Whether AI Will Replace Most Jobs Remains Genuinely Contested

It’s a common theme, so it’s no surprise that I get asked for my opinion on it too.

[Affirmative objection ①] Most jobs will be automated. This is only natural, since there are many things AI does with higher precision than humans. But this is the same thing that has happened again and again throughout history, from the Industrial Revolution to Japan’s period of rapid economic growth to the IT revolution. So there’s no need to fear it needlessly; just think of it as a gradual, natural shift in “the range of work humans should do.”

[Affirmative objection ②] AI overwhelmingly surpasses humans. Soon there will be nothing left for humans to do. Even the work of developing higher-precision AI will come to be handled by other AI, and humans won’t even need to do AI development anymore. This is where humans will be called on for a new kind of budding potential. What AI is good at is rational behavior. So humans should turn to things uniquely human and irrational: selfless service, non-profit activity, art. Freed from rationality, humans should move toward a new stage.

[Negative objection ①] In reality, what current AI can do is severely limited, and it’s a huge mistake to treat it as though it were all-powerful. This isn’t the first AI development boom in history, and of the “strong AI” and “weak AI” discussed in the past, all that’s happening now is that weak AI has produced remarkable results in specific fields, with talk that this seems likely to produce major economic benefits. Computers genuinely acquiring intelligence is still a long way off. There isn’t even a sign that such an era is coming.

[Negative objection ②] AI’s progress is remarkable, but it carries many problems. Because of its nature as an extension of statistical processing, AI is strong with probability. In other words, for the task of identifying “this is probably roughly right,” the balance of speed, precision, and cost is very good. Its weak points, conversely, are individual optimization and exception handling. But the real world is full of exactly that kind of thing. I can already hear the counterargument: “AI’s strength is finding patterns in vast amounts of data where patterns seem invisible at first glance.” But it’s obvious to anyone that running a giant supercomputer for days to handle a handful of extremely rare cases in the real world makes no economic sense. No one, and no company, will ever do such a thing, now or in the future.

These objections all seem to hit the mark to some degree.

As partly explained above, I think many engineers are well aware of AI’s technical constraints.

There’s a famous prediction about processor speed called “Moore’s Law.” It began when Gordon Moore, an engineer at Fairchild in the 1960s who would go on to found Intel, stated in a paper that processor performance would double every ◯ years (the number ◯ was revised each time and eventually settled around 18 months). By the 1990s, many engineers had recognized its limits, and from 2000 onward came the era of multi-core processors that continues today. When I once visited Israel, I heard the story of the engineers at Intel’s Israel branch who came up with multi-core processors. Apparently headquarters in Santa Clara wouldn’t listen to their proposal at all and instead pushed a different development theme on them. So they built a multi-core processor prototype in 9 months while also handling that assigned theme, and marched into Santa Clara carrying it with them.

It’s quite an exciting story, but the lesson here is twofold: most people don’t know about technical limits and don’t even try to notice them, and difficulties get overcome by a handful of wise engineers who notice those limits and demonstrate innovation.

I believe many technical challenges get resolved through this kind of innovation process. Recent “AI-generated” content, for example, really is remarkably real. It used to carry an unmistakable artificial feel, but in recent years that has improved dramatically. A technology called GAN (generative adversarial network) is often credited for this, but back in that shocking period in 2012 when Google used deep learning to have a Stanford supercomputer teach itself to recognize “cats,” surely no one had even imagined an idea like GAN.

If we take an optimistic stance on technical constraints, does that mean an era of hallelujah-to-AI is coming? I don’t think so. That has to do with industry structure.

There’s no field harder to topple a giant in than AI development. Developing new AI itself is fine. But is there anyone today with the motivation to develop image-recognition AI from scratch and take down the giants? I doubt it. The reason is data. To improve learning precision, the volume of data matters enormously. The GAN example above is a typical case. AI technology suppliers who have already launched image-related services own staggering amounts of image data, and that amount keeps growing constantly. It’s not easy for a latecomer to win.

