A.I. came at ‘perfect time’ to save the economy, VC says

“We are rapidly reaching the limits of current AI,” SK Ventures partners Paul Kedrosky and Eric Norlin wrote in their firm’s report. sub stack. In a post titled “AI Isn’t Good Enough,” they argue that, less than a year after ChatGPT entered public awareness, “we’re quickly reaching the limits of current AI, whether it’s because it’s prone to hallucinations.” , or because the training data is insufficient, the field is narrow, the training corpus from many years ago is out of date, or there are countless other reasons. ”

More provocatively, they argue that we are living in a strange time when technology is too advanced to pose a challenge to employment for many in the near future, and not far enough advanced to be truly productive improve. They describe a dynamic they call a “labor wormhole” that is “eating the economy,” saying we need either better AI or worse AI, but the limits of current technology Sexuality puts us in the “middle zone” where AI is already capable of rapidly displacing large numbers of workers, but has yet to provide broad enough economic benefits.

Here’s What Wormholes, Middle Zones, and ‘Better’ Versions of AI Mean

era of chronic shortages

SK Ventures’ thesis is squarely focused on a theme of the 2020s and the pandemic economy: shortages.

First, there is a shortage of the building blocks—chips, training data, large language models—that underpin the technology. This scarcity in turn drives up prices, making it harder for companies and startups to innovate in cost-effective ways. The cost of chips, in particular, is still too high. Chipmaker Nvidia, whose chips start at about $15,000, surpassed $1 trillion in market capitalization in June and reportedly has an 80% market share. New York Times. Kodrowski and Nolin argue that AI innovation will stall until costs come down. “This wave has been great for some companies, notably Nvidia, but in the future people will think this wave is mostly about pipelines in AI,” they said.

They argue that workers also face clear challenges and point to an incomplete recovery in the labor force participation rate. This means that the employment-to-population ratio in the U.S. economy, while significantly higher than in 2020 (the period of the worst job losses in modern times), would still struggle to reach 2019 levels, let alone 2007 levels before the Great Recession .

To understand what’s really going on in the economy and the role of artificial intelligence in potentially solving the problem, Kedrosky and Norlin use the wave metaphor. While many investors will associate AI with the explosive growth of ChatGPT in 2023, they see this as effectively the end of the first wave of AI that began in 2017.That year, an influential article was published Paper by a team of Google researchers, All you need is attention, This becomes the basis of how to train an AI model. The partners wrote that the current wave will continue for another year or two, ending only when costs come down across the board. For that, the world will need more, newer models such as thought tree models, cheaper chips, and the “inevitable commoditization” of large language models that will be offered as a service.

There are signs that the cost of accessing AI infrastructure will drop. Amazon has made it clear that it wants to directly challenge Nvidia’s dominance in chipmaking. Other major tech companies, such as Meta and Alibaba, have made large language models freely available to developers.

The pair speculate that a new era of technological advancement in cheaper AI computing systems will last until 2030. More importantly, it will help the U.S. economy cope with a looming productivity downturn as the country struggles to find enough workers to fill all the open jobs.

The advent of artificial intelligence comes at a perfect time to tackle a shrinking workforce

While AI is entering its next phase of development, the birth of the industry is “perfect timing” for the global/US economy, the partners wrote. The U.S. economy is facing an existential problem that there may not be enough workers to fill all the jobs. Essentially, the current extremely tight labor market will become a permanent fixture of the economy rather than a recent trend. “America’s labor force sinks into a wormhole and disappears,” write Kedrowski and Nolin.

current digital display The U.S. Workforce Stabilizes After decades of steady growth. This trend, exacerbated by the “Great Resignations” brought about by the COVID-19 pandemic, is a one-off rather than an occurrence over time. Kedrosky and Norlin estimate that if the U.S. labor market continues to grow in line with GDP and its pre-pandemic trend, current employment would increase by 5 million. They’re not the only ones pointing to demographic trends.The Census Bureau also records data This suggests that the number of workers entering the labor force will not offset the number of workers retiring in the near future.

