the human abstraction

Abstraction. It is a wonderful word, for me. When i was much younger, i used to work on everything from designing kernels for big software companies to running high-level robotic interfaces. Abstraction. We were indoctrinated since children about the dogma that everything could be abstracted, that we needn’t know the details as long as we knew the input and output along with edge cases. And we designed our systems that way, a towering edifice of logic.

Yet, I was not satisfied. Computers were the easy abstraction, we built them after-all. Once we reached the limits of von Neumann architecture and sat at the awkward intersection between the macro and the quantum world, there was senescence. And this bother us all; after all, we were used to the high-flying, fast-paced world of technology moving ever forward.

A century ago, a president of the USA declared that they would pump money into solving the ultimate computation device: the brain. It was all back-slapping and champagne for the first couple years as new technologies were rolled out and a vast sea of data was unleashed. Yet, there was no real abstraction, no development of a concrete theory of the mind. When I met someone, knowing their parameters—genome, environmental exposure, etc.—would not help me decipher what they would do next. That was deeply disturbing to me. And many others.

We began in earnest one day after a conference in San Diego several decades back. It started off as a pie-in-the-sky project with goals approaching Apollo, yet our challenge was orders of magnitude more daunting, for we barely knew what we would find in the end and how it would help us. Yes, the human abstraction was an obsession that grew day-by-day to consume entire departments and funding agencies.

You see, the thought of people as something other than a computational I/O, a set of well-designed biological relays, slowly started to fade from our minds. It happened first with C. elegans. The original work of Brenner and colleagues to elucidate every neuron paid off 70 years later, when a simulation of the entire worms nervous system and other organ systems was complete. It could predict with nearly 90% accuracy what a particular worm would do given you knew its genetic composition and environmental parameters. It was a triumph.

Then followed several other model organisms, including the heroic project, MELSTAR, that completely modeled D. mel. That was thirty years on. People could smell it. We were getting close to mammals. And then we would have a new form of computing, one that could bridge the precision of the past with the creativity and adaptability of natural beings. People talked about technological singularity, of creating computers smarter than man. But this was not our aim. And it never became a reality—we feared that event so much that great care was taken to handicap every new algorithm or newly designed circuit to prevent that possibility.

We entered the 22nd century with the expectation that by the 23rd century we would have solved humanity. That we would be able to abstract people. Given a sample of someone’s DNA, a record of their life’s movements (easily obtainable via phone and other records), and several other parameters, we would be able to predict both what they would do and in the near future when computing power had caught up, when they would do it. An old movie, The Matrix i believe it was called, was constantly referenced, many claiming they foresaw a day in the near future when we’d start taking subjects to enslave. But we didn’t need hosts, just parameters.

Somewhere around the early half of the 22nd century, we solved the avian, rodent and non-human primate brains, nearly simultaneously. A team at Stanford-Berkeley (Stanford had completed its hostile take-over of its nearby neighbor several decades back) help show that the interactions and eventual social hierarchy of several littermate mice could be modeled completely. Teams at Tokyo University, Peking University, and several Max Planck institutes then followed with the non-human primate. The avian simulation came out of a Harvard-MIT collaboration.

And so, here we are. Today. The first simulation of a human mind has been run…and failed. For reasons unknown to us, it failed to predict the output of patient 1. Everything had been synced, the patient was continually hooked up to allow real-time monitoring of blood content, relative neurotransmitter counts (microdialysis and voltammetry had advanced quite a bit), brain region-specific mRNA expression levels, and a host of other parameters. We had every possible parameter about patient 1’s life as well: their precise location every 10 seconds since birth, exactly who they interacted with, what they ate, who they slept with (and when)...everything. Short of measuring the activity of every cell in their body, a feat we still couldn’t accomplish even with the monumental improvements in fMRI and other imaging modalities, there was nothing more we could do. No more arguments could be passed to the program.

But there is hope. The simulation still had the same personality as patient 1. That was cause for celebration: for now, when i meet a person on the street, all it would take is a couple of their parameters, and we could skip the unnecessary social interactions; months or years of shared time; and heated, angry conversations. The human abstraction would be complete and then we could known the inputs needed to achieve a specified output. Wonderful.