What do you think?
Rate this book
Darwin's Devices: What Evolving Robots Can Teach Us About the History of Life and the Future of Technology
What happens when we let robots play the game of life?
The challenge of studying evolution is that the history of life is buried in the past—we can’t witness the dramatic events that shaped the adaptations we see today. But biorobotics expert John Long has found an ingenious way to overcome this problem: he creates robots that look and behave like extinct animals, subjects them to evolutionary pressures, lets them compete for mates and resources, and mutates their ‘genes’. In short, he lets robots play the game of life.
In Darwin’s Devices, Long tells the story of these evolving biorobots—how they came to be, and what they can teach us about the biology of living and extinct species. Evolving biorobots can replicate creatures that disappeared from the earth long ago, showing us in real time what happens in the face of unexpected environmental challenges. Biomechanically correct models of backbones functioning as part of an autonomous robot, for example, can help us understand why the first vertebrates evolved them.
But the most impressive feature of these robots, as Long shows, is their ability to illustrate the power of evolution to solve difficult technological challenges autonomously—without human input regarding what a workable solution might be. Even a simple robot can create complex behavior, often learning or evolving greater intelligence than humans could possibly program. This remarkable idea could forever alter the face of engineering, design, and even warfare.
An amazing tour through the workings of a fertile mind, Darwin’s Devices will make you rethink everything you thought you knew about evolution, robot intelligence, and life itself.
The challenge of studying evolution is that the history of life is buried in the past—we can’t witness the dramatic events that shaped the adaptations we see today. But biorobotics expert John Long has found an ingenious way to overcome this problem: he creates robots that look and behave like extinct animals, subjects them to evolutionary pressures, lets them compete for mates and resources, and mutates their ‘genes’. In short, he lets robots play the game of life.
In Darwin’s Devices, Long tells the story of these evolving biorobots—how they came to be, and what they can teach us about the biology of living and extinct species. Evolving biorobots can replicate creatures that disappeared from the earth long ago, showing us in real time what happens in the face of unexpected environmental challenges. Biomechanically correct models of backbones functioning as part of an autonomous robot, for example, can help us understand why the first vertebrates evolved them.
But the most impressive feature of these robots, as Long shows, is their ability to illustrate the power of evolution to solve difficult technological challenges autonomously—without human input regarding what a workable solution might be. Even a simple robot can create complex behavior, often learning or evolving greater intelligence than humans could possibly program. This remarkable idea could forever alter the face of engineering, design, and even warfare.
An amazing tour through the workings of a fertile mind, Darwin’s Devices will make you rethink everything you thought you knew about evolution, robot intelligence, and life itself.
288 pages, Hardcover
First published April 3, 2012
11 people are currently reading
167 people want to read
Ratings & Reviews
Friends & Following
Create a free account to discover what your friends think of this book!
Community Reviews
Displaying 1 - 14 of 14 reviews
June 26, 2022
Possibly the first time I've seen pop culture references cited in the footnotes?
Very simplified and aimed at an extremely popular audience - letting us in on what science is really like, explaining everything in very plain language, assuming we know nothing. So probably quite good for that particular audience, but not so much for me.
Very simplified and aimed at an extremely popular audience - letting us in on what science is really like, explaining everything in very plain language, assuming we know nothing. So probably quite good for that particular audience, but not so much for me.
May 20, 2017
What does evolving robots teach us about the history of life? or the future of technology? These are questions that are too large to be cashed in by this book. Instead, what we see is how this science is done from rather too close for my tastes:
* elaborating details while remaining non-technical in a way that promises no broader understanding of evolutionary patterns or natural laws,
* conforming to the relevant scientific literature in a way that feels defensive,
* addresses philosophical and conceptual questions in a way deliberately designed to be confusing for the stated purpose of hazing students,
* jokey but not witty in a way that seems unseemly for written work,
* married to a particular experimental approach,
* and providing a perspective of the technological future that is as narrow as the military's next funding round.
Having said all this, evolutionary robotics is still too compelling a subject to be taken down by even the most mis-focused presentation if that presentation is made sincerely. There are a number of great references, and you'll get a nice crash course in the area, as well as some vivid demonstrations of how not to proceed methodologically. If you are a student, you will get a feel for what it's really like to work in this area (and particularly for what it's like to have the author as your advisor, who does a lot to give credit to students and collaborators, which is a really decent thing to do). The book even ends on a set of apologies for not given a broader picture of evolution, evolutionary robotics, or bio-mechanics and provides references for where you can fill in the picture.
