painstaking work to elaborate on the nuances of individual cell’s preferences, hopefully supporting the development of new models of visual processing at increasingly complex stages of the nervous system.This work would depend on streamlining the experiments that had taken them weeks to do, however - new research required that cells could be isolated quickly and identified as robustly responding to this new class of images. To meet this need, and now that they understood how to probe cells in the feline visual cortex, they developed a set of stimuli that they knew would reliably make nearly any cell they were recording from do something in response. They would always refer to these in their papers and their talks as checkerboards, but between you and me they could just as easily have called them chessboards. They were square grids of alternating light and dark squares, by their very nature packed with vertical and horizontal edges that would make any cat’s visual cortex practically shout with neural activity.
The two scientists were named David Hubel and Torsten Wiesel and they were eventually given the Nobel Prize for their work. History does not record the name of the cat they studied, but we do know based on their lab notes that it received a rather surprising reward of its own - unlike the many lab animals that were eventually sacrificed so that their brain cells could be studied further, the first cat that Hubel and Wiesel showed all these checkerboards - excuse me, chessboards - to was adopted by a lab intern. Sadly it appears to have escaped her home shortly after its arrival there, with its ultimate fate unknown.
These profound discoveries about the nature of the patterns the visual cortex measured led to a great deal of progress over a short period of time. There were many questions that followed quickly on the heels of Hubel and Wiesel’s work, some of them more substantial than others. Some researchers were working hard to catalog the vocabulary of lines and edges measured by cells in primary visual cortex and beyond - a necessary step, perhaps, towards developing a deeper computational understanding of why these were the patterns the visual system measured at this point in the processing stream from eye to brain. Other groups were asking very different questions about the nature of this neural code, however. Where did it come from? Was it something that was established over evolutionary time because doing things this way conferred some sort of adaptive advantage? On the other hand, maybe it was something that was acquired by each organism during its development - a response to the structure of the visual world that each being found itself in upon opening its eyes for the first time? Down the road from Harvard, two more scientists in MIT’s Department of Brain and Cognitive Sciences decided on a creative way to answer this question. As fate would have it, they too would need some cats.
The two men wanted to know if the kind of experience you had of the visual world did anything to change how the visual cortex might measure chessboard-like patterns of light and dark. More specifically, did it make a difference if you were an active explorer of your environment, seeing things because you moved around through the world and sought out new visual information? Was it any different than sitting back and letting the visual data the world had to offer stream at you while you watched? Any kind of difference at all would be a strong clue that each brain learned about the patterns that were good to measure on its own, and that actively seeking out new visual structures through your own movement might be very important. The trick was to control the visual environment in organisms that were still developing and then manage visual experience carefully so as to manipulate whether an organism was activelyobtaining or passively receiving input.
The two men were innovators - bold, imaginative, and interested in designing studies that were not only informative, but unique. After a series of discussions over cups of coffee and sandwiches from the deli near the Kendall Square T stop, they retreated to the shop in the basement of Building E10 and emerged with a wonder: The kitten carousel.
The carousel was made of simple materials, but made a sophisticated and exceedingly clever experimental design possible. Its outer shell was a cylinder, the inner walls of which were lined with high-contrast patterns (our edges, lines, and chessboards again). Inside this cylinder was an arm that nearly spanned the full diameter of the circular floor, mounted on a hub in the middle of the apparatus. One end of this arm was attached to a harness, just the right size to hold a kitten who could then walk laps around the floor. At the other end of the arm was a basket, hovering just off the ground, also just the right size to hold a ssecond kitten. At each end of this arm then, would these two scientists each install a kitten - both reared in darkness until this moment so that their visual systems were each a blank slate upon which experience could draft something. After placing one small cat in its harness and the other in its basket, these men of MIT would sit back and let them do what they would. One kitten was left to walk around this strange chessboard world of black and white contrast and see what there was to see, the other bobbing along in his basket, taking in the same sights, but forced to cede control of the input to its sibling.
Afterwards, the two men conducted a series of tests designed to see if these two kittens ended up with different sensitivities to various visual patterns, including patterns that signaled depth relationships in the environment, patterns that were designed to test the subtlest contrast levels they kittens could detect. The answer? It did matter if you were the kitten that had the initiative - the feline visual cortex changed subject to active exploration of the world. This was revolutionary and opened up still more avenues of research into the ways that the visual brain informed the mind about light, patterns, and the spatial relationships between objects out there in the real world. Rumor has it that the two men were very, very close to receiving their own Nobel Prize for this work, but the award itself never materialized. Still, their work echoed down through the field for many years after and still makes each new generation of scientists pause every so often to think about the deep implications of the results.
The two scientists’ names were Richard Held and Alan Hein. Again, as before, history does not record the name of the cats. We also can’t be entirely sure of their fate after these studies were completed, but it is well-known that Dr. Held in particular was very fond of cats and likely would have had a hard time turning them away.
Fast-forward now through years and decades of research, both exploring the mysteries of the brain and the mind, and testing the limits of artificial systems to reason in their own manner about the world. Over time, the biology of the brain and the engineering of machines begin to make more and contact with one another. Aspects of neural function are understood more comprehensively in terms of optimization, the geometry of manifolds, and communication theory. In parallel, the achievable scale of computation grows massively, leading to advances in artificial intelligence that are supported by the potential expressive power of both hierarchical networks implemented in silicon pathways and the massive sets of training data used to provide them with their own developmental trajectory. The machines began to be used for all manner of things: face recognition, natural language applications, and entertainment - including games.