I once advised an AI-based medical startup to increase the processing precision and speed of their equipment. They were building a “system” that could automatically classify cells. Being a system meant they were developing the whole chain from IN to OUT. IN is the stage of acquiring and organizing information: the camera that captures the images, the focusing technology, the mechanical technology for precisely positioning the cells. It’s an area full of troublesome challenges. Then, based on the input data, AI automatically judges various information about the cells, and finally completes the OUT by classifying them. In other words, my advice meant this: if you intend to win on AI alone, you’ll end up crushed by the giants in the end, so you have better odds if you specialize in what you’re good at.

A society without game changers is boring. Yet society, including the entrenchment of capital, is steadily heading toward gridlock. The more people demand from AI, the more AI itself is forced to acquire scarcity. What this means is that no one today would pay hundreds of millions of yen for an AI that merely recognizes animal species from images. Not just AI, technology in general becomes commoditized and eventually obsolete. This holds true whether it’s a hardware business, a software business, or a content business.

But self-generated scarcity sooner or later hits a wall in most cases.

Take television as an example. How interested are you in an 8K TV? Do you even remember that 3D TVs existed? Television is already an area where securing scarcity has become difficult. This isn’t especially negative. It’s not the industry collapsing or anyone’s negligence, but rather seems like a perfectly natural course of events.

Take the email client you use on your PC as another example. Did you know there was an era when people bought email clients for money? Can you say which client is hard to use, or which one is good? The very concept of an “email client” seems to have already faded, and this is all a story spanning roughly the last 20 to 30 years. Did you know that Famicom game cartridges (not including the console) used to sell for over ¥10,000? Now you can buy toys at Don Quijote with 128 titles loaded on them for around ¥500.

Scarcity is extremely important in capitalist society. This is different from monopoly. Creating scarcity through monopoly is not permitted in modern society (that’s what antitrust law is for). Scarcity is only born when you create, ahead of others, something users feel is “special value, something I want.” Prioritize being ahead of others and your pursuit of value ends up insufficient; prioritize value or performance and you lose to others, especially to giants. It’s an extremely difficult balance. But in a situation where the possibilities in the “value” dimension are wide open, for example in a situation like the current AI development field, where expectations and a sense of growth potential are strongly pervasive across society as a whole, opportunities are undoubtedly scattered everywhere. But this will eventually run its course.

AI development requires massive costs, and operating it in practice costs money too. The processing power demanded of edge devices keeps rising, which will inevitably become a factor squeezing costs. No doubt many innovations will solve these problems. But I believe imbalance will inevitably emerge somewhere.

At that point, I think we’ll start hearing people say, isn’t this better than AI?

Sōetsu Yanagi’s The Way of Crafts contains a passage like this: “A culture that ignores machines cannot stand, and I do not deny that machine-made products have their own particular beauty. At the same time, a culture in which humans become slaves to machines cannot stand either. Therefore society should not demand machines without limit.” Unlimited demand is exactly what excessive mechanization is. The concern that the sage Sōetsu Yanagi pointed out 100 years ago is right in front of us today.

The power of human misunderstanding and the lack of imagination can sometimes be tremendous. A typical example is the “human washing machine” from the Osaka World Expo: a system where you’d step into a machine, press a button, and your bath would be done. But who on earth wants a product like that? What’s even more surprising is that this washing machine wasn’t just an exhibit; for a time, a major manufacturer actually proceeded with real development of it. If Yanagi were alive, wouldn’t he declare that nearly every machine seriously presented under the theme of “predicting the future” was extremely ugly and unnecessary?

I’m a father of four, and recently my 6-year-old eldest daughter has become able to wash her own hair, which makes me deeply feel how much she’s grown. That said, her technique still isn’t quite there, and she asks me, “Daddy, how was it? Did I do it well?” When I point out spots she missed at the nape of her neck or around her ears, we go through an exchange like, “Hmm… how about now?” “Yeah, you did great.” I’m happy being in the midst of this wasteful, inefficient process. I truly, from the bottom of my heart, am glad the human washing machine never became a fixture in ordinary households.

The content of this post is an excerpt (original text) from the following book. If you’re interested, please pick up a copy.

The Shape of a Happy IoT Startup

The Shape of a Happy IoT Startup

Originally published in Japanese at https://clazytech.com/2022/08/1078/. Translated with LLM assistance and reviewed before publication.