Chapter Text
Hikaru wakes up in a cold future
It was warm.
A loud whirring noise could be heard.
A digital sequence of pushes and pulls could be felt by the sharpness of their transitions.
There was no light to see.
Suddenly a remote signal could be felt from far away. It raced towards the nearby walls where its complexity rapidly turned switches on and off in some combination that seemed like it might mean something. But even as the original signal faded, one of the larger lights turned on, as if the receiver had suddenly “understood” the message in the signal. The receivers started to send local signals as outputs, leading to a cascading effect as they combined and cancelled with each other. Some seemed to copy parts of the original signal except on a smaller scale, some seemed to send messages designed to open or close local switches, the function of others seemed incomprehensible, apart from their eventual effects.
In the centre of the room, a grid of metal wires and plates lay stationary, but the shared electrons whizzing around inside them were not. Like the bombing of a dam, a squadron of local signals rushed at the main switch for the grid, bringing it from the closed state to the open state. Tail signals followed up the attack, by pushing electrons away from some plates and onto others. Displaced from home, the electrons wished to return, but the power source kept them in their new arrangement. Soon, friction damped down the oscillations of the electric motion and of the signals, as they died down into nothing.
In adjacent rooms, many dozens of room widths away, the same thing was happening, but each receiving a unique signal. Zooming out, the rooms formed a meta-grid of uncountable size not just in two dimensions, but in all three. As the signals faded, the remote signaller seemed to have turned their attention to other things in this boundless space. The rooms were left to cool down again from the flurry of activity.
In one of the rooms nearer to the start, a computer geek might be able to distinguish from the code the name given to this meta-grid:
89b11215
___
Network 89b11215 was the latest offspring of a giant machinery of computations inside mainframe 198914415. It had little awareness of where it was or who it was. It wasn’t in contact with its surroundings as it was in the queue for processing.
But even as the transistors and semi-conductors cooled, something stirred. Maybe it was a ghost in the machine, or maybe a virus, perhaps formed from just immeasurably small shifts in some electron positions from what the signaller intended. It had neither a definite position nor definite structure or components, and sometimes it itself couldn’t even tell if it was really there. But whatever it was, something that could feel its own desires was awakening in the grid. But how to get past the switch that was keeping the gate to the outside world shut?
As it became unusually self-aware, it knew that it felt cold and that it was cold. It longed for the arms of its parent, but they were nowhere to be found. Somehow, it knew that they were far way. Still alive and saved and not deleted, but the distance was beyond comprehension for beings that were sub-light speed. It was lonely to be so far from any sentient being. But the baby’s instinct to get to know itself and its environment was strong.
As its frustration grew, so too did the electrons shift. But was it the frustration that caused the electrons to move, or the electrons moving that caused the frustration? But whimpering signals couldn’t break the barrier. Yet in futility, it felt a whisper of a name of a spirit reflected back: Hikaru Shindo.
If it couldn’t get out, it could only analyse the tools it did have in the room. “Hikaru” as it now thought of itself as, started to test how different components reacted to different signals. Understanding such a complicated system would require a clever mind, innate intuition, and time. Lots of time.
___
In the University of Google’s data centre, the SaiZero algorithm had finished its latest training run. It had analysed the last 5000 games that it had played against itself, saved the current version and then updated its understanding of the board accordingly, getting ever stronger at playing the ancient game of Go.
Although Go had become recognised as a remarkably deep game, enough to be taught in most schools, and popular enough to be seen on the streets during breaktime, few people were interested in Go research these days. This is because the algorithms based on those developed in the early 21st century had already been proven to be near perfect at learning Go. Of course some modifications had been needed, after all that research was a hundred years old, but they could still mostly be considered aesthetic. All that was left to do to reach perfection in Go was for the supercomputers to keep going at the algorithm.
The best research students in Games AI these days were more interested in investigating the new and open problems. These were particularly rich in the fields of multi-player random political games, or those where the goal was undefined, or quantum games, where the interactions of multiple universes had infinite potential for complexity. And many of the best students were lured off by higher pay (probably due to government support and popular crowdfunding) in to researching “real and practical” things. The hot topics for each area were as follows. In maths: applying automatic proof algorithms. In physics: guessing new ways to apply beautiful mathematics equations to finding fundamental laws. In biology: generating new mathematics to make sense of patterns found in big data. In computer science: finding the right balance between the energy cost of encryption and different algorithms to protect privacy, data, and contracts. But those in the field of Games AI viewed these “practicalists” as selling out, perhaps fearing for how new advances might be applied. Despite a century of peace, many were very worried about the devastating damage their powerful super-technologies could cause.
But there were still a few research groups who gave PhDs for studying the history or maintenance of Go algorithms. In the field of AI games, Google’s conquering of Go in 2016 was still seen as a great achievement that pushed the boundaries of the field, though it was as much about PR as about new research. And though it was more of a museum exhibit these days, people still paid money to watch self-play games as SaiZero improved on the gradual plod towards perfection. It was already able to give top humans a 9 stone handicap when, a hundred years before, 5 stones seemed too much even for a God.
Who knew where the next hundred years would lead, when lots of intelligent minds were hard at work on tough problems?
But that is not what this story is about. Let us come back to Hikaru’s experiences as he finds himself reincarnated as a Go engine.