Robots may soon be able to reproduce - will this change how we think about evolution?
From the bottom of the oceans to the skies above us, natural evolution has filled our planet with a vast and diverse array of lifeforms, with approximately 8 million species adapted to their surroundings in a myriad of ways. Yet 100 years after Karel Čapek coined the termrobot, the functional abilities of many species still surpass the capabilities of current human engineering, which has yet to convincingly develop methods of producing robots that demonstrate human-level intelligence, move and operate seamlessly in challenging environments, and are capable of robust self-reproduction.
But could robots ever reproduce? This, undoubtedly, forms a pillar of “life” as shared by all natural organisms. A team of researchers from the UK and the Netherlands have recently demonstrated a fully automated technology to allow physical robots to repeatedly breed, evolving their artificial genetic code over time to better adapt to their environment. Arguably, this amounts to artificial evolution. Child robots are created by mixing the digital “DNA” from two parent robots on a computer.
The new design is first sent to a 3D printer that fabricates the body of the robot, then a robotic arm attaches a “brain” loaded with control software inherited from the parents, along with any new components, such as sensors, wheels or joints, selected by this “evolutionary” process. A digital replica of every new robot is also created in a computer simulation. This enables a novel type of evolution: new generations can be produced from a union of the most successful traits from a virtual “mother” and a physical “father”, combining the benefits of fast but potentially unrealistic simulated evolution with the more accurate assessment of robots in a real physical environment. The new robots therefore inherit traits that represent the best of both types of evolution.
While this technology can operate without a human in the loop, it also allows for collaboration with a human “breeder”: just as humans have been selectively breeding crops since the dawn of farming, the robot breeder could influence selection of robots with particular traits. One might even imagine breeding farms, producing robots adapted to specific conditions and user requirements. They might be bred for qualities such as battery life or carbon footprint, just as we breed plants for drought-resistance or taste.
Such farms should be subject to the same strict controls and ethical considerations as, say, breeding of genetically modified crops, for example enabling an entire facility to be shut down at the touch of a button, or limiting supplies of raw materials. Furthermore, it is also important to consider the possibility that evolution might result in robots exhibiting malicious or harmful behaviours and put appropriate preventive measures in place.
The idea of digital evolution – imitating biological evolution in software to successively breed better and better solutions to a problem over time – is not new. It can be traced back to the 1960s when engineers in Germany programmed a computer to evolve the optimal design of a jointed plate subject to turbulent airflow. Since then, “evolutionary algorithms” operating inside a computer have been used to design everything from tables to turbine blades, by simply telling the evolutionary process what metric it should seek to optimise (for example, the power generated by the turbine blade). In 2006, Nasa sent a satellite into space with a communication antenna designed by artificial evolution.
We are now at a breakthrough moment.While scientists have always been confident that digital evolution could be effective as an optimisation tool, its creativity in producing original and unusual designs that would not have been conceived by a human has been more surprising. The creativity of biological evolution is clearly apparent in the natural world. In the Cuban rainforest, vines have evolved leaves shaped like satellite dishes that amplify the signals propagated by echolocating bats to direct them to its flowers, increasing pollination. In the freezing Southern Ocean, fish manufacture their own “anti-freeze” proteins to survive.
But numerous examples of creativity in digital evolution have also been observed. Asked to find behaviours for a six-legged robot that would enable it to walk even if it had been damaged, digital evolution discovered multiple ways of walking that used only subsets of the legs, even discovering a way for the robot to move if allits legs had been snapped off, by shuffling along on its back. In another case, it evolved an electronic circuit on a chip where elements of the circuit were disconnected, exploiting electromagnetic coupling effects specific to flaws in the silicon on the actual chip.
Digital evolution now finds application in avenues that we might imagine to be uniquely human, for example in creating music and art (even winning an award in a University of Wyoming art competition where judges were unaware the winning picture was created by an algorithm). While this may sound to the uninitiated like artificial intelligence, digital evolution is a specific subset of that wider field.
The idea of harnessing evolution to design robots is particularly appealing, especially in cases where humans have little knowledge of the environment the robot should operate in – for example, undersea mining, clean-up of legacy waste inside a nuclear reactor, or using nano robots to deliver drugs inside the human body. Unlike natural evolution which is driven simply by the goals of “survival and reproduction”, artificial evolution can be driven by specific targets. Once this evolutionary process is set in chain, and with the technology outlined above, of a computer system instructing a 3D printer to create improved models of the robots for these particular environments, we have the beginnings of a theoretical framework for a self-sustaining robot population that is able to reproduce itself, and “evolve” without too much input from humans.
Which isn’t to say that humans would be redundant. Digital evolution will probably be a collaborative process between human and machine, with humans providing descriptions of whatis desired while evolution provides the how. So for example a human might demand an “energy-efficient robot made from sustainable materials to move heavy waste inside a reactor”, leaving evolution to figure out how this can be achieved. Advances in manufacturing technology that facilitate automated and rapid prototyping in a range of materials including flexible soft plastics have played an important role in enhancing our ability to replicate evolution on practical timescales.
If this all might seem to border on science fiction, there is a serious point. Robots clearly have a role to play in our future, whether in revolutionising healthcare or undertaking tasks too dangerous for humans. We are rapidly using up stores of raw materials on our planet, and current manufacturing processes increase carbon emissions and create serious problems with waste disposal. Perhaps the creativity of evolutionary methods will enable the design of new types of robot, unfettered by the constraints that our understanding of engineering, physics and materials science impose on current design processes.
From another perspective, until we discover extraterrestrial life, biologists have only one “system” on which to study evolution. Just as the Large Hadron Collider provides us with an instrument to study the intricacies of particle physics, perhaps a reproducing system of robots provides a new instrument to study fundamental questions about life itself.
Emma Hart is a professor in the