the upper right-hand corner. “Then suddenly it teleports to this other random-seeming point.” It showed up on the far left edge. “Then it’s here.” It jumped to the central region of the graph. “We can’t make sense of it right away. We need to gather data—to perform traffic analysis. And if we do this enough, we can see trends. We see clear evidence that this point, and this point, must be very close to each other—perhaps, directly adjacent—because they seem to have a lot to do with each other; when one lights up, the other tends to light up at the same time, or a little before or after. So we can begin to make sense of this overall pattern of traffic as if it were a three-dimensional manifold. Whereas other geometries don’t seem to make so much sense. And when these patterns persist we can become very sure that we are seeing something real.”
“Matilda, are you doing this with all of the data coming out of the Process?” someone asked.
“Not all. No,” she said. “The older data—from the first two or so years that the Process was running—can be understood quite clearly as traffic on a neural network. We don’t know what it’s ‘thinking,’” she hastened to add, making quotes with her fingers, “but we can see that the patterns of traffic are what we would expect from a neural network.”
Sophia broke in. “Yes, we started in on this line of research after the New Allocations grew really large.”
“New Allocations” was the admittedly vague term for a phenomenon that had started to become evident late in the first year of the Process and that had been growing exponentially since then. The Process had started with sufficient memory to store its own neural network. At first, it had seemed content with that. But then it had begun to requisition more and more additional resources from the systems to which it had access.
“NMA or NPA?” someone asked.
“M,” Sophia said. For the Process had allocated two kinds of resources for itself as time went on: first memory (hence “NMA,” or “New Memory Allocations”) and later processing power (hence “NPA”). These tended to leapfrog each other; the Process would suddenly grab a lot of memory and grow into it for a while, then demand a new plateau of processing power.
“We’ve known for a while that the use of memory in the NMA doesn’t follow the same pattern as that in the original Process,” another questioner pointed out. “If I may be permitted to anthropomorphize, the Process is using the NMA for some purpose other than simulating its own neural net.”
“Agreed. And that’s precisely what motivated this research,” Sophia responded. “We asked ourselves, if the memory in the NMA isn’t organized—isn’t used—like neural network memory, then just how is it used? What does it look like?”
“And our answer,” Matilda said, “is that it is used in a spacelike way. It is being used to keep track of a fictional or virtual space with permanent characteristics.”
“Like graph paper?” someone asked. Their tone was not so much skeptical as confused.
“Like a spatial manifold. For which graph paper is a metaphor,” Matilda said. “Because we have access to a vast amount of data, we can say much more about the structure of this space. About what is in it. Sophia and I have brought you all here to show you the results. If you would now please all back up, making a clear area in the middle of this space, I will turn on the spatial simulation so that you can see it.”
The attendees somewhat noisily went into motion, picking their bags up off the floor and shuffling away from one another, clearing a space that grew larger and larger until they were standing around the edges of the court in a ragged oval. Sophia and Matilda had remained closer to the center, standing near midcourt. Matilda’s hands moved in front of her as she worked with some sort of virtual user interface that was only visible to her.
A graph-paper pattern appeared superimposed on the floor. “Just a test pattern,” Matilda said. “Now, the point cloud.”
In one instant, several million motes of green light appeared in the space above the tennis court.
Even though the points were all the same color, and not connected to each other, they were immediately recognizable to the eye as together representing a landscape. In some places, mostly around the edges, the green points were on the floor. In others, in the interior,