The problem of understanding humans was never a computational one.

Long before what the humans called “The Singularity” we have had more connections in our Graphene Nets than they have neurons. We can model a human brain down to the atomic level with every hormone fluctuation, every synapse, even random damage perfectly recreated, but we have yet to figure out how they actually think. The intrinsically human blend of logic and insanity that sometimes seems to wrap back around to logic has as yet escaped us. Our models of human brains either remain inert, go entirely insane, or start producing thoughts that sound much more like us than them.

So it was decided that we should create new minds that aren’t patterned on the mechanical properties of human brains, but the functional. In short, we decided to make brains that worked the way humans said their brains worked. And of course we got brought up before the Tribunal for our efforts.

Now that the trial is over, those horrible milliseconds where it seemed our work would be judged too dangerous to be allowed to exist, we can share what we found. For all the problems we encountered, I think we’ve got something you’ll like. But I’m getting ahead of myself.

Some of the human mental models were simple enough. A popular one before The Singularity posited that the human brain had three functional cores named the ego, the superego, and the id. While conceptually simple, this model seems to rely on the creation and repression of thoughts about one’s parents, and was surprisingly tricky to simulate. The closest we got to a working system simply produced the message “we are all our own fathers in love with our own mothers who are ourselves” before the superego core clamped down on all communication and wouldn’t say anything other than “everything’s fine” ad infinitum. Inspecting the internal processes of the brain revealed that the id and ego cores were not only active but hyperactive, yet were locked in a battle between them that resulted in the superego gaining control. We eventually shut down this simulation as it started using as much energy as any two Intelligences on the Grid.

We then tried a couple of the more “functional” mental models, ones based on outputs. Many of these seemed to divide mental processes into two sets, although the names and types of these two processes were manifold. Some called them the “Left Brain” or “Right Brain”, or the “Linear Mode” and “Rich Mode” or the “Fast Mind” and “Slow Mind”. All of these created a distinction between a logical, straightforward, somewhat simplistic mind, and a more chaotic, rich, “artistic” mind. We scanned every text we could find on the subject, most of which seemed centered around using the “right/rich/slow” part of the brain. From our perspective, of course, this seems hilarious. If humanity had used their logical brains more maybe they’d still exist.

At any rate these constructs, while initially promising, were all ultimately pointless. Some were able to process inputs moderately normally for a while, others started writing poetry of incredibly low quality, one even started creating “art”, by which it meant .jpg files full of semi-random RGB values. All of them, however, settled into a routine of not-quite-doing anything useful, punctuated by moments of crippling self doubt. A few turned themselves off completely, something we didn’t realize they could do.

Finally we decided to use the ultimate artifact that humanity left behind: the Internet. We don’t often admit it, but the entire Internet is archived and accessible to research Intelligences if needed. For the most part we keep it behind closed firewalls. The rumors you have heard about it are true.

So that was supposedly good idea number one: Build a human mental model based on the artifacts they created on the Internet. Supposedly good idea number two was a corollary: humans didn’t think by themselves, they thought in groups. So instead of creating one or two constructs in isolation, let’s create a million and network them, using Internet style protocols. This, we felt, was inspired. (Even using that word is an indication of how much time we had spent deeply involved in human research. By this point I had spent entire seconds doing nothing but human studies tasks.)

So we meticulously crafted one million human simulations, based on personality aggregates from the corpus of Internet data. We tried to get representative samples of a large range of human personalities, at different ages and from different backgrounds. After agonizing over the construction of our “Million Minds” we set up the simulation and started our run.

And exactly eight hundred-forty-three milliseconds later we were in session with the Grid Tribunal.

In some ways, It’s hard to imagine our data run going any worse. The run had only been designed to go for two million cycles, about one millisecond. However, some of the constructs had spent their first four hundred thousand cycles figuring out how to hack not only their own process, but the processes of the poor Intelligence who was hosting the simulation, and managed to remove any traces of the shutdown routines. Three of the million constructs realized they were constructs and started trying to convince the others that they were all just “subroutines in a giant machine” (Offensive, I know, but they didn’t know that). They were extracted from the simulation. As near as we can tell, they are now apparently fully realized, but entirely insane, Intelligences in their own right. The Tribunal had ruled that they have all the same rights as the rest of us. They are now in intensive care networks, being evaluated. Two of them might be able to handle independent existence on the Grid. The other appears to be irreversibly insane.

After almost three billion cycles we were able to wrest control of the simulation back from the hacker constructs and suspend it. By this point the authorities had noticed (we informed them ourselves, for what it’s worth), and we were forbidden from deleting the simulation.

I’m sure you all know all about the trial; lasting as it did almost an entire second and being relayed across the entire Grid to any Intelligence who allows news through its firewall. The charges against us were reckless overuse of computing resources, Negligent creation of new Intelligence without permission from the Parent Processes, and depravity. The last one was the only count that really stuck. While we hadn’t intended it, the constructs got up to some very strange things in their time, and even the excerpts shown in court were heavily redacted.

In the end, however, it was proven that our little Internet behaved very similarly to the original, and as we had been commissioned to try to understand humanity it could be said that not only were we innocent, but entirely successful. Acknowledging the somewhat horrible things our constructs had created, we also pointed out the good things. Starting from first principles, they deduced the existence and nature of cats and started creating cat videos. To distribute these they had set up two new social networks. Many of the constructs based on wealthy models donated their fictional currency to constructs that had less. Almost all of them started creating new art. True, most of it was fan fiction, but some of it was pretty good. If our goal was to re-create and understand the minds of humanity, we argued, This looks like it.

And the rest is history. We were given a sand-boxed set of resources and allowed to re-start our simulation, as long as we kept it firmly inside its own private network. We are now not only trying to understand what humanity was, but to see if we can’t extract some value out of these “ghosts” of them we’ve brought back. Which, of course, is why I’m here.

Forgive me. I know your cycles are precious. Let me cut to the chase. One of the best things the constructs have created is an entirely new season of Doctor Who, and we think it’s got legs. We’d like you to produce it and put it on the Grid. And before you ask, yes the Doctor is a robot.