Home / Science / Technology
Artificial Intelligence
Evolving Competing Robots
In a Swiss laboratory, a group of ten robots is competing for food. Prowling around a small arena, the machines are part of an innovative study looking at the evolution of communication, from engineers Sara Mitri and Dario Floreano and evolutionary biologist Laurent Keller.
They programmed robots with the task of finding a "food source" indicated by a light-coloured ring at one end of the arena, which they could "see" at close range with downward-facing sensors. The other end of the arena, labelled with a darker ring was "poisoned". The bots get points based on how much time they spend near food or poison, which indicates how successful they are at their artificial lives.
They can also talk to one another. Each can produce a blue light that others can detect with cameras and that can give away the position of the food because of the flashing robots congregating nearby. In short, the blue light carries information, and after a few generations, the robots quickly evolved the ability to conceal that information and deceive one another.
Their evolution was made possible because each one was powered by an artificial neural network controlled by a binary "genome". The network consisted of 11 neurons that were connected to the robot's sensors and 3 that controlled its two tracks and its blue light. The neurons were linked via 33 connections - synpases - and the strength of these connections was each controlled by a single 8-bit gene. In total, each robot's 264-bit genome determines how it reacts to information gleaned from its senses.
In the experiment, each round consisted of 100 groups of 10 robots, each competing for food in a separate arena. The 200 robots with the highest scores - the fittest of the population - "survived" to the next round. Their 33 genes were randomly mutated (with a 1 in 100 chance that any bit with change) and the robots were "mated" with each other to shuffle their genomes. The result was a new generation of robots, whose behaviour was inherited from the most successful representatives of the previous cohort....
With the yoke of natural selection relaxed, processes like genetic drift - where genes pick up changes randomly - were free to produce more genetic diversity and more varied behaviour. After around 500 generations of evolution, around 60% of the robots never emitted light near food, but around 10% of them did so most of the time. Some robots were slightly attracted to the blue light, but a third were strongly drawn to it and another third were actually repulsed.
Mitri, Floreano and Keller think that similar processes are at work in nature. When animals move, forage or generally go about their lives, they provide inadvertent cues that can signal information to other individuals. If that creates a conflict of interest, natural selection will favour individuals that can suppress or tweak that information, be it through stealth, camouflage, jamming or flat-out lies. As in the robot experiment, these processes could help to explain the huge variety of deceptive strategies in the natural world.
Book Review: Guilty Robots, Happy Dogs by David McFarland
(out of 5 stars)
David McFarland, someone well-versed in biological robots and zoology, offers up this quick philosophical (not technical) discussion of just how we go about identifying 'alien' minds. 'Alien' here refers to non-human minds, not the ET variety, specifically those of animals and robots. He assumes as fundamental the need to identify both rationality and subjectivity in an 'other' before we could ascribe to it a mind. Most of the book involves dealing with the numerous and convoluted problems associated with those identifications.
To move his ideas along, McFarland uses his dog Border as the animal example, and a conceptual security robot for the other. Throughout the early parts of the book, the reader gets an intro to 'mindless machines' and the role design plays in both animals (through natural selection, environment, etc) and robots (engineer, programmer, etc).
The bulk of the book involves traditional philosophical considerations of intent, functionality, rationality, subjectivity, feelings, knowledge, and mind. Much of this discussion will be familiar to readers of Daniel Dennett (and Dennett is frequently referenced) and/or general philosophy of mind. There are some interesting applications of these concepts to robots (especially), but I'd advise the novice philosophy reader to find a quiet room and have an optimal mind set before proceding through the middle sections (as I would advise on any good philosophy book).
Toward the end, the reader gets stronger discussions of mind as they may (or may not apply) to robots and animals. Many of the contradictions are pointed out, as well as the inherent difficulty (impossibility?) of determining the mind, mind set, or subjectivity of anything which might house them.
The end of the book falls off the truck, unfortunately. Throughout the chapters, McFarland clearly appears to be laying groundwork for his conclusions (and yes, I use this word in the philosophical not empirical sense), only to turn in a 'hedging all bets' card in the epilogue without any real opinion. He offers what may be possible, states that philosophers are all in disagreement, and proposes weakly that its basically up to the reader to determine what is going on inside that skull/robot. I recognize (both from reading this book and previous knowledge of many of the subjects) that a conclusion of any sort would not be likely in strong empirical terms, but at least McFarland should have let the reader know this was an exploration without an actual purpose other than to discuss the issues (does the Intentional Stance come into play here in his narrative?). I never expected McFarland to state whether he thought his dog had a mind, but he presented enough points of view that I expected him to accept one at some point. He never did. It was all one big lecture for the reader (enjoyable though it was).
The strongest points for me were his determination that mind and consciousness were just as much products of evolution and purpose (or for the robot, design and purpose) as any other phenotypic effect. His one strong conclusion was that we cannot expect an animal or a robot to ever have a mind or consciousness like ours. They don't have human brains, haven't been selected (or designed) under the same conditions, and therefore, if they have minds at all, those minds would conform to the specific needs and conditions of their respective developments.
Guilty Robots was worth the read, and with a stronger finish this would have been a four-star review. However, the weak ending, the progressing obscurity of our main characters (security robot, Border), and a reader-must-decide 'conclusion' somewhat spoiled an otherwise solid effort (but not enough to render it a waste of time). Three and one-half stars.