I should spell out more about my problems with the approach presented here. The hypothesis is that evolutionary robotics can, by virtue of being embodied, be subject to physical constraints and dynamics that can too easily be mishandled by simulation. This seems reasonable. However, this is mistakenly taken to mean that we should use robots instead of simulating, instead of doing both. There are two problems with this: 1) using simulations will help you find the scenarios most worth the significant trouble of using robots to evaluate. This is why computer models are used so significantly in industry: to reduce the expense of physical prototyping by better directing those efforts. 2) The hypothesis that evolutionary robots will be more appropriate for finding the right evolutionary dynamics should be treated as a hypothesis, itself needing experimental validation, instead of being treated as an axiom.
Overall, this is not a book for a popular science audience, but maybe is instead written for an undergraduate presumed to be a proxy for such a reader or someone more narrowly specialized. I think almost anyone interested in models of life would be better off reading "Bio-Inspired Artificial Intelligence: Theories, Methods, and Technologies" (https://www.goodreads.com/book/show/4...). However, if your interests are narrowly in-line with the author's, then you might find this book profitable.
* elaborating details while remaining non-technical in a way that promises no broader understanding of evolutionary patterns or natural laws,
* conforming to the relevant scientific literature in a way that feels defensive,
* addresses philosophical and conceptual questions in a way deliberately designed to be confusing for the stated purpose of hazing students,
* jokey but not witty in a way that seems unseemly for written work,
* married to a particular experimental approach,
* and providing a perspective of the technological future that is as narrow as the military's next funding round.
Having said all this, evolutionary robotics is still too compelling a subject to be taken down by even the most mis-focused presentation if that presentation is made sincerely. There are a number of great references, and you'll get a nice crash course in the area, as well as some vivid demonstrations of how not to proceed methodologically. If you are a student, you will get a feel for what it's really like to work in this area (and particularly for what it's like to have the author as your advisor, who does a lot to give credit to students and collaborators, which is a really decent thing to do). The book even ends on a set of apologies for not given a broader picture of evolution, evolutionary robotics, or bio-mechanics and provides references for where you can fill in the picture.
I should spell out more about my problems with the approach presented here. The hypothesis is that evolutionary robotics can, by virtue of being embodied, be subject to physical constraints and dynamics that can too easily be mishandled by simulation. This seems reasonable. However, this is mistakenly taken to mean that we should use robots instead of simulating, instead of doing both. There are two problems with this: 1) using simulations will help you find the scenarios most worth the significant trouble of using robots to evaluate. This is why computer models are used so significantly in industry: to reduce the expense of physical prototyping by better directing those efforts. 2) The hypothesis that evolutionary robots will be more appropriate for finding the right evolutionary dynamics should be treated as a hypothesis, itself needing experimental validation, instead of being treated as an axiom.
Overall, this is not a book for a popular science audience, but maybe is instead written for an undergraduate presumed to be a proxy for such a reader or someone more narrowly specialized. I think almost anyone interested in models of life would be better off reading "Bio-Inspired Artificial Intelligence: Theories, Methods, and Technologies" (https://www.goodreads.com/book/show/4...). However, if your interests are narrowly in-line with the author's, then you might find this book profitable.
September 30, 2018
Before I read this book I had no idea that people built robotic fish for research. When I thought of robots I thought of Asimo or robots that move on land. What can I say. i thought the book was interesting to read. Thought I did not understand some of the stuff but, it was fascinating. It makes me want to build my own fish robots. Tadro was fun to read about even though I didn't understand everything. I also liked when he wrote about the fist vertebrates. I had never even wondered about what the first vertebrate might have looked like.
April 11, 2020
More than a century ago, by the same twisted logic, Social Darwinism was born: a blind transfer from one emerging science to find solutions in another field. The trick is the solutions fit well with the hopes and beliefs of the prophet. And it has to be an emerging science or people will know the fraud.
June 12, 2012
In my opinion, the first half of this book prior to chapter 6 is an indispensable introduction to the world of evolutionary robotics - it reaffirms some of the limitations in general evolutionary thinking and address the concerns and laments Darwin had himself about evolution - how it is not as simple as it looks, and that natural selection is still a part of our live in other ways (particularly behavioral evolution). However, the second half of this book grows tiring. It's not the heavy mathematical considerations or the extensive review of methodology, but surprisingly is the authors voice - it becomes too personal and impersonal at the same time.
