Fertility and the AI event horizon

Combining two issues that I don't fully understand

Photo by Possessed Photography on Unsplash

I don’t feel that I have much of a grasp of either the fertility issue or the AI issue. But I see some interesting parallels between the two issues, which might end up being interrelated. More specifically:

1. Both issues seem to have become dramatically more important over the past 5 years.

2. Both issues might radically change the world over the next few generations, perhaps more than any other single factor.

3. In both cases, the “medium projection” is itself quite stunning in its implications.

4. Rapid AI progress might create a sort of event horizon, beyond which it is impossible to forecast any societal trend, including fertility.

I’ll start with a few observations on fertility, and then in part 2 consider AI progress.

Part 1: China’s very odd history of fertility

When I was young, my mother took me to the University of Wisconsin to hear a talk by Paul Ehrlich, who wrote The Population Bomb. By the time I was old enough to study economics, it was already becoming apparent that he was wrong. That fact turned me into a sort of fertility skeptic, that is, skeptical of any sort of forecasts about future trends in population and living standards.

A bit later, India and China made attempts to reduce fertility, in ways that led to some pretty severe violations of human rights. This made me skeptical of government meddling in fertility decisions. (I’m not saying that skepticism is justified in all cases, just telling you where it came from.)

China’s one child policy was particularly controversial. But I don’t think enough thought has been given to whether it was even effective. The standard view seems to be that it was somewhat effective, but at a severe cost to human rights. But was it?

The FT has an interesting article discussing China’s more recent policy of encouraging more births. This graph caught my eye:

For anyone familiar with China’s one-child policy, this graph is astounding. Indeed it’s so odd that I fully expect someone in the comment section to explain to me that I’ve misinterpreted the graph in some fashion.

China adopted a one-child policy in 1980, and ended the policy in 2015. Do you see the problem? Prior to 1980, China’s birth rate had been plunging sharply. After 1980 the birth rate actually increased, not falling back to 1980 levels until 1993. But during this period of time, other East Asian countries were seeing declines in their birth rates. It’s not obvious that the one-child policy had any effect at all!

Even I would not make such an extreme claim; we have plenty of anecdotal evidence of Chinese women being forced to have abortions. And there was an upward bump in the number of births in 2016 and 2017, when the restriction was lifted. But then the number of births (which had been stable during the previous 15 years) began a precipitous decline, which has continued up to the present time.

If someone were told that China adopted a draconian fertility policy during the period from 1980 to 2015, and were not told what the policy was, then a glance at this graph might lead that person to assume that it was a pro-fertility policy.

According to the FT, experts are even more skeptical of the current policy of encouraging more births:

However, experts are sceptical that official measures to bolster the birth rate will persuade young people to start families, especially as rising unemployment and tepid economic growth have reined in spending.

Wang Feng, an expert on Chinese demographics at the University of California, Irvine, said officials were resorting to the same “playbook of using administrative power to achieve demographic goals” that was evident during the one-child policy era, the 35 years from 1980 when families were restricted to having one child.

While Beijing successfully stopped couples from having multi-child families, it was harder to use administrative powers to achieve the opposite result, he said.

A BBC News article expresses skepticism regarding China’s claim that the one-child policy prevented 400 million births:

"The 400 million figure is based on the assumption that if the one-child policy hadn't come in, the fertility rate would have stayed the same," says Gietel-Basten. This would have been unique to China among countries experiencing similar industrial and economic developments in that period, he adds.

For South Korea and Singapore the fertility rate in 2013 was 1.2 and in Japan, 1.4. It was three in the Philippines and 2.3 in Indonesia.

Indeed it’s possible that the policy might have even led to a few extra births:

Another expert, Cai Yong of the University of North Carolina, has suggested the one-child policy may have caused anxiety, prompting people to have children younger, which could increase the likelihood of having a second.

The one child policy contributed to gender imbalance in China, although as with so many other variables, the data is a bit murky:

For each girl born in China there are 1.16 boys, according to the official Xinhua news agency. The CIA World Factbook, external states that only Liechtenstein has a higher ratio, of 1.26 boys born for every girl.

Sexually selective abortions have been cited as a major cause of China's unusual imbalance.

But Gietel-Basten says the data for China is "poor", and the births of many girls are not registered if parents have broken the rule by having two children. When it comes to school age, the number of boys and girls enrolling tends to become more equal, he adds, saying officials often turn a blind eye.

Over the course of my life, I’ve seen many supposedly earthshaking issues come and go. Do you recall the panic about teenage mothers during the 1990s? Never mind.

So my gut instinct is to be skeptical of the current fertility panic. And the same time, I cannot think of any logical reason why fertility pessimists are wrong. Even under the “medium scenario”, the current birth rate is so low that many countries will experience precipitous population declines. And birth rates may go even lower. None of the factors that seem to be depressing fertility appear likely to go away; indeed they may intensify over time. So maybe this time really is different.

And that leads to my next point. The looming AI boom creates an event horizon, beyond which it seems almost impossible to forecast.

Part 2: Does AI change everything?

Consider three possible scenarios:

1. AI progress slows dramatically, for some hard to specify reason.

2. Superintelligence turns out to be impossible in the medium term, but AI allows us to create billions of robots with human level intelligence, automating almost all jobs.

3. Superintelligence is created, and moves the world in a direction that I cannot even imagine.

What strikes me the most is that even the medium scenario (viewed as quite pessimistic by many Silicon Valley types) is almost insanely optimistic. Even without superintelligence, you can envision a world where we all live like “billionaires” (except perhaps for exclusive goods like oceanfront homes), with all the work done by robots. Imagine UBI where the “b” stands for billionaire.

What would fertility look like in that sort of world? I have absolutely no idea. It’s possible that we’d have so many fun toys to play with that fertility would drop even lower. At the other extreme, perhaps technology would radically lower the cost of child bearing and child rearing, making parenthood vastly more fun. Indeed in a world where most people were not working, perhaps large families would be the center of life, the thing that gives meaning to life. Either extreme seems possible.

More broadly, virtually all prediction seems to become impossible at the point where AI becomes the dominant force in the economy. For instance, using the usual metrics, our current fiscal regime is increasingly unsustainable. But that could change if AI led to supercharged growth. Indeed, I find it hard to think of any area of society where the future course of events is not completely dominated by AI in one way or another, at least once we get beyond the event horizon. And some AI experts believe that this event horizon is only a few years away.

My view is that we should not conduct our fiscal policy on the expectation that we’ll win the lottery, that AI will lead to fabulous riches. At the same time, many AI experts suggest that our odds of winning this lottery are close to 100%, and some of the experts who disagree don’t predict business as usual, they predict the extinction of the human race. Neither view leaves much room to worry about the public debt.

I find the current state of affairs to be slightly surreal. There’s a sense of everything being in abeyance, up in the air until we figure out what AI will do. I actually lean a bit toward the pessimistic side of things, the view that superintelligence may not be possible in the near or even medium term. But even this view allows for AI to radically transform society in a way that is qualitatively different from any previous technology. And what if I’m wrong, what if AI does far more than I current expect?

Photo by Tomasz Frankowski on Unsplash