The rise of the machines, or the rise of the humans? Or both?
Writing about robotics and artificial intelligence (AI) in 2017 must be similar to writing about the internet in the 1980s: “What will we do with all our productivity savings and free time once this new technology creates a fantastic new utopia?”
It is easy to be cynical because the predictions out there are pretty wild. One of engineering’s biggest heroes of the moment, Elon Musk, leader of robotic-forward companies SpaceX and Tesla, predicts AI as humanity’s “greatest existential threat”. Founder and chairman of the world’s largest retail platform Alibaba (China) Jack Ma recently said the world’s best chief executive could soon be a robot; “It remembers better than you, it counts faster than you, and it won’t be angry with competitors.”
What technology can do now is already amazing. AI sorts our vast amounts of data into specific, usable quantities at lightning speed using the cloud and supercomputers; Google finds your search results, building information modelling (BIM) informs design, Amazon predicts your next purchase.
Robots meanwhile take care of our repetitive and programmable tasks, often in extreme or dangerous environments: probes in deep space, or insect- and worm-like robots cleaning up nuclear meltdowns or fixing machinery in deep tunnels.
But increasingly, the line between these two forms of automation, AI and robotics, is being erased. It is the marrying of these technologies, the “brain” of AI and the “body” of robotics, that has many worried or excited about humanity’s future.
But author of Rise of the Humans: How to outsmart the digital deluge Dave Coplin is firmly in the latter camp. As UK Microsoft’s principal tech evangelist, he is a staunch defender of the tech industry.
But he admits the first mistake was using the name “Artificial Intelligence”.“Because what we’re looking at today is neither artificial nor intelligent. When we talk about AI we’re really talking about another form of automation.”
“Machine learning”, “DeepMind”, “Neural networks” – whatever you call it, what people like Musk and Ma are investing in is the mimicking of humans’ basic cognitive processes. Coplin explains: “Anything that takes you less than a second to think of the answer, can actually be easily automated today. Typically that’s anything that follows a pattern.”
Istock 185323250 crop3
But because entire professions are built on such patterns, AI is now taking away certain tasks from workers. Call centres, supermarket checkouts and taxis are the first to see changes. Today even tasks in the upper levels of banking, accountancy and the legal system are changing rapidly; given a standardised process, historical data, and connecting variables, a computer can now spit out a likely probability when asked complex questions.
“There’s a number of tasks in civil engineering that will benefit from this kind of automation, from design to maintenance,” says Coplin. “If we can feed the algorithms with the right data, the robots will be able to do it for us.”
A widely cited 2013 University of Oxford study claims robots could displace 35% of jobs today by 2035. Many in the tech world disregard this scenario altogether, saying that since the 1960s nearly all warnings about automation and high unemployment have proved incorrect.
Istock 185323250 1 cropped
But this time around could be different, both in the speed and depth of changes coming. Pro-Robotics body the International Federation of Robotics (IFR) says that in 2015 industrial robot sales increased by 15% to 253,748 units. The IFR expects 2.9M industrial robots to be deployed worldwide by 2019, which is 1M more than in 2015. Annual worldwide AI revenue could grow from $643.7M (£498.9M) in 2016 to £30bn by 2025, according to market analysts Research & Markets.
Coplin argues that drastic changes will come, but through augmenting humans, rather than replacing them. “If the algorithms and robots can do 40% of the tasks I do today, that doesn’t mean I sit on my backside for the rest of the day, it means I go and do the things that the algorithms and robots can’t do.”
Coplin says three major areas of humanity will be unachievable by AI or robotics for the foreseeable future: emotions, creativity and accountability.
We can get machines to imitate Rembrandt’s eye or replicate Mozart’s style, but they cannot in themselves create new things
Dave Coplin, Microsoft
“So at the moment I’ve got algorithms that can interpret your emotions; I can take a quick look at your face and tell if you’re happy or sad and what that might mean. But the algorithm doesn’t inherently know what ‘being sad’ means. It can’t contextualise it: I can see it, but there’s not much I can do with it.
“The second skill is creativity: yes we can get machines to imitate Rembrandt’s eye or replicate Mozart’s style, but they cannot in themselves create new things. So we have built a computer that can beat the best human in the world at chess – but that computer could not do something as simple as hum a tune while playing; it couldn’t invent a different game.
“The third and final area is what I term ‘accountability’. A world of AI, in particular, is a world of probability. When you’re given an answer by an algorithm, you’re not given a certainty, you’re given the most likely answer in the least amount of time. Today that’s okay because the decisions we’re making based on algorithms are: ‘Which restaurant should I eat at?’; ‘Which mountain bike should I buy?’ In the next few years we’ll be making much deeper decisions.
Making deep decisions
Citizens today are already taking on these deep decisions: finding a lifelong partners on Tinder, getting political news via Facebook, Google DeepMind’s new role in the NHS.
Coplin gives his 11-year-old son as an example, saying his son is learning nothing in school to prepare him for this brave new world. “What we have to do is equip students with the right sort of judgement skills and accountability to be able to know what action to take, based on that probability.” And what if the son grows up to be an engineer?
“If you bring that back to civil engineering, it’s going to be designing projects that are going to be given all this input and they’re going to need to be creative and know how to engage with other humans, and finally, they’ve got to know that when they’re given a probability by one of the algorithms, they do their human due diligence to make sure it’s going in the right direction.”
Companies need to start a “data culture”, not everyone becoming data scientists, but understanding at a basic level how data works
The right direction, according to powerful tech companies is for more robotics, more AI. Companies like Microsoft have built their fortunes using extremely talented computer programmers, incredible masses of data and supercomputers to crunch data. But many of these companies have taken on risk regarding tax arrangements, fake news and antitrust issues.
