Don’t fear an AI-induced jobs apocalypse just yet

From The Economist:

“I think we might exceed a one-to-one ratio of humanoid robots to humans,” Elon Musk declared on March 1st. Coming from the self-styled technoking of Tesla, it was not so much a prediction as a promise. Mr Musk’s car company is developing one such artificially intelligent automaton, codenamed Optimus, for use at home and in the factory. His remarks, made during Tesla’s investor day, were accompanied by a video of Optimus walking around apparently unassisted.

Given that Mr Musk did not elaborate how—or when—you get from a promotional clip to an army of more than 8bn robots, this might all smack of science-fiction. But he has waded into a very real debate about the future of work. For certain forms of ai-enabled automation are fast becoming science fact. Since November Chatgpt, an ai conversationalist, has dazzled users with its passable impression of a human interlocutor. Other “generative” ais have been conjuring up similarly human-like texts, images and sounds by analysing reams of data on the internet. Last month the boss of ibm, a computing giant, forecast that ai will do away with much white-collar clerical work. On March 6th Microsoft announced the launch of a suite of ai “co-pilots” for workers in jobs ranging from sales and marketing to supply-chain management. Excitable observers murmur about a looming job apocalypse.

Fears over the job-displacing effects of technology are, of course, nothing new. In early 19th-century Britain, Luddites burned factory machines. The term “automation” first rose to prominence as the adoption of wartime innovations in mechanisation sparked a wave of panic over mass joblessness in the 1950s (see chart 1). In 1978 James Callaghan, Britain’s prime minister, greeted the breakthrough technology of his era—the microprocessor—with a government inquiry into its job-killing potential. Ten years ago Carl Frey and Michael Osborne of Oxford University published a blockbuster paper, since cited over 5,000 times, claiming that 47% of the tasks American workers perform could be automated away “over the next decade or two”. Now even the techno-optimistic Mr Musk wonders what it would mean for robots to outnumber humans: “It’s not even clear what an economy is at that point.”

. . . .

The immediate problem for advanced economies is, then, not too much automation but too little. It is exacerbated by the fact that, for large businesses, automating has been hard to get right in practice. It is likely to prove no easier with the latest buzzy ais.

Rage for the machine

Mechanical arms on a factory floor performing repetitive tasks such as welding, drilling or moving an object have been around for decades. Robot usage historically centred on the car industry, whose heavy parts and large batches with limited variety are ideally suited to the machines. The electronics industry, with its need for precise but repetitive movements, was also an early adopter.

More recently the list of industries embracing robots has widened, observes Jeff Burnstein, president of the Association for Advancing Automation, an American industry group. Advances in computer vision have made robots more dexterous, notes Sami Atiya, who runs the robotics business of abb, a Swiss industrial firm. Lightweight “collaborative robots” now work side-by-side with humans rather than in cages, and autonomous vehicles ferry objects from one spot to another in factories and warehouses.

At the same time, robot prices have tumbled. The average price of an industrial robot fell from $69,000 in 2005 to $27,000 in 2017, according to Ark Invest, an asset manager. Last December abb opened a 67,000-square-metre “mega factory‘‘ in Shanghai where robots make other robots. Installation costs have come down, too, with newer “no code” systems requiring no programming expertise, notes Susanne Bieller, general secretary of the International Federation of Robotics (ifr), another industry group.

As a result of better technology and lower prices, the global stock of industrial robots grew from 1m in 2011 to nearly 3.5m in 2021 (see chart 2). Sales at Fanuc, a big Japanese robot-maker, rose by 17% last quarter, year on year; those of Keyence, a Japanese firm that acts as an automation consultant to the world’s factories, shot up by 24%. Though down from the frothy peaks of 2021, when bosses sought alternatives to human workforces incapacitated by covid-19, robot-makers’ share prices remain a fifth higher than before the pandemic.

. . . .

For all that growth, however, absolute levels of adoption remain low, especially in the West. According to the ifr, even South Korean firms, by far the world’s keenest robot-adopters, employ ten manufacturing workers for every industrial robot—a long way from Mr Musk’s vision. In America, China, Europe and Japan the figure is 25-40 to one. The $25bn that, according to consultants at bcg, the world spent on industrial robots in 2020 was less than 1% of global capital expenditure (excluding the energy and mining sectors). People spent more on sex toys.

