As technology evolves and society embraces artificial intelligence, we need to change the way we think about robots and autonomous machines.
When you think of robots, what do you see? Chances are your mind’s imagery is dominated by characters from television and movies.
From Robby the Robot in Forbidden Planet to Star Wars’ R2-D2 to The Terminator and Wall·E, movies show robots doing their thing completely detached from human control.
This fictitious world of futuristic creatures has almost certainly coloured how most people think about robots, and until now that hasn’t really mattered. But as autonomous machines become rapidly more common, we need to rethink our perceptions.
Such mechanical marvels have already infiltrated modern society, from robotised assembly lines to self-driving trains to highly automated warehouses with barely a human in sight. These machines never sleep, never take holidays and are becoming increasingly cost effective, which is why their numbers are growing each year.
The International Federation of Robotics (IFR) publication World Robotics 2015: Industrial Robots reports annual robot sales rose 29 per cent to a record 229,261 units in 2014.
The IFR looks mostly at industrial robotics, where industries such as car part manufacturing and electronics are deploying large numbers of autonomous machines. By 2018, the IFR projects annual sales to reach 1.3 million.
However, today’s robots do not look much like Robby the Robot or R2-D2. The “brain” in a self-driving train is just a collection of sensors and chips distributed through the engine unit. They also aren’t thinking for themselves; even referring to them as autonomous machines flatters them. They are either following a programmed course of action or are tethered to a human operator.
The perception gap
The distance between the robots in our heads and the ones in the real world is most obvious in a new book by David Mindell, Our Robots, Ourselves: Robotics and the Myths of Autonomy.
Mindell is a professor of history of engineering and manufacturing at the Massachusetts Institute of Technology and founded robot technology company Humantics Corporation. He has spent many years working with these devices.
Mindell’s central insight is that we overestimate the abilities of robots and other autonomous machines and, conversely, underestimate how much they might help us. We sense, process and analyse far better than they do, and even the most autonomous robots still depend on instructions from human beings.
For the moment, one of the great challenges of these devices is to make them less frustrating to humans and to make them work with people deeply and fluidly.
Mindell’s book sometimes describes robots as partners, but often they seem like tools. He explains the process of working with a robot called Jason while exploring the deep-sea floor. He thinks of Jason not as another human, but as a telescope.
CPA Q&A. Access a handpicked selection of resources each month and complete a short monthly assessment to earn CPD hours. Exclusively available to CPA Australia members.
"It was something you looked through,” he writes. “In fact, when it worked best, it was invisible.”
Tomorrow’s devices will be better, Mindell argues, but they will probably still remain tools. They will have different interfaces, but at times we will still need to direct them. While we might be happy for our car to drive us to work, he suggests we would still want a big red button with “stop” written on it to deal with the times when something goes wrong.
Indeed, automotive giant Daimler’s Highway Pilot System, which it expects will deliver the world’s first semi-automated road trucks by 2020, will retain a human co-pilot. A Daimler spokesperson says the company “expects a gradual process where the level of automation is increased step by step, if the conditions allow it”.
Robots in the real world
Semi-autonomous robots are commonplace in areas humans would find unpleasant or dangerous. For example, in the mining industry, BHP Billiton, Rio Tinto and Fortescue Metals Group are all investing in highly automated systems for minerals extraction, haulage and processing.
Their reasoning goes beyond safety. Rio Tinto chief executive Sam Walsh explained last October why his company has become a world leader in driverless, autonomous mining trucks. “The companies that thrive will be the most productive and efficient operators – and we are – and those who remain at the bottom of the cost curve – which we will,” he said.
Related: What will happen to Australia when the mining boom ends?
The most dangerous and costly field of autonomous machines is warfare. One of the most enduring visions of a movie robot is Arnold Schwarzenegger’s Terminator, a remorseless human-shaped killing machine that hunts its prey with unstoppable intent.
The closest we’ve got to that is the US Army’s more humble “robotic pack mule”, the legged squad support system or LS3 now under development.
From the standard military plane, engineers have developed remotely piloted drones, which have proven successful for surveillance missions and are increasingly used for small area strikes. However, as Mindell points out, drones are really just weapons with very long-range triggers, which are only ever pulled by humans.
According to Professor Toby Walsh, artificial intelligence expert from The University of New South Wales, the military is one area where research is aiming to give machines more autonomy. He believes it’s only a matter of time before decisions are left entirely to military machines, with niche programs already designed to achieve this.
“We will be putting autonomous systems into the battlefield and they will literally be making life or death decisions,” he says. “This is a really challenging moral threshold.”
