The business consulting firm McKinsey & Company just released a sobering new report entitled “Modeling the Impact of AI on the World Economy,” projecting the economic dislocations over the next few years that are likely to result from the spread of artificial intelligence (AI).
McKinsey & Company is not a pack of sci-fi dreamers. They’re paid by some of the richest and most influential businesses in the world, and they’re paid to be right. That doesn’t necessarily mean McKinsey & Company are always right, but it does mean that their estimates need to be taken seriously.
Two bottom lines: (1) the impact will be enormous; (2) the gap between the haves and the have-nots will grow faster than ever.
The first such gap the firm analyzes is the gap between countries. China and the US are far ahead of the rest of the world in AI development. Last year, 48 percent of the world’s AI startup investment occurred in China and 38 percent in the US. That left the rest of the planet scrambling for the remaining 14 percent. A few more advanced countries like Canada, South Korea, and Sweden may hold their own, but most of the world is going to get clobbered.
The second gap is between firms. Businesses that embrace AI are far more likely to be winners. “Front-runners tend to slowly concentrate the profit pool of their industry in a winner-takes-all phenomenon,” the report states. This may lead to the phenomenon of increasing concentration and the rise of ‘superstar’ firms.” What McKinsey calls the “long tail of laggards,” comprised of 60 to 70 percent of firms worldwide, are projected to lose around 23 percent of their cash flow compared to today. That kind of drop will destroy a great many businesses.
Humanists, though, are probably more interested in the third gap—the one between winners and losers in the workforce. McKinsey predicts that the total wages paid to workers “in the repetitive and low digital skills categories”—e.g., truck drivers, cooks—will decline by more than a third. By contrast, people who are skilled at developing and using AI tools will be riding the gravy train.
So we’ll just retrain all those truck drivers to be software engineers, right? Even to the limited extent the United States can do that, the report points out that the time these folks spend unemployed between jobs is likely to increase, depressing aggregate demand and stunting economic growth.
Though the overall tone of the report is positive, it does cite one chilling example: when Britain industrialized in the first half of the nineteenth century, output per worker grew dramatically while wages did not. As the profit share of national income increased, labor’s share of income declined. There are good reasons to suspect that pattern will be repeated as the AI revolution unfolds. In fact, the Financial Times recently reported the déjà vu insight that the share of company revenues going to profits is the highest in many years, while the proportion destined for employees’ pockets has shrunk.
Other considerations also generate concern. The McKinsey projections only go out for ten years—but AI growth isn’t going to stop then. It’s going to accelerate. It’s one thing to note, as the report does, that the horse industry was replaced by the auto industry over a century ago with a great net increase in employment. But there seems to be no upper limit on how far AI can go, on how many human jobs it can replace. Skilled workers ultimately die; software just keeps getting stronger.
I suspect that the report’s assumption that low-end, repetitive jobs are the ones at greatest risk may be misplaced. Some of the most intriguing recent AI advances involve replacing tasks done by high-income professionals, not just replacing grunt work. More and more diseases are being better diagnosed by AI than by humans, and a robot doctor has passed China’s national medical exam. In the legal field, an AI recently did a better job than a team of experienced lawyers in identifying issues in a contract. In education, AIs are already being used to mark student essays, and to fine-tune personalized learning. In finance, AI may replace up to 45 percent of portfolio management roles, and many banking jobs as well. AI writes many of the news articles you read (but not this one). AI can create beautiful visual art and captivating music. It seems the only thing AI cannot do is tell us what its limits are.
The real key, I believe, is to focus on who owns the artificial intelligence. At any point before I retired, I would have loved to have been replaced by a machine—if I owned the machine. If someone else owned it, not so good. Billionaire hedge fund founder Ray Dalio says artificial intelligence and automation are already causing such a dramatic wealth gap that “a national emergency should be declared.” Oxfam found that 82 percent of the growth in global wealth last year went to the top 1 percent of individuals ranked by wealth.
All AI growth has to be financed somehow—it doesn’t happen by itself. Traditional forms of capital finance reinforce ownership in the hands of the already wealthy. But as I’ve written before, well-proven alternatives exist, that finance growth in ways that build ownership into ordinary people.
In the UK, the Labour Party last week announced that if it comes to power, it will require firms of more than 250 employees to create “ownership funds” to give their workers stakes in the business. I don’t know the details of how this will work, and the scheme may be flawed. (Carrots tend to work better than sticks.) But it is at least pointed in the right direction—the only direction that offers long term hope. AI and the robots it controls are going to get stronger and stronger. There isn’t a single thing we can do to stop that, and we shouldn’t want to stop anything with so much potential for improving the human condition. But if that AI is owned by the same tiny fraction of the population that owns most productive capital today (or, if McKinsey is right, an even tinier fraction), then humanity will have converted what ought to be the greatest boon in history to its greatest scourge.