. . . .

Businesses are now beginning to embrace generative ai, too. But as with robots and process automation, bedding in the new technology will not happen overnight. Allen & Overy, a law firm that in February launched a virtual legal assistant with Chatgpt-like powers, requires its lawyers to cross-check everything the bot spits out. cnet, a tech-news site, starting in November quietly published 73 articles written by a bot, first to the consternation and then the delight of journalists, after the articles were found to be riddled with errors.

Link to the rest at The Economist

18 thoughts on “Don’t fear an AI-induced jobs apocalypse just yet”

  1. The execution is always the problem. The more complex the task to be taught to a robot, the more it has to ‘know’ and be able to speedily compute in real time.

    What a 15-year-old learning to drive does in spite of raging hormones, a half-developed brain, and pimples, needs to be codified, preferably perfectly, with an expert database and every eventuality pre-programmed. Most people have NO idea of the complexity this involves. Helping a fragile old lady take a bath – a good use for a robot – could accidentally kill her.

    It also makes me realize what absolutely wonderful machines humans are.

  2. Musk wasn’t really talking global numbers. More like first world numbers (300-500M) or total global cars (1.4B) and that took about a century. Also his aspirational number is more like 1M (Marsville! 😉 ) So he’s only looking forward 30-50 years.

    More importantly robot adoption is going to accelerate for the next 20 years because of the labor shortage already upon us, the reshoring of manufacturing (the new factories will be heavilly automated, paying good salaries, not FOXCONN sweatshops), and most importantly, robot *leasing* which is already a growth business. The latter model will be most prevalent in agriculture outside the row crops (fruits, vegetables, orchards, etc). Startups are emerging that buy robots and rent them for temporary use in places where there isn’t enough work to justify a purchase. (Think fruit picking.)

    Here’s some of the agriculture robot builders:

    None of these robots will be taking anybody’s job, just making them more productive and keeping them safe. While filling a need without exploiting the uninvited.

    Hard to see a downside: more productivity, no social disruption.

    • Based on my conversation with local farmers for apples and nuts, agricultural robots need huge improvements in usability (right now the orchard has to be designed for the robot, not the other way round) and cost (currently unaffordable for anyone small to moderate in size). Yes, I know it’s being worked on.

      Second, the OP isn’t really knowledgeable about robots, it shows in calling Keyence an “automation consultant”. Also, I think the majority of automation is, and will continue to be, specialized automated machines (for example, pick and place machines for PCB assembly, wire bonders for semiconductor, etc) rather than general purpose robots.

      And while there is a place for general purpose no code robotics and cobots, both are currently over hyped. In summary, yes, automation of various sorts has its place, but a big challenge is using the right kind in the right place. Just blindly automating is a huge waste of money.

      • “…the majority of automation is, and will continue to be, specialized automated machines (for example, pick and place machines for PCB assembly, wire bonders for semiconductor, etc) rather than general purpose robots.”

        Why overly complicate things?
        There is a place for ML in today’s industrial/agrarian robots in discrimination functions, but general purpose flexibility doesn’t justify the cost. For now.

        The robots Musk is developing aren’t for industrial use. For one thing, they’re too slow and fragile by design. They’re meant to be relatively cheap. $20,000 or less.

        By now Musk has a demonstrated M.O: he sees where things are headed and positions himself a step or two ahead of the endpoint he envisions. And he makes his money helping things along. Electric cars, reusable space boosters, satellite internet…humanoid robots.

        For a better read at where his humanoid robots fit, look to Japan and Korea and their eldercare robots. Hollywood has done at least two movies along those lines, most notably ROBOT & FRANK.

        General purpose, light duty robots would be suitable for necessary jobs that don’t add enough value to command a living wage. In offices, hospitals, and homes. Schools maybe but they’d have to be heavily armored. Space stations and the moon, too. Mars, maybe. In Singapore they have a humanoid robot receptionist, named Nadine. Walmart greeter might also work. The idea is they’ll go where humans go, do what humans could do can’t or is beneath them.

