‘Humanoid’ robot wave signals change on the production line

From steam engines to conveyor-belt assembly lines and robots on the factory floor, the manufacturing industry has long been a pioneer of new technologies. Artificial intelligence now looks set to become the next and, perhaps, biggest leap forward. But what will it mean for jobs over the next decade?

Use cases: controlling plants, recommending equipment fixes, designing products, assembling parts

Manufacturing is already highly automated, with sensors, software and computer networks monitoring the output, data, pressure and temperature of factory machines and industrial processes. Such connectivity has become essential on sites that are sometimes square miles wide.

“In a refinery or petrochemical plant, there can be thousands — if not tens of thousands — of instruments, equipment and valves [needed] to, for instance, manage 250,000 to 500,000 barrels of oil per day and process that into gasoline,” points out Jason Urso, chief technology officer in the software division of Honeywell, a US industrial conglomerate.

Within 10 years, more than 80 per cent of manufacturing facilities could be using AI to help run these “control systems” and fix problems with them, he forecasts. If, for example, a machine emits an unusual sound, a factory worker can ask the AI software to analyse that sound, summarise the problems associated with it and recommend remedial action, Urso says.

Some manufacturers already invest in this type of AI. United States Steel Corporation, for example, has said it will use generative AI software from Google to guide its workers through truck repairs and ordering parts.

AI is also playing a bigger role in product design. For example, AI-powered software can help automotive engineers make multiple three-dimensional vehicle designs in minutes instead of days, says Stephen Hooper, vice-president of software development, design and manufacturing at US software supplier Autodesk.

“You [can] build 3D [computer designs] of styling for new vehicles in a fraction of the [current] time,” he notes. “You can control characteristics like the wheelbase, the vehicle type . . . and [the AI] will derive hundreds, if not thousands, of alternatives”.

Hyundai has used Autodesk software to help design parts for a prototype car, the wheels of which can transform into four legs to walk and climb — making it a potential rescue vehicle. 

In factories, robots have long been used to assemble parts but the next generation will be AI-powered “humanoid” robots, working alongside humans. These will have sufficient dexterity and learning capabilities to handle processes such as picking and sorting items into categories, experts say.

Early versions could be operating within the next five years, predicts Geordie Rose, co-founder and chief executive of Canadian start-up, Sanctuary AI, which is aiming to create the world’s first robots with “humanlike intelligence”. Its latest Phoenix model is 5ft 7in tall, weighs 70kg, and can walk at up to 5km/h. It is operated by humans but, Rose forecasts, will eventually mimic human memory, sight, sound and touch.

Demand for such humanoid manufacturing robots is going to be “significant”, according to a recent Goldman Sachs research note — particularly in electric vehicle assembly.

“The central premise of this approach is that you can build a machine that’s humanlike in the way that it understands the world and acts on it,” explains Rose. However, building a machine that can react like a human “is obviously a lot harder than [building] a machine that . . . can do a couple of things that people can do”.

Sanctuary’s robot can already sort mechanical parts as fast as a human, but even Rose acknowledges further improvement is needed. “The question is, how long it will take [for our robots] to go from the lab to being on the factory floor,” he says. “And that’s a very difficult question to answer.”

Eventually, robots equipped with artificial general intelligence (AGI) — the same level of intelligence as a human — will be capable of designing and making things, Rose predicts. “You could ask a sufficiently powerful AGI [robot] to design a new battery and then manufacture it.”

Jobs lost: production-line workers, quality-control assessors, and machine operators

Adding AI to manufacturing robots — which do not demand pay rises or go on strike — has the potential to make millions of traditional manufacturing roles redundant.

Pascual Restrepo, associate professor at Boston University and an expert on industrial robots, points out that non-AI robots have already replaced between 6mn and 9mn manufacturing jobs, globally, since the 1980s. About 500,000 of these were in the US alone.

Now, most experts predict AI will lead to more jobs cuts in manufacturing. When technology leaders around the world were surveyed last year by recruitment company Nash Squared, they estimated that 14 per cent of jobs in manufacturing and automotive industries will be lost due to “automation” technologies, including AI, over the following five years.

Production-line workers, quality-control assessors and machine operators seem most at risk of being replaced by AI. Gabriele Eder, head of manufacturing, industrial and automotive at Google Cloud, Germany, suggests that, in these areas, AI-powered machines and equipment can “often operate with greater precision and consistency than human operators” — requiring less human intervention in manufacturing processes.

“Our members are very much worrying [about AI taking their jobs],” says Kan Matsuzaki, assistant general secretary at IndustriALL, an international union representing more than 50mn workers in mining, energy and manufacturing. He adds, however, that his members acknowledge the possibility of AI bringing benefits, such as improving safety in manufacturing.

Training manufacturing workers in applying alongside AI could help them adapt and minimise job losses, but opportunities may be limited. “When you reach like 55 years old . . . can those workers be retrained to become [an] AI machine . . . specialist, for example?” says Matsuzaki “[It] is very difficult to do that.”

New jobs: machine monitors, robot programmers, digital champions, forensic AI scientists

However, some experts predict that AI will create more new jobs in manufacturing than it eliminates. They note that manufacturing companies are keen to hire, rather than fire, workers — but are hampered by a global shortage of people with manufacturing skills.

New AI-related manufacturing jobs will include running AI machines, monitoring their performance, programming robots, and working in “cross disciplinary teams” with equal expertise in data science and manufacturing, experts predict. At the same time, old jobs will change and become more tech-focused, rather than being replaced by AI, says Marie El Hoyek, an expert in AI and industrials at consultants McKinsey.

“Some of the manufacturing roles will need to evolve,” she says. “I imagine [in the future] you would need digital champions who are core manufacturing people but know how to translate their needs and their work into digital language to the digital team and say ‘this is what I need you to resolve’.”

AI will increase demand for “forensic AI scientists”, typically from a technology background, who analyse the performance of AI systems, says Cedrik Neike, chief executive of digital industries at German technology company Siemens. “[We] need to have experts which [understand] where things go wrong to fine tune them,” he says.

How widely these AI systems are deployed remains open for negotiation, though. “The ultimate question is, who will benefit from this AI?” says Matsuzaki. “If you introduce AI and automation robot[s] in manufacturing workplaces . . . you can reduce your number of workers, which means the productivity will gain [and] profit will gain . . . But there’s nothing for the workers.”