Matthew Griffin, award winning Futurist working between the dates of 2020 and 2070, is described as “The Adviser behind the Advisers” and a “Young Kurzweil.” Regularly featured in the global press, including BBC, CNBC, Discovery and RT, Matthew’s ability to identify, track, and explain the impacts of hundreds of revolutionary emerging technologies on global culture, industry and society, is unparalleled. Recognised for the past six years as one of the world’s foremost futurists, innovation and strategy experts Matthew is an international speaker who helps governments, investors, multi-nationals and regulators around the world envision, build and lead an inclusive, sustainable future. A rare talent Matthew sits on several boards and his recent work includes mentoring Lunar XPrize teams, building the first generation of biological computers and re-envisioning global education with the G20, and helping the world’s largest manufacturers ideate the next 20 years of intelligent devices and machines. Matthew's clients include three Prime Ministers and several governments, including the G7, Accenture, Bain & Co, BCG, BOA, Blackrock, Bentley, Credit Suisse, Dell EMC, Dentons, Deloitte, Du Pont, E&Y, HPE, Huawei, JPMorgan Chase, KPMG, McKinsey, PWC, Qualcomm, SAP, Samsung, Sopra Steria, UBS, and many more.
WHY THIS MATTERS IN BRIEF
For the first time we are seeing how it’s possible for robots in the future to design and evolve, manufacture and assemble themselves without human input.
Experts at the University of Oslo, Norway have discovered a new way for robots to design, evolve and manufacture themselves, without input from humans, using a form of artificial evolution called “Generative design,” and 3D printers – although admittedly the team, for now at least, still has to assemble the final product, robot, when it’s printed.
Generative design is something we’ve talked about several times before and it’s where artificial intelligence programs – creative machines, if you will – not humans, innovate new products – such as chairs and even Under Armour’s new Architech sneakers.
The labs latest robot, “Number Four,” which is made up of sausage like plastic parts linked together with servo motors, is trying out different gaits, attempting to figure out the best way to move from one end of the floor to the other. And while you might look at this video and think it’s weird, or funny remember that this is just the start. Today it’s evolving, trying to learn how to move from A to B in the most efficient manner, but tomorrow – well, it could be “evolving” anything, and all at a much faster rate than humans.
By constantly monitoring its own progress and comparing it with previous attempts, over time it gets visibly better at this simple task.
“It’s now testing variations of its original movement pattern,” says Kyrre Glette, associate professor at the University of Oslo’s research group for robotics and intelligent systems.
The movement isn’t perfect though – of course. After all it’s trying to evolve and evolution is never a simple task. In nature for example around 99.9% of “evolutions” go nowhere and the animals die out, but that’s what evolution is all about – trying new things, failing, and trying again until something works.
In this case the robot number four’s instructions come from a computer simulation and since a computer simulator doesn’t perfectly model the real world there’s bound to be some mismatch between the virtual world and the real world, say the team.
The robot was conceived in a “virtual womb” – a generative design computer program that tries out thousands of different simulations and solutions to how the finished machine could best move around its given environment, which in this case is the carpeted floor of Glette’s lab. Glette and his team don’t tell the computer how to solve the problem, they only introduce certain parameters – like the fact that the robot should move from A to B and the AI, the creative machine, does the rest, iterating thousands of times, keeping the best versions and iterating again.
“It’s artificial evolution. It starts off with very simple combinations of these sausages and simple motors that can move them,” says Glette, looking at an on-screen visualization of the evolution process. Eventually, there are some solutions that manage to move forward just a tiny little bit. Then they are saved and those solutions are then used for coming generations and modified slightly. Eventually you get better and better solutions.”
By running a thousand individual virtual robots through a thousand generations, the computer can come up with a working model within a couple of hours and then it tells a 3D printer to make a real world version of this “pinnacle of evolution,” and all the robotics team needs to do is stick it together using servo motors.
That said though, we are already seeing the rise of 4D printing – printers that both print and can then automatically assemble the newly printed products, so one day there’s no reason why the robots can’t, or won’t, be able to design themselves, print themselves and assemble themselves. And moving one step further on, they might also soon be able to grow themselves in the lab, in the same way BAE are growing drones using new molecular assemblers.
Impressive as these systems might be, they’re still limited to work within the parameters we humans give them. In the future, Glette’s robots might figure out how to get around unexpected obstacles – and could even 3D or 4D print new body parts in the field in order to accomplish the task.
As expected though hundreds of prominent scientists and experts are already warning about the inherent danger of computers becoming smarter at an ever increasing rate, let alone evolving themselves – and then let alone, again, of letting them be used to evolve and manufacture new robots.
In many people’s eyes this new research has overtones of Skynet – after all while many people think of Skynet as just an “evil” computer program many people forget that it must have been able to design, evolve and manufacture the Terminator robots. After all, they didn’t just materialise out of thin air did did they?
The legendary physicist Stephen Hawking, for example, last year, went on record to warn people about the dangers of runaway AI.
“Once humans develop artificial intelligence it could take off on its own and redesign itself at an ever increasing rate. Humans, who are limited by slow biological evolution, couldn’t compete and would be superseded,” Professor Hawking warned. He also signed a letter in July condemning a potential AI arms race, and this year the UN will debate banning so called “Killer robots”.
“Once you’ve got artificial general intelligence, you’ve got a computer that’s got all the same smarts as you and I have, that means it can learn. So it’s learning 24/7, it doesn’t get tired,” says Keith Downing, a professor of artificial intelligence at the Norwegian University of Science and Technology.
“So once you get artificial general intelligence, you’re going to quickly accelerate into something they call artificial super intelligence. That’s where the computer is thousands of times smarter than us. So if they’re also smarter than us,” he says, “they could trick us and we could be in trouble.”
Glette though isn’t overly worried about a robotics takeover. That’s partly because he knows what it takes to get a simple, sausage-like robot to figure out to best way of moving across a carpeted floor.
“There’s a lot happening with self-driving cars and Google telling you when to take the next train home,” Glette says, “but we see that there is this complexity challenge, and I don’t think there is any real breakthrough at the moment. I think we will get smarter and smarter solutions, but no real human like intelligence any time soon.”
In the meantime though he and his team hope to make the robots, and its virtual evolution system, clever enough to operate in real life situations in the future though. One scenario would be sending robots into a nuclear disaster zone to solve any unexpected problems by adapting to their environment, perhaps even printing new tools or body parts in the process.
A self-learning robot could be handy for exploring distant planets, too. But the leap from that scenario to one of computers deciding they are smarter than us and acting on it is, hopefully, some time off – that said though many experts believe that that date is 2045. So maybe you should book your plane tickets to Mars now and avoid the forthcoming robo-geddon.