“What I fear is that companies will make a very substantial saving, by automating their business, and the worst thing we can do is to bank that saving,” says Coplin. “In the broader economic context, and political context we’re in, it’s going to be very hard for anyone not to want to bank all those savings.
World’s knowlege freely available
“But if you look at the Internet as an example: yes a lot of people have got very rich, but we’re in a world, regardless of developed or developing, where everyone has access to the world’s knowledge. This has fundamentally changed our society and AI will do something quite different again.”
So what on robotic earth is a lowly civils firm, engineer or labourer supposed to do to prepare for the future?
“There are some things on a personal level, thinking about a world with an ambiguous future, and how well equipped you are for change,” says Coplin.
Explore the AI world
“At an organisational level it is really a request for people to explore this world, understand what it might mean, and once explored you need to expand. AI is relatively simple to experiment with – the algorithms will have been written by the big tech companies, they’re out there. So start to experiment and see what it could do. Build up the insights.
“Companies need to start a ‘data culture’, not everyone becoming data scientists, but understanding at a basic level how data works, so you know the right questions to ask. That helps an organisation to adapt as quickly as possible.”
Also attempting this change in culture is the UK government. Following on from reports on robotics in medicine and manufacturing a new UK Robotics and Autonomous Systems Network’s White Paper is being written on “Robotics for Resilient Infrastructure” for the Department for Business, Skills and Industrial Strategy.
Infrastructure Plan needs robotics
Co-author of the report University of Leeds’ Raul Fuentes argues the government’s National Infrastructure Delivery Plan, which commits £100bn by 2020-21, will need robotics. The plan will set the government’s objectives: boost productivity, raise international competitiveness, support growth and create jobs.
But is it realistic to expect robots will create jobs? “We provide data that show that, yes, jobs disappear, but they reappear somewhere else. And this has been proven time and time again throughout history. Some people might argue this is different with robotics and automation but we believe it can help as well,” says Fuentes.
Further, Fuentes, a trained civil engineer, says the biggest threat is not going to be “job losses” but skills gaps. “Most importantly I believe is what’s going to happen to graduate engineers when they come into their jobs is that you start to design in BIM, designing a portal frame, simple things, and those are the things that can be designed by a combination of AI and computer methods.
I don’t see any robots taking over the world anytime soon. They’re tools to assist humans
Barry Lennox, Manchester University
“Experienced engineers will always be needed to check the designs. But they’re only able to do it because they’ve been able to learn, building up from simple tasks. And that disappears. So how are we going to produce engineers in the future? I think that’s the big problem.”
Fuentes’ advice is not to learn to code or get a robotics degree, but rather pay attention to your own company, they might already be on to it.
“What I can tell you for sure, this is happening. Most companies have a robotics division, but most are employing experts from outside. But in the not very distant future they all will have somebody dedicated to that.”
Taking over city maintenance tasks
Fuentes has laid down one gauntlet, staking a claim on robotics performing all city maintenance tasks in Leeds – roadworks, streetlight replacement, pipe repairs – by 2050. The long-term goal reflects the size of the challenge. “Think of a robot: it has a brain, which is the electronics board or computer. Then there’s perception, which is directly linked to that.
“But then there is all the dexterity, manipulation, locomotion and autonomy – that’s going to be the obstacle,” says Fuentes.
Elsewhere at the forefront of academia, Manchester University professor of robotic systems Barry Lennox says plenty is happening in civils and robots. And things are happening increasingly quickly. “You can see the kind of things that students can do with electronic components that cost £5, they can do incredible things for very little money, and they’re getting better day by day.”
He leads a consortium investigating how to decommission nuclear sites robotically. These automatons are operated by people, not with AI. The robots are designed for drudgery, servitude and early death. Lennox does not believe in a robot apocalypse. “I don’t see any robots taking over the world anytime soon. They’re tools to assist humans, keeping us away from dangerous environments.”
“I think you [civil engineers] need to understand what robotics can do, are likely to be able to do. We’re also developing a robot that can inspect and repair pipes. In five to 10 years you could see robots in pipes going along, seeing where the leaks are, blockages.”
Another growth industry in coming decades will be 3D concrete printing. As Skanska innovation manager David Lewis says: “The dream is that construction workers on site with their ipad or tablet, they have the latest revision of the BIM model and they can press print,” he says.
“We’re not that far away from that.”
BIM and automation
Indeed, BIM has sped up automation advances, with an increased uptake of condition- and preventative-based maintenance in road and rail maintenance. “Smart roads” could soon be predicting hazards, delays and maintenance requirements, while rail experts are already using sensors to make train maintenance more forensic.
Hong Kong’s Mass Transit Railway Corporation (MTR) has been on AI for years already, planning, scheduling and optimising its engineering work. And deep under Hong Kong some serious robotics are in action, with sensors and snake-like machines examining and cleaning TBM cutter heads. Highly ambitious and high-risk, in poor ground and under high pressure – a robotic solution is necessary.
But investing in new territory can be difficult. “Five years ago we tried to build some robots. It was a big mistake,” explains Bouygues technical director Christophe Portenseigne. “In factories robots stay in one place. In our industry, they need to move,” he says.
This is a simple observation about a problem which could have been solved easily through earlier collaboration. But commercial realities, intellectual property rights and privacy concerns – rightly or wrongly – often stand in the way of progress.
Coplin says if we want a future with less robot failures it will necessarily mean less closed ecosystems and more shared information.
And as the tech companies’ arms race continues, companies will come for more data to fuel their AI algorithms and further personalise their services.
Of course we’re cynical about this, and we’re right to be cynical. But that doesn’t mean we should try to get the gift in the right way.”