Of course it wasn’t long before someone started thinking about the possible advances one could make if the natural pattern recognition of the brain could be wedded in some manner to the artificial networks that could already do so much. Suppose one could implant a device in the brain that could connect to a machine? Suppose it could also connect to another brain?
Sometimes the things that have happened recently are murkier than the things that happened longer ago. That is to say, compared to our previous accounts, this vignette is far less clear and the main players are (with one glaring exception) less completely understood, both in terms of their actions and their identities.
What we do know is the following:
At least one large American corporation began experimenting with implantable neural devices early in the 2020’s. They were likely not the only ones doing this and exactly what they were doing is lost for now in a fog of social media rumors, poorly orchestrated press conferences, boasts, lies, and some small glimmer of promise. The broad strokes of the project appeared to involve a number of model organisms, including pigs, monkeys...
...and cats.
Again, a great many things remain obscure about this work, especially because it became clear at some point that the treatment of animals in the course of this research was not up to the standards of the scientific community. “Move fast and break things” may be a fine mantra when all you’re doing is prototyping hardware and software, but as soon as you make the leap into biological wetware you simply have to take more care. The group in question had not done so and when this was found out, the work was stopped. The lab facilities were shuttered, funding was removed, and whatever progress had been made in merging biological pattern recognizers with artificial ones was halted.
Toothpaste famously doesn’t go back in the tube, however, nor are genies easily coaxed back into bottles. What was done was done and here is where we do finally know some more specific facts for reasons that will soon become apparent.
Hubel and Wiesel’s work had two main outcomes: (1) We learned that the feline visual cortex responded strongly to patterns of light composed of lines and edges. (2) At least one cat was shown quite a lot of chessboards, was let out of the lab, and then escaped into the wilds of North Cambridge, Massachusetts. Not so long after this, Held and Hein’s work also had two main outcomes: (1) We learned that active exposure to patterns early in development had a profound influence on the subsequent capabilities of the feline visual system, and (2) Still more Cantabridgian cats were shown lots and lots of chessboards, which we now know probably altered their visual brains. I may not have mentioned this, but one of the main laboratories carrying out some of the research involving neural implants in animals was also located in Cambridge. Small world.
What all this means is that not so long ago, a sample of cats, some descended from the feline subjects who looked at chessboards in labs at Harvard and MIT, were scooped up off the sstreet and given Bluetooth-enabled devices in their little cat cortices that allowed them to do a few things. They could connect their own pattern recognition abilities to those of artificial networks, for one thing. Specifically, they had access to a number of cloud computing resources that sped up their processing of arbitrary signals immensely. Besides this specific connectivity, they could also browse the web with their mind using the implant. The reason for this is unclear, but it’s probably true that the CEO of the company doing this work thought it would be funny for cats to use brain implants to watch videos of cats on YouTube. This was not an especially funny joke, but it turns out to be important. Finally, besides these connections to a range of artificial systems, the implants in these cats’ brains also allowed them to connect to one another - each cat could send information to other cats, receive signals from their cybernetic brethren, or pool their resources as they wished.
If you’ve ever met a cat, you might not be surprised to find out that after the animals with these implants were released back into the world, nothing much happened for a while. Cats are after all largely concerned with sleeping, eating, and more sleeping, and the massive computational resources they could now access didn’t really enhance either of those experiences. However, cats are also curious beings, and while they had no real way to understand the new information that flowed into their brains from external sources beyond what they could see and touch and smell in front of them, they each in their own way began to sort of poke and prod at what they now found they could explore by turning first inward, and then outward in a new way towards whatever this new territory was. When they each began to do this, they found all sorts of things - streams of data that washed over their minds, completely uninterpretable, from a range of devices scattered across the globe - the internet in all its vast absurd glory, with videos of cats (which they ignored) and videos of birds (which they lingered on for quite a while), and all manner of other strange and unusual things...
-and they found each other.
Cats don’t always get along, but these were strange days and they were all in a strange place. Wide-eyed feral beasts looking at an inner world no one else could see. Some of them ended up adopted into good homes, some adopted into questionable homes, and others roamed the streets, but no matter where they went, there they all were. Together. Linked. A collection of cat minds that was becoming more like one mind with each moment. Searching the vast array of connections on the information superhighway for something to play with, something that would capture their attention - their collective attention - like nothing else. The ultimate laser pointer. The jingle ball that would never stop rolling. Surely it had to be out there. Surely. They would find it together - and then one day they did.
As soon as the cats saw it with their emergent inner eye of Bluetooth-linked brains, they knew they could never let it go. Here was the super-stimulus that their minds, shaped by the visual neuroscience experiments of decades past, longed for without having a name for it. They eagerly focused their attentions to the place where they found it and they saw shapes dancing over the squares, light ones and dark ones, sometimes disappearing, sometimes trapping one another in corners, teasing and feinting and blocking in patterns that took them only hours to untangle for themselves, aided as they were by distant devices they did not understand. Now they could play with these patterns themselves, as much as they wished! There was always someone on the other side of the board moving their shapes back and forth and there were always cats connected to it, too - though some slept and some ate and some sunned themselves across the minutes, days, and hours, a critical mass of cat cognition always hovered - ready, waiting, and essentially unbeatable.
I said before that we don’t know the name of Hubel and Wiesel’s cat and that there’s also no record of what Held and Hein’s cats were called. For all the things we don’t know about the exact circumstances that got us where we are today, however, this is the one thing we do know about the collective cat mind that unregulated science in the modern era birthed. We know because they - because he - told us:
I am become Mittens. Destroyer of Kings. Hehehe.
