Transcription
The discussion about whether jobs will be lost due to AI is, well, heating up in various places. There's a phenomenon called the Jevons Paradox, and with AI, the cost of various intellectual labor is decreasing. For example, the development cost for creating software is also decreasing. As a result, won't more engineers be needed? Hello, everyone. I'm Takahiro Ano. Today, I'd like to talk about the theme of "What's the deal with AI, really?" As many of you know, AI is becoming capable of doing many things, and there's a discussion about whether jobs will be lost due to AI, whether there will be AI unemployment, which is, well, heating up in various places. Since there are various perspectives on this, I'd like to take a look at what the reality is today. First, let's see what people in the industry, for example, those at AI development companies, are saying. At the top here, what is Dario Amodei, the president of Anthropic, saying? He said that within 1 to 5 years, about half of entry-level white-collar jobs will be affected. And that the unemployment rate might become around 10% to 20%. Dario Amodei said this last year. I think Dario's perspective is one that's easy to understand. Since AI is becoming capable of doing many things, it's not so much people doing field work, but rather people working in offices, so-called white-collar workers, who will be affected, not blue-collar workers. And among white-collar workers, it's not so much senior people who do difficult work and make difficult judgments, those who are older and experienced, who will be affected. Rather, it's junior-level people who are heavily involved in research tasks or administrative work who will be affected. This is one perspective. Dario Amodei apparently made this statement in 2025 last year, but as of now, one year later, the unemployment rate is not yet 10% to 20%. These are quite high figures, but it hasn't reached 10% to 20%. So, the situation hasn't become like that yet. For your reference, the unemployment rate in the US is around 4%, and in Japan, it's in the 2% range. An unemployment rate of 10% to 20% is a very large number, and it hasn't happened yet.
Another perspective, different from this, which is often talked about in society, is what Sam Altman of OpenAI is saying. He talks about "AI washing," where layoffs or restructuring due to other reasons are being justified by AI. What he's saying is that many tech companies are currently conducting large-scale layoffs. And some people are blaming AI for this, but he says, "Wait a minute." For example, in 2020, there was the COVID-19 pandemic, and during that time, many companies hired extensively. They hired a lot. And now, the backlash from that is leading to layoffs, so AI shouldn't be that much of a factor. He's saying that everything is being attributed to AI for layoffs. So, Sam Altman is saying that while layoffs and rising unemployment might happen in the future, it's possible that these are actually happening due to other factors, and people are blaming AI. This is also a slightly different perspective.
And then there's one more, quite interesting, perspective. There's a phenomenon called the Jevons Paradox. With AI, the cost of various intellectual labor is decreasing. For example, the development cost for creating software is also decreasing. Normally, one might think that if 10 people can now do the work that 100 people used to do, then 90 people will no longer be needed. This is an easy-to-understand perspective. However, there are many examples in the past where efficiency increased due to similar technologies, but the number of jobs did not decrease. In 1865, there was an English economist named Jevons. Jevons was researching the coal problem. At that time, in 19th-century Britain, coal was being used extensively due to the Industrial Revolution. Research and development also progressed, and efficiency improved so that the same amount of coal was not needed. This was wonderful. One would think that the amount of coal needed would decrease. However, what happened was that because coal could be used so cheaply for this work, many fields emerged that wanted to use more coal. Companies wanted to use it, and regions wanted to use it, which led to an increase in demand. As a result of increased demand, the total amount of coal used also increased. So, while it sounds like coal consumption should decrease if efficiency increases, more coal was consumed at that time. The more efficient it became, the more it was consumed. This is the Jevons Paradox. And the same situation is expected to happen with AI. Simply put, the cost of creating software has decreased. However, because software that was previously so difficult can now be created with this much effort, people think, "Well, then let's try making this app," or "Let's try making this kind of software." In this way, demand increases in various applications and industries, and as a result, won't more engineers be needed? This is the story of the Jevons Paradox.
So, broadly speaking, there are three perspectives. Like Dario, who says that if AI can do more, then employment in those areas will be lost. No, that's not right. The second perspective is that while there's a perception that jobs are being lost or are about to be lost, it's actually AI that's being advertised as the problem, while it's just reducing what was previously over-hired. The third perspective is, no, no, as AI continues to improve efficiency, people will want to use it in more places, so the number of software engineers needed will increase. These three completely different perspectives coexist.
Therefore, let's look at the data to see what various people are saying and how they are presenting it. I apologize for the messy handwriting, but please take a look. First, let's look at the figures from major global economic organizations. First, the IMF. According to a report by the IMF (International Monetary Fund), about 40% of global employment is expected to be affected by AI. It's important to note here that "affected" does not necessarily mean "unemployed." For example, it means that productivity will increase, or that work can be done in a different way than before. So, about 40% of global jobs are expected to be affected by AI. In developed countries, that percentage is even higher, with about 60% of jobs expected to be affected by AI. Furthermore, the report states that of that 60%, about 27% are jobs that are likely to be replaced and whose demand will decrease due to AI, or jobs that carry the risk of wage reduction. The remaining 33% are expected to have a complementary effect, making work easier with AI, or playing a complementary role in changing how we work. Therefore, the IMF states that about 27% of jobs in developed countries may be replaced by AI. So, it's a rather pessimistic outlook.