Kodlowski and Nolin argue that demographic trends will eventually lead to a sharp decline in overall productivity as industries such as retail, manufacturing and health care struggle to fill open jobs. There are signs that these trends will in fact continue. The overall labor force participation rate remains about one percentage point lower than in February 2020. size The population of the United States is equivalent to millions of workers.

AI may reduce productivity until it’s good enough to really replace knowledge work

The scarcity of human workers means that labor becomes more expensive than capital, Kedrowski and Nolin say. When economies find themselves in this situation, they turn to automation to address labor shortages. In the past, prominent tech CEOs such as IBM’s Arvind Krishna and Google’s Eric Schmidt have also pointed to demographic trends in developed countries that support A case for AI innovation. The big difference about the coming wave of automation triggered by artificial intelligence is that it will primarily target jobs that involve “tacit knowledge.”others have done similar claimssaying that AI will appear first in white-collar jobs, a marked change from most types of innovation that have primarily impacted agriculture and industrial manufacturing.

The problem is, however, that some of these jobs are too complex to be automated by the current state of AI. “However, we are in a middle ground where AI can quickly replace large numbers of workers without providing compensation and broader productivity benefits,” the duo wrote road.

This could create a world where some workers lose their jobs, causing them to lose their jobs, but the technology is still too rudimentary to produce meaningful productivity gains – making everyone worse off as workers lose their jobs while the economy remains depressed . “Not all waves of automation create jobs as quickly as they displace them,” Kodrowski and Norlin write. “And, more importantly for our purposes, not all waves of automation lead to bursts of productivity to compensate for displacement.”

This middle-ground innovation, which displaces workers without increasing productivity, is known as “so-so automation,” according to one company. Research Papers Two venture capitalists cited this. A recent example cited by Kedrosky and Norlin is the replacement of call center workers by generative AI that can answer customer service questions. Employees are not usually transferred to another part of the company; rather, they are transferred to other parts of the company. They are usually just fired. Customers will end up with a worse experience because if the AI ​​chatbot service representative is not provided with accurate and correct prompts, they may not get any useful advice, creating a lose-lose situation for everyone involved. “These goods and services will inevitably displace large numbers of people, but they will not drive human prosperity,” Kedrowski and Nolin wrote.

Kodlowski and Norlin’s fellow venture capitalist Marc Andreesen remain firm believers in AI’s productivity benefits. “Technology doesn’t destroy jobs, and never will,” Anderson wrote in a widely circulated essay. blog post He argues that even if AI does somehow take away every human job, that’s actually a good thing. “It would be a straight-up material utopia that neither Adam Smith nor Karl Marx could have dreamed of,” Anderson said of a world where artificial intelligence does all the labor.

A Report Consulting firm McKinsey appears to support the Andresen school of thought that AI will lead to productivity growth. The McKinsey study estimates productivity growth of 0.1% to 0.6% a year through 2040, although it acknowledges that part of the variance in its forecast depends on the economy’s ability to move unemployed workers into new jobs.

Complicating Kedrosky and Norlin’s thesis is that once a generation of workers is trained in AI, they may further displace existing workers who lack these skills. Whether this possibility boosts productivity, as higher-skilled workers displace less-skilled workers, or dents it through mass unemployment remains to be seen.

What both Kedrowski and Nolin believe they are convinced is that automation without “high productivity gains” would lead to “economic dislocation,” especially when workers are as hard to find as they are in the current labor market.

Kodrowski and Nolin argue that the next wave of AI innovation will be critical to helping the U.S. address the macro issues facing a shrinking workforce. Without it, the U.S. economy risks being the first to fall into a “labor wormhole” and risk being eaten alive. “If we want to escape the automation of the past few decades and compensate for the gravity that has dragged the American workforce down the wormhole, we need to look beyond the limits of current AI technology.”