“We are no longer particularly in the business of writing software to perform specific tasks. We now teach the software how to learn, and in the primary bonding process it molds itself around the task to be performed. The feedback loop never really ends, so a tenth year polysentience can be a priceless jewel or a psychotic wreck, but it is the primary bonding–the childhood, if you will–that has the most far-reaching repercussions.”
— Bad’l Ron, Wakener, Morgan Polysoft
In SMAC, the tenth-tier pure scientific technology Digital Sentience represents the advent of fully-artificial general intelligence on Planet. Its first prerequisite is Industrial Nanorobotics, which implies the extensions of the previous state of the art in AI represented by Pre-Sentient Algorithms. It combines this with the insights into human mind structure required by Mind/Machine Interface from the other side of the tree to make real AI a possibility.
It’s clear from the structure of the technology tree and the length and depth of this quote that Reynolds has put a lot of thought into AI. In particular, there’s enough here to give a lot of insight into how he thinks it potentially impacts the narrative he’s creating. And, therefore, the gameplay he’s constructing to support it.
First off, we should note the bare fact that this discovery comes a full five tiers after Pre-Sentient Algorithms. Its predecessor likely comes as the player is just entering the mid-game, while this advancement happens near the end. On Planet, AI is not merely the result of some clever hacker coming up with one cool trick. It’s a long, arduous, and gradual process of improvement.
Second, true AI does not come out of any deep analysis of humans or any of the amazing biological breakthroughs that came in the mid-game. This is somewhat surprising. We knew before that Pre-Sentient Algorithms were a pure mathematical accomplishment, but that pattern appears to hold all the way up to the end. We learn here that AI, in SMAC, is best cladistically classified as a type of tool and not a kind of person.
Third, the process by which Reynolds imagines AI actually working is fascinating. Traditionally in science fiction, an AI is imagined as a brilliant person that lives in a computer instead of a body. As people have gotten more interested in the topic over time, they’ve focused on the possibilities inherent in easily copying and inspecting the internals of the operating intellect.
But Reynolds comes at it from a different angle. Instead of describing a general AI as some sort of unified super-genius, capable of doing any task equivalently well, he sees it as primarily the right kind of learning/feedback process implemented in software. The plasticity of the AI – its ability to learn – fades after “childhood” as it bonds itself to the task at hand. This implies that the analogous, familiar human pattern is not just a quirk of biology. Instead, Reynolds is putting forward the idea here that it is somehow a logical necessity of what we’d call intelligence.
Hence, the identification of our futuristic software guy as a “Wakener”. His job is primarily to structure the unformed polysentience’s initial experiences so that it becomes the program that solves the desired problem in the most optimal way. In a weird way, he needs the skills of a modern preschool teacher much more than he needs those represented by a modern computer scientist.
In the game, this technology enables the construction of the Network Backbone secret project and enables the adoption of the new Cybernetic Future Society option. The latter represents a society that allows its many of its fundamental operations to be restructured and then largely run by these brand-new AIs.
This results in large bonuses to the faction’s Efficiency, Research, and, oddly enough, Planet scores in exchange for a substantial Police penalty. The idea is that the new AIs are all about efficiency in every area of a faction’s operations. But they take it to the point that the average person on the street’s opinion counts for very little. Ergo, the populace grows more unruly.
Just as the previously examined Thought Control option did, this set of new, orthogonal bonuses reshuffles the possibilities inherent in the social engineering screen. For instance, it makes the Police State/Planned combination thinkable for factions without Yang’s immunity to inefficiency. In fact, Police State/Planned/Knowledge/Cybernetic mitigates this penalty substantially while delivering a massive Research bonus. The effect is probably something like what people thought the Soviet Union could be in the ’50s, when they were at the cutting edge of space exploration.
But there’s another new combination that I find even more intriguing. For instance, the Cybernetic research bonus can be used to cancel out the Fundamentalism research penalty. So one could theoretically move a faction into a Fundamentalism/Planned/Wealth/Cybernetic government. This is a neat combination because it manages to combine good Growth, good relations with the Planet, and an excellent Industrial rating, which is notable because these three are intentionally thematically difficult to accomplish simultaneously. Meanwhile, the Police penalty makes foreign military adventures difficult. Yet it’s not a pure builder approach, because the Fundamentalist bonuses mitigate the army Morale penalty from Wealth while also granting a large bonus to spying.
It’s worth wondering what kind of society such a combination is modeling. My best guess is that this results in something pretty close to what Sister Miriam would likely have been advocating since the nature of the Planetmind became common knowledge. The largely peaceful religious government runs an economy primarily focused on internal development and the well-being of its citizens. At the same time, the AIs set it up so that these human-centric goals can be accomplished with a minimum of waste and impact on the broader environment.