Goldman Sachs, a financial institution, stated in its 2023 report that about two-thirds of jobs are expected to be affected. The report states that two-thirds will be affected and one-quarter will be replaced. Calculating from these ratios, approximately 300 million people are expected to be replaced, according to the report. On the other hand, the World Economic Forum (WEF) presents the opposite view, stating that about 78 million jobs are expected to increase. In this report, it is stated that while one technology may lead to a decrease in jobs, another may lead to an increase. And the combined result is this. Various technologies are listed, and for each, it states how many jobs will decrease and how many will increase. In the field of AI and information technology, 11 million jobs are expected to be created. On the other hand, 9.2 million jobs are expected to be lost. Adding these up, the net effect is an increase of about 1.8 million jobs due to AI, according to the WEF. Various technologies are listed, and as a result, the total is expected to increase by about 78 million people. So, the WEF predicts that jobs will increase due to the direct impact of AI alone. Therefore, the forecasts of various economic experts in the world are also slightly different and divided. Some people say that jobs will decrease significantly, while some institutions say that jobs will increase even with AI.
Next, let's look at what's happening on the ground, according to reports. A survey by Stanford University suggests that the impact is already starting to be felt. When examining various job categories that are likely to be significantly affected by AI, it was found that among those aged 22 to 25, i.e., younger age groups, the number of jobs decreased by 13% since 2022. This is a 13% decrease over three years, from 2022 to 2025. This survey suggests that entry-level jobs for younger people are already being affected.
According to figures from the Brookings Institution, a US think tank, there is data showing that unit prices are decreasing. In freelance jobs that are considered easily replaceable by generative AI, the number of new contracts has decreased by 2%, and sales have decreased by 5%. This report indicates that the impact on unit prices is already starting to be felt. The third item is an article from the Los Angeles Times. It reported that even students who majored in computer science at Stanford University are having difficulty finding jobs after graduation. This is quite shocking. A while ago, Stanford University was one of the top universities in the field of technology and information science, so it was unimaginable that graduates would have trouble finding jobs. However, this is what is happening, according to the article.
Therefore, looking at the current situation, while mass unemployment is not yet occurring, it seems to be happening to some extent among younger people. Or, it's not that people are unable to find work, but rather that the unit price of work is starting to decrease, especially in job categories that are highly susceptible to generative AI. These are the kinds of reports that are starting to emerge, and this seems to be the current situation in the United States.
Finally, what about Japan? When researching various aspects, Japan is a very unique country. And that is because society cannot function without AI. Looking at the forecasts, in Japan, according to statistics released by Recruit, there will be a shortage of 11 million people, not jobs, by 2040. There will be jobs, but about 10 million fewer people to work them. This is according to Recruit. By 2040, which is only 15 years away. A rather alarming figure of 10 million people being short in 15 years is being presented. Therefore, the shortage of people is more serious than AI unemployment. Similarly, Persol, with a slightly different timeframe of five years, predicts a shortage of 3.84 million people by 2035. While the exact figures differ slightly, both of these statistics suggest that a shortage is highly likely. So, the situation is quite different from that in the United States.
Now, let's look at the current situation. I don't think there's been a significant change starting to happen yet. Regarding new graduate hiring, according to a survey by Kyodo News, companies that responded that they will reduce new graduate hires for 2026, compared to last year, have outnumbered those that will increase them for the first time in five years. So, the tight market for new graduates has eased slightly for the first time in about five years. However, looking at the numbers, it's not a drastic change, and it seems to have reached a pause. To what extent this is due to AI's impact is probably not very significant yet. Also, there are occasional articles mentioning personnel transfers in Japanese companies. Mizuho Financial Group, for example, announced news that they will be making about 5,000 personnel changes gradually over 10 years. In Japan, due to regulations on dismissal, companies cannot easily fire people compared to US companies. Therefore, large-scale restructuring is unlikely to occur, and instead, they will proceed by having people retrain and move to different departments.
So, it can be seen that in Japan, the impact is starting to be felt in entry-level job categories, particularly in new graduate hiring. However, it can also be viewed as just a cycle of economic fluctuations, and it's only the first time in five years that such a situation has occurred. Overall, strong signals have not yet appeared, and this is likely the current situation.
Therefore, regarding AI unemployment, I have presented the current situation in Japan, the US, and the world, looking at three different perspectives and data, including forecasts and reports. Based on that, I recently heard something from someone at an AI company. This year, the company, which used to mainly accept graduate students for internships, is now seeing a lot of undergraduate students. More undergraduate students are coming for internships than graduate students. This is quite strange. While a significant proportion of science and engineering students in fields like AI and information science typically go on to graduate school, the ratio of undergraduate students has suddenly increased in about a year. This is an uncomfortable situation. When asked why, it turns out that the undergraduate students applying for internships are rapidly thinking that software engineering jobs will disappear. This is apparently called "gyururisation" in that circle. They are apparently calling it "gyururisation." Because they think they are "gyururised," they want to get an internship at that company as soon as possible, within the next year, and get in while there are still jobs available. This is apparently what is happening. While the extent to which this is perceived as large-scale and what the actual outcome will be is debatable, it is undeniable that students are starting to think this way.
So, what do I, Ano, think? I have no idea, but one thing I can say is that even if there aren't strong statistical signals right now, this could start to happen rapidly. So, even if this year is not much different from last year, there's a scenario where companies could suddenly change gears next year and stop hiring new graduates. I think this is a possibility, and it's important to be aware that while we look at the current situation, it's not guaranteed that next year will be a continuation, and there's a possibility of a sudden surge in numbers. I also don't know what will happen in the future. However, if there's a sudden surge, the country needs to have solid policies in place and be prepared, and as individuals, we need to think about what we will do when that sudden surge occurs and take action accordingly.
So, that concludes my discussion on "What's the deal with AI unemployment? May 2026 edition." I will continue to monitor these figures by looking at various reports. If you found this interesting, please subscribe to the channel and give it a thumbs up. Thank you. [Music] [Music]