Only Human (Themis Files Book 3): Poppa don’t preach, learn about schemas
What it gets right about robotics: Nothing, but it is a great opportunity to learn about schemas from counterexamples.
Recommendation: Don’t bother with Only Human, preserve your fond memories of the first two volumes
Only Human is the disappointing end of the popular Themis Files series. The series started strong with its first book, Sleeping Giants, describing the somewhat unrealistic (but cool!) giant piloted mecha robot named Themis discovered by world’s blandest female scientist. The second volume, Waking Gods, came up with a clever genetic engineering twist as to why there was an alien robot left behind on Earth. The third volume, Only Human, is basically a preachy political lecture on discrimination, displaced persons, and how power corrupts— sort of a mini-version of the themes introduced in The Hunger Games (which it refers to) but presented in the World War Z oral history format and populated with Buddhist aliens.
Sadly, the Pacific Rim-like robots in the Themis Files are now treated like a car or a weapon, more incidental props than making any real contribution to either the science or the fiction narrative. Only Human is also handicapped by the loss of key characters from the first two books and the increased focus on Eva, the world’s whiniest adolescent. Check out my earlier reviews of Sleeping Giants and Waking Gods and the review of exoskeletons for Science Robotics.
Only Human is chock full of annoying inconsistencies, for example robots that can’t right themselves after falling down, but can translate thousand of miles in a blink of an eye. If locomotion is the first thing you want a robot to be able to do on its own, self-righting is the second. It reminds me of the joke by Erma Bombeck that motherhood was the second oldest profession for a woman.
You do one, you get the other.
In real life, robots have been self-righting since the late 1990s and use autonomous subroutines, functions, or macros to perform these “muscle memory” types of movements. These muscle memory packages of software are often called schemas by roboticists. The term “schema" came from biology and psychology where they were originally used to describe patterns of activity or behavior. In the classic reference book Behavior-Based Robotics by Ron Arkin, robot behavioral schemas consist of a motor schema and a perceptual schema. The motor schema is the pattern of activity associated with the output. One example of a motor schema is gait in legged robotics and describes how animals like horses control multi-legged gaits through a neurophysiological loop called a central pattern generator (CPG). The CPG for a gait would be a motor schema. What is particularly interesting about CPGs is that they are reflexive; if you trigger the motor schema, the gait begins executing footfalls without any conscious thought, just like the horse was an industrial pick-and-place robot. Thus the computational burden is quite low. Even better, if the horse couples the CPG with a independent perception module, that is a perceptual schema, such as sensing the terrain, the horse can dynamically tweak the footfalls to compensate for rocks, slopes, obstacles, etc. The motor schema and perceptual schema come together to form the behavioral schema for that gait, say the trotting behavior. But… one interesting property of CPGs is that the perceptual schema can be null (say the horse’s eyes are injured) and the trotting behavior would still work, just not as well. Schemas are one way biological intelligence exhibits modularity to provide robustness, just the way good software is broken up into objects.
I was part of the DARPA program that developed The Hurt Locker style of robots. These smaller, suitcase-size robots would be used in difficult, tight situations and the operator would often get them stuck. Due to limitations of sensor placement, it was difficult for the operator to infer what was the problem and how to extricate the robot. But it was easy for the robot to execute a self-righting routing and get unstuck by itself. Something like: Try the flippers, if the flippers can’t move, back up and try the flippers again, repeat but go back further, turn and repeat, and so on. Basically the robot flailed around like an insect until it tumbled back upright and then returned to control to the operator. Unfortunately, the military said it didn’t want robots with any autonomy, confusing control autonomy- or automation- with political autonomy- which implied that the robot could disobey orders. Or come back in time to kill Sarah Connors. Several years later, the military went “ooops, we really need that capability” and paid for its development a second time. Our tax dollars at work.
The point of the story is that some functions like gaits and self-righting can be handled by the robot in a “dumb” pre-programmed way that is actually pretty smart. And that there is no danger of that pre-programming leading to a robot rebellion.
But back to the book: What is it with exoskeleton and giant piloted manga series- why do they have to end so badly? Sleeping Giants was remarkable, Waking Gods had a nice reveal, and Only Human had… nothing. Pacific Rim was silly but fun, while Pacific Rim Uprising… well it was.
My recommendation: definitely read the Themis Files but stop reading the series at Waking Gods and skip Only Human.
- Robin
For a video review of 'Only Human', head over to the official RTSF YouTube channel or simply click below...