Concerns about machines deciding who lives and who dies led to the launch of the Campaign to Stop Killer Robots international coalition in 2013, backed by Human Rights Watch and Amnesty International.
Building in learning
Nasir Ahsan, a Sydney University PhD graduate in philosophy, marine robotics and machine learning, has directed a fleet of remote submersibles for the Ocean Exploration Trust vessel Nautilus, and also worked with devices from aerial drones to mining robots. He is now director of Sydney-based start-up company Abyss Solutions, which is creating semi-autonomous vehicles for surveying waterways and dams.
Ahsan says these exploration vehicles are capable of making some decisions, but they must still be piloted via a tether. However, the goal is to deliver greater autonomy so they can make more decisions about the path they follow and the features they survey.
Abyss Solutions has developed low-cost geo-positioning technology using state-of-the-art machine learning algorithms, to help the devices know where they are.
“Without that, it would never be possible to build any kind of autonomy into the vehicle,” says Ahsan. “The vehicle needs to know where it is. When it knows, it can then compare its position to a pre-defined trajectory programmed into it and can correct its position to stay on that trajectory.”
Ultimately, Ahsan hopes this work might apply to projects such as Nautilus, letting it cut the tether to its remote submersibles.
“Tethers are good, but they are difficult to support and manage,” he says. “And whenever you have a tether, that means you have a ship behind it, and that can cost A$50,000 a day to operate.”
The push for true autonomy
A host of researchers continue to pursue true robotic autonomy or at least far greater autonomy than the world currently possesses. Undersea researchers such as Ahsan want to free submersibles from their tethers to make them easier and cheaper to run.
Space researchers want robots that can explore other worlds without waiting minutes or hours for signals to travel back and forth to Earth.
Military researchers worry that while their weapons rely on remote human operators, they are vulnerable to having their communications blocked.
To create this vision, researchers will need better sensors, improved battery life and, more importantly, superior processing power to crunch sensor data and make decisions.
As Professor Walsh points out, we are already happy to entrust our lives to machines, even with their limited capacities. Autonomous trains traverse London’s Docklands Light Railway and will run on the new Sydney Metro when it opens in 2019; and planes spend most of their time flying on autopilot.
The job question
A glance at a modern car manufacturing welding line with its peculiar absence of actual people is enough to make you wonder where robotics might be taking us. It’s easy to forget, however, that car lines have looked this way for more than two decades. Professor Walsh notes there are many manual tasks that won’t be automated any time soon.
Demand for airline pilots is soaring and surgery will still be done by surgeons for at least a decade – and in all likelihood, much longer.
“Replicating the sensation of the scalpel and whether it is cutting through fat or muscle is a really challenging engineering task,” he says.
Mindell makes similar observations. Railways were first imagined to replace horses, but in fact they offered a great deal more. Likewise, drones on Earth and robots on Mars don’t replicate humans; they do new things.
He dispels as a myth the idea of machines taking over human jobs.
“Rarely does automation simply ‘mechanise’ a human task,” he writes. “Rather, it tends to make the task more complex, often increasing the workload.”
The most important myth of all for Mindell is that machines will develop minds of their own. Granted, automation can take on some tasks previously performed by humans.
“The machine that operates entirely independent of human direction is useless,” he says.
Automation changes the type of human involvement, but does not take people out of the picture entirely.
The road ahead
If Google’s vision is realised, autonomous robots will be as common as cars on our roads. In fact, they will be one and the same.
Since 2012, Google has trialled self-driving cars in some US cities – such as Austin, Texas, and Mountain View, California – and major automobile companies are now jumping on board.
US research firm BI Intelligence estimates that about 10 million self-driving cars will be on the road globally by 2020.
“One thousand people die on the roads of Australia each year,” says Professor Toby Walsh, artificial intelligence expert from The University of New South Wales.
He says that number “will go towards zero” with autonomous vehicles, and he predicts that will drive rapid acceptance of driverless systems.
Calling on the robots
Deep sea research: The crushing pressure on the ocean floor makes it no place for humans and is also problematic for remote control systems.
Fire and rescue: Autonomous robots can withstand higher temperatures and aren’t choked by toxic smoke.
Human care: An ageing population sees increased investment in technology that can autonomously monitor and assist people in need 24 hours a day.
Mining: This industry uses autonomous robots for monotonous work and also for dangerous jobs, such as drilling. Machines are more cost-effective than fly-in, fly-out workers.
Space exploration: Most space probes already have a high level of autonomy, made necessary by the distance of space, limitations of light speed and the time it takes for control signals to reach them.
9 technology trends you need to know about