4 min read

Failure to launch

Digging deeper into the world's declining fertility; are large language models (such as ChatGPT) civilisation-altering or pure hot air; how to slow scientific progress; and immigration, working class wages and the honeypot effect.
Failure to launch
Photo by Tim Mossholder / Unsplash

1—Failure to launch

The world's population will soon begin to decline due to plunging fertility rates. But why is that the case when stated fertility preferences – which "are extremely strong predictors of actual fertility behaviours, stronger than covariates like religion, education, race, income, or any other socioeconomic or cultural variable... have not declined much"?

According to Lyman Stone, by only focusing the demographic decline many policymakers may be missing the "the 'tragedies' under the 'statistics'":

"Whatever role economic and technological shocks may have had, they have not led people in most countries to report desiring fewer children. Actual fertility has fallen even as desired fertility has not in most of the high-income countries of the world. Thus, as with marriage, the likeliest story on falling fertility in the last two decades is not one of people simply freely choosing not to have so many children. Rather, fertility has most plausibly fallen because of economic 'failure to launch' among young people, long delays in career stability, excessive housing costs, exploding childcare costs, rising student debts, and other adverse circumstances, not least the oppressive panopticon of social media which makes prisoners of us all."

You can read Stone's full essay here (~14 minute read), which suggests looking beyond the statistics and focusing more on rectifying the causes of the decline in fertility, such as reducing drug and alcohol use and the resulting 'deaths of despair'; helping people achieve stability earlier in life; and improving housing affordability through liberalised zoning policies.

2—AI - bull or bear?

ChatGPT launched in November 2022 and has taken the world by storm. But will it – and its inevitable successors – be "civilisation-altering", or "pure hot air" like Web3 (blockchain), relegated to the fringes of society once the novelty wears off?

That's the question recently posed by Google's François Chollet, who provided a bull and bear case for large language models (LLMs) like ChatGPT:

"The bull case is that generative AI becomes a widespread UX paradigm for interacting with most tech products (note: this has nothing to do with AGI [artificial general intelligence], which is a pipe dream). Near-future iterations of current AI models become our interface to the world's information.

The bear case is the continuation of the GPT-3 trajectory, which is that LLMs only find limited commercial success in SEO, marketing, and copywriting niches, while image generation (much more successful) peaks as a XB/y industry circa 2024. LLMs will have been a complete bubble."

You can read Chollet's full twitter thread here (~4 minute read), which concludes that LLMs will keep getting better but will most likely be confined to "consumer products, and perhaps even in education and search".

3—How to slow down scientific progress

"Leo Szilard—the physicist who first conceived of the nuclear chain reaction and who urged the US to undertake the Manhattan Project—also wrote fiction. His book of short stories, The Voice of the Dolphins, contains a story 'The Mark Gable Foundation,' dated 1948, from which I will present to you an excerpt, without comment."

4—The honeypot effect

If a government restricts immigration, it raises the wages of the local working class by reducing the supply of unskilled labour. But even ignoring the possibility of capital/labour substitution, Dean Hoi found that in the long run doing so actually hurts local workers:

"The Chinese Exclusion Act [1882] had a significant, negative long-term effect on American workers. My estimate is that workers in locations exposed to the Act earned on average 6-15% less over their working lives than their counterparts in other locations.

The negative effects were strongest for low-skilled and unemployed workers.

The exclusion of Chinese immigrants not only failed to improve conditions for working-class Americans, but made them substantially worse off in the long run."

But why? According to Hoi, the negative outcomes were due to the 'honeypot effect':

"A closer look suggests the Chinese Exclusion Act was initially successful in boosting low-skilled wages and the employment of Americans in low-skilled jobs in the regions it had an effect.

This created a 'honeypot' – American workers in those locations increasingly took and remained in low-skilled jobs. They became significantly less likely to become educated, meaning they fell behind their counterparts in other locations on the occupational ladder."

You can read Hoi's full summary here (~4 minute read), in which he concludes that the higher wages caused by the Act incentivised many workers to stop progressing up the career occupation ladder, reducing their lifetime earnings.

5—Further reading...

👮‍♀️ 1,500 supporters of former President Jair Bolsonaro have been detained in Brazil "after the Congress building, presidential palace and Supreme Court were ransacked on Sunday".

🐦 But still up nearly 300% since Musk took over: "The number of active users on the Mastodon social network [an open source, federated Twitter competitor] has dropped more than 30% since the peak and is continuing a slow decline, according to the latest data posted on its website."

🐭 "Disney CEO Bob Iger told hybrid employees on Monday they must return to corporate offices four days a week starting March 1".

❌ "China suspended issuing short-term visas to South Koreans and Japanese on Tuesday... in apparent retaliation for restrictions imposed on Chinese travellers over Covid concerns."