He tries to remedy this in the last section of the last chapter by apologizing for his lack of mention of other's works - and kind of gives the reader a sense that this is more of a memoir of the work done by his lab, but otherwise, interesting first half - the second half would make for a better technical OR personal book, but doesn't belong with the first parts. I would recommend this to anyone interested in evolution or robotics, but probably not to biologists or engineers in general.
He tries to remedy this in the last section of the last chapter by apologizing for his lack of mention of other's works - and kind of gives the reader a sense that this is more of a memoir of the work done by his lab, but otherwise, interesting first half - the second half would make for a better technical OR personal book, but doesn't belong with the first parts. I would recommend this to anyone interested in evolution or robotics, but probably not to biologists or engineers in general.
March 12, 2016
A thoroughly enjoyable and well-written journey thru the trials of scientific exploration, embodied robots, and evolution. Mr. Long presents the true tale of his attempts to prove the origins of the vertebrae. His matter-of-fact telling and occasional bits of humor are entertaining as well as enlightening. I only give four stars to books that enliven my brain and cause the arcs of my thoughts to travel new paths and Darwin's Devices is such a book.
I originally picked it up because I like to build robots and this book looked to give me some tasty tips. It didn't. But it did cause me to think about robots in a different light. Chapter 5, "The Life of the Embodied Mind", is very thought provoking and a chapter I'll read over and over again.
Besides, any book that references Valentino Braitenberg's Vehicles is well worthy of attention to even the dilettante roboticist!
I have to admit, some of my favorite parts are those few times when he explains why his failures were actually successes. Science rules!
I originally picked it up because I like to build robots and this book looked to give me some tasty tips. It didn't. But it did cause me to think about robots in a different light. Chapter 5, "The Life of the Embodied Mind", is very thought provoking and a chapter I'll read over and over again.
Besides, any book that references Valentino Braitenberg's Vehicles is well worthy of attention to even the dilettante roboticist!
I have to admit, some of my favorite parts are those few times when he explains why his failures were actually successes. Science rules!
November 17, 2014
More than a popular science book, "Darwin's Devices" is a first-hand account of carrying out an experiment in evolutionary robotics, from beginning to end. Long walks the reader through the rationale for the experiment, his methods, results, mistakes, successes, and interpretation. I think it is important that the mistakes and hardships be included, so that science is not made out to be some shiny thing done by geniuses in white coats. Basic research, real science, is weighed down by the sweat, tears, trial, and error that pervade all human effort. Still it moves on. The topic follows the work of John and his students closely. All traces of futurism and crystal-ball-gazing about the progress of technology are confined to the last chapter.
July 11, 2016
Disappointing and tedious
The central thesis of this book is that evolvable robots are more effective for investigating and demonstrating evolutionary systems than software simulations. Unfortunately, I believe the book supports the exact opposite conclusion. Evolutionary algorithms can't be convincingly modeled with three or four agents and a dozen generations, as the ambiguous results of these poorly conceived experiments demonstrate. Thousands or millions of agents and generations can be simulated in software, along with much more interesting and realistic selection pressures, genetically coded attributes, and environmental interactions.
The central thesis of this book is that evolvable robots are more effective for investigating and demonstrating evolutionary systems than software simulations. Unfortunately, I believe the book supports the exact opposite conclusion. Evolutionary algorithms can't be convincingly modeled with three or four agents and a dozen generations, as the ambiguous results of these poorly conceived experiments demonstrate. Thousands or millions of agents and generations can be simulated in software, along with much more interesting and realistic selection pressures, genetically coded attributes, and environmental interactions.
June 26, 2013
If the fossil record is incomplete, where do we turn for answers? Robots, duh. Biorobotics researcher John Long turned to evolving robots, specifically, to try to answer questions about veterbrae evolution in fish. Darwin’s Devices is his tale of the amazing Evolvabots and his quest for answers.
Read the full review on GeekyLibrary.
Read the full review on GeekyLibrary.
October 2, 2012
Interesting overview of evolutionary concepts and how robotic creatures are modeling evolution generation by generation. I particularly enjoyed the final word on robots in military work and as battlefield-evolving combatants.
August 12, 2012
Fascinating that scientists can build extinct creatures, but all in all this didn't keep my attention the way I had hoped.
September 23, 2012
Meandering discussion of the use of mechanical fish to study evolution.
December 1, 2012
mild. not bad. mild.
Read
December 11, 2013Featured on Science for the People show #226 on August 16, 2013, during an interview with author John Long. http://www.scienceforthepeople.ca/epi...
Displaying 1 - 14 of 14 reviews