        It may sound insensitive to some ears, but the niche they’ll be best fitted for is “house elf”. or their historical precursor.

        • The requirements for the Apple picking robot was quite extreme (and costly); in the future, it might be more relaxed.

          Shaking a tree is totally different level; heck, people have been doing that (shaking plum trees is mentioned in On the Banks of Plum Creek). Also, I believe for some trees (walnuts, pecans) it’s best to wait until the nuts naturally fall – I’ll ask about that.

  3. Looks like the robots are winning. For several years, they have been mopping the floors of big box stores. A guy with a mop used to do that.

    The self-checkouts are easily a majority in the local WalMart. The last one I saw was a unit they had stuck on what had been a human station. Those jobs used to be human.

    At Sam’s Club the iPhone is taking the self-checkout jobs away from the robots. I go through the store, scanning each item with my iPhone. Then I walk out and a human at the door does a quick audit scan.

    Amazon warehouses are full of robots running around the floor doing what humans used to do.

    People fret about robots writing books and legal briefs, while the robots are content to start at the bottom and work their way up.

    • I’ve seen all those.

      At the local Walmart they station one employee to monitor six auto checkouts, mostly to help scan produce.

      Also: before the pandemic no supermarket had online shopping, now they all do, even the local independents. Pickup or delivery. Home depot, too.
      Drug stores take prescriptions straight from the doctor via email and deliver.

      Processes are streamlined all over.
      And they’re not putting anybody out of work because those jobs aren’t in high demand anyway.
      The economy is changing.

      • Um, Felix? Whole Foods. Went online for home delivery not long after their acquisition in 2017. (Admittedly, Amazon had the infrastructure and expertise largely developed already.)

        I was getting one auto-fill medication delivered to my door back in 2009.

        Now, I admit that we still pretty much only have Amazon orders delivered. We are very picky about our produce and meats. It’s not free, but Prime membership is much less than what we’d rack up in normal shipping charges.

        Some things still aren’t economic, too. We’re going across town (about 30 miles round trip) next week to pick up a load of flooring. A bit less than $500 worth, and ship to store is free – but they want $250 to ship an order to the house. Nope, nope, nope…

      • And they’re not putting anybody out of work because those jobs aren’t in high demand anyway.

        Of course they are putting people out of work. In your example five cashiers were put out of work. No demand by employers = putting people out of work. Changing demand is a significant factor in changing an economy. Demand doesn’t fall because the economy changes. The fall in demand is the change.

        • You’re assuming those five workers exist and aren’t working someplace else for a higher salary than they would’ve had as a Walmart cashier. And that one person replacing the five gets a higher salary without the checkouts.

          Also, as a local factor: after the big hurricane and the quakes, PR lost over 300,000 people, all working age, who weren’t waiting for things to return to the ststus quo ante. Most weng to Florida. The advantages of bilinguism and the US labor shortage. Remember, the US is over 4.3M workers short.

          TL:DR – As boomers retire their replacements in the workforce come from millenials and zoomers. Remember, boomers are the biggest generation ever, genx is a third smaller than the boomers (50M vs 75M). Millenials, as the boomer’s kids, are a bigger cohort than genX, but have lower labor participation than their parents. Zoomers aren’t done arriving but are also low labor force participants.

          The shortage won’t be “solved” until the millennial kids (if any) hit the work force around 2040. The “if” being the declining fertility rate and life choices of the millenials.

          The US has the best demogrsphics of the develooed world but it still falls short of replacement demographics. It has to make up the difference via brain drain and automation, hardware and software.

          • I assume nothing. The positions those workers held no longer exist. They don’t exist if the workers are on unemployment, and they don’t exist if they are now editors at Random House. They were put out of work.

            The subsequent fate of those workers is a second question. In a healthy and growing economy, they all went over to Random House, took jobs making solar panels, or became coders. With a few setbacks, this has been happening in the West for the last 500 years. It’s always a crisis endangering the very existence of the planet.

            Today it’s the result of computers and AI. In the past it happened because of the steam engine. I’m old enough to remember when every elevator in the country had a guy sitting on a stool moving the car from one floor to another and pulling the gate open. How many people under the age of fifty have ever dialed the operator?

Comments are closed.