Anthropic raised eyebrows recently in its extremely optimistic predictions of how AI will boost economic growth. Obviously, since it is preparing its initial public offering of its stock, the firm has an interest in convincing potential investors that the AI boom is endless. But the AI boom has many proponents, both investors and politicians who frequently hype the economic gains that AI will generate, to support even higher stock prices and more data centers in their districts.
Productivity
The basic and oft-repeated argument is that AI will boost productivity, like the industrial and computer revolutions before it, so AI-powered economic growth will solve all problems, including high debts loads, displaced workers, life enhancing innovations, etc. But productivity is not so simple.
In economics, productivity is actually a measure of worker productivity, so the formula for the rate of economic growth is productivity times the growth in the number of working people. If productivity rises by 15% while unemployment rises by 15%, then there’s no increase in economic growth.
Anthropic predicts in one scenario that over 15% of office workers may need to transition to more manual jobs. If they take those jobs at lower pay, then their productivity drops, which would also lower economic growth. For some reason, AI proponents frequently ignore the negative effects of worker disruption when estimating future economic growth.
Competition
Anthropic also bases its economic growth estimates on AI shortening the business cycle. Instead of entrepreneurs toiling in garages for years to develop new products and technologies, AI will do it almost instantly. However, a faster business cycle is not so simple economically.
Part of the current business cycle is that the inventor gains a competitive advantage allowing her to charge a premium price. Over time, competitors offer similar products and prices fall, eventually reducing many goods and services to commodities, where the price is only slightly above the actual cost to produce.
AI will reduce the competitive business cycle as well. Your company may lower its costs by using AI, but AI will also enable new competitors to reproduce similar products or services. Due to competition, prices will drop. If volume increases 15% but prices drop 15%, there is little economic growth.
I use AI for free every day to answer questions, and I expect it will be ubiquitous. I sometimes catch AI errors, and it corrects itself instantly. AI firms may believe that they will be able to control who has access to their AI results, so they may believe that they can continue to charge a premium price for those results. But the self-learning, self-replicating nature of AI models and agents suggests that AI results will be available virtually everywhere for free or for extremely low prices.
If there is a business opportunity, AI will identify the most efficient way to exploit it and make that result available instantly. Other countries and their companies also have AI, so we should expect both a proliferation of competition and a rapid destruction of competitive advantage. That will certainly boost innovation and will create new jobs, but it will also cut profits to commodity levels almost immediately. There is no certainty that the most expensive AI companies will win in the marketplace, and logically, the lowest cost producer—whether in China or at some new cheap company—should be expected to win in a commoditized industry. For some reason, AI proponents frequently ignore the negative effects of faster competition when estimating future economic growth.
History
Every technological revolution has caused some unemployment and disruption.
The industrial revolution drew workers into city factories. Many of these were women and children who previously worked on farms. As farm equipment mechanized, farm productivity rose too. Industrialization caused economies to boom, as farm animals and horses were replaced by machines, and many previously low paid rural workers were able to earn more in company towns, despite tough conditions and exploitation. Mills mass produced cheap clothing to replace handmade clothes. In many cases, the economic and social benefits of industrializing outweighed the costs.
In the case of mining, such as the Michigan copper mine above, the introduction of the steam drill doubled productivity, as one worker could now do the job of two. But it also caused mining unemployment, lower wages, strikes and lower prices. When prices dropped low enough, the mine closed, replaced by lower cost production elsewhere. Not everyone benefited from industrialization, as the old folk song John Henry reminds us.
The computer revolution also changed the nature of work. My first job still had a secretarial pool in the middle of the office surrounded by managers’ offices. Computers replaced the secretaries, and the managers needed to learn computers or retire. The factory floor was also automated. By replacing routine work, demand for skilled workers increased as did economic inequality.
We are still living in this revolution, as the Internet continues to concentrate productivity gains and wealth production into fewer hands, often controlled by a handful of leading companies and owned by a handful of billionaires.
The AI Revolution
Technology revolutions do not all follow the same script. Farm animals and unpaid women working at home lost their old jobs in the industrial revolution, but life improved for most people. Switchboard operators and assembly line workers lost their jobs due to computerized automation, but customer service reps and high skilled jobs gained.
AI will target the knowledge industry, but not just as a continuation of the computer revolution. Most of the highest skilled and highest paying jobs will be lost to AI. While some careers require a real human face, touch or trust, a surprising number of jobs can be completed by AI.
I recently checked into a hotel where the clerk was on a screen, so the clerk could be replaced by AI. Tech billionaires frequently hype robotics, and I’ve taken Waymo rides several times. Robotics is advancing at an unprecedented pace, and robots are now making robots. The future is coming faster than you can say “danger, danger Will Robinson”.
Conclusion
The assumption that AI will cause our economy to double every few years with little downside is wrong. The assumption that all AI companies will make massive profits forever is wrong. When the auto industry began the US had 90 companies within a few years. Quickly that reduced to three, and one almost went bankrupt due to foreign competition.
The assumption that AI data centers will not quickly be made obsolete by the rapidly evolving technology is also flawed. AI will undoubtedly find more efficient ways to run in the near future than today’s water-hogging fossil-fuel polluting inefficient data centers.
Downsides are a given in every revolution. I’m sure those early industrialists would have been horrified to see the war machines of our world wars. AI is already being used in warfare, speeding up lethality. AI is being used by criminals, in theft and fraud. And I’m sure AI will produce future problems that mere humans can’t imagine.
Selling AI’s benefits while ignoring the downsides is an industry-wide fraud. Allowing corrupt politicians to help perpetrate this fraud on the American people, instead of preparing us for massive job disruptions and other negative consequences, is unconscionable. Doing so while ignoring the climate crisis may be suicidal.
Human stupidity (HS) is more apparent than AI in the one-sided, baseless assumption boosterism so common today. Instead of allowing a handful of AI egoists and egotists to misdirect us, we need common sense regulation, people-first solutions and billionaire taxation. So vote for Democratic candidates and solutions!