According to one venture capital firm, the wave of AI innovation and investment sweeping Silicon Valley and Wall Street is coming to an end. It is also “largely nonsense,” but nevertheless holds the key to solving the major problem holding back the US economy. What happens next is the crucial part, the company adds.
“We are rapidly reaching the limits of current AI,” Paul Kedrosky and Eric Norlin, both partners at SK Ventures, wrote of their company’s Substack. In a post titled “AI Isn’t Good Enough,” they state that less than a year after ChatGPT exploded into the public consciousness, “we are rapidly reaching the limits of current AI, whether due to its propensity for hallucinations, inadequate training data in narrow fields, outdated training corpora from years ago, or countless other reasons.
More provocatively, they argue, we find ourselves in a strange age where technology is too advanced not to jeopardize many people’s jobs for the foreseeable future, and not even close enough advanced to deliver real productivity gains. They describe a dynamic they call a “workforce wormhole” that is “eating up the economy.” They say we need much better or much worse AI but the current limitations of the technology leave us in a “middle zone” where it’s already capable of quickly displacing large numbers of workers but it’s not yet delivering enough broader economic benefits on.
Here’s what they mean by the wormhole, the midzone, and a “better” version of AI.
The era of chronic shortages
The SK Ventures argument focuses entirely on the theme of the 2020s and the pandemic economy: shortages.
First, there is a shortage of the building blocks that make the technology possible: chips, training data, large language models. That scarcity, in turn, has driven up prices, making it more difficult for companies and startups to innovate in a cost-effective way. Chip costs in particular have remained exorbitant. Chip maker Nvidia, whose market cap hit the $1 trillion mark in June and reportedly has an 80% market share, has a starting price of about $15,000 for its chips, according to the New York Times. Until costs come down, AI innovation will stagnate, Kedrosky and Norlin argue. “This wave has been great for a few companies, especially Nvidia, but going forward it will mostly be about forwarding the AI house,” they said.
There is also a clear and distinct challenge for workers, they argue, pointing to the incomplete recovery in employment rates. This means that the share of people employed in the US economy, while markedly higher than in 2020, when there was the greatest job loss in modern times, will struggle to reach 2019 levels, let alone 2007 levels before the Great Recession.
To understand what is really happening in the economy and the role of AI in potentially solving it, Kedrosky and Norlin use a wave metaphor. While many investors would associate AI with ChatGPT’s explosive growth in 2023, they argue that this is actually the end of the first wave of AI that began in 2017. In that year, an influential article was published by a group of Google researchers: “Attention is All You Need”, which became the basis for training AI models. The current wave will last another year or two and will not end until costs fall across the board, the partners wrote. To do that, the world will need more and newer models, such as tree-of-thought models, cheaper chips and an “inevitable commoditization” of large language models that will be offered as a service.
There are indications that the cost of accessing the infrastructure on which AI is built will fall. Amazon has already made clear its intentions to directly challenge Nvidia’s dominance in chip manufacturing. While other major technology companies such as Meta and Alibaba have made their large language models available to developers for free.
The new era of lower-cost technology advancements in AI computing systems will last until 2030, the two theorize. More importantly, it will help the US economy cope with the impending declines in productivity that the country is struggling to find enough workers to fill all open positions.
The perfect time to tackle the shrinking workforce
Even as AI enters the next phase of its development, the industry’s birth came at the “perfect time” for the global/U.S. economy, the partners write. The US economy is facing an existential problem where it risks not having enough workers to fill all of its jobs. Essentially, the current extremely tight labor market will be a permanent part of the economy and not a recent trend. “The American workforce fell into a wormhole and disappeared,” write Kedrosky and Norlin.
Current numbers show that the U.S. labor force remains stable after decades of steady growth. The great resignation caused by the COVID-19 pandemic has exacerbated this trend, causing it to happen all at once rather than unfolding over time. Kedrosky and Norlin estimate that if the U.S. labor market had continued to grow in line with GDP and pre-pandemic trends, there would currently be an additional 5 million workers. They are not the only ones pointing out these population trends. The Census Bureau also recorded data pointing to a near future in which the number of aging workers will outweigh those retiring.
Kedrosky and Norlin believe that demographic trends will eventually lead to a drastic drop in overall productivity as industries such as retail, manufacturing and healthcare struggle to fill open positions. There is some evidence that these trends are, in fact, here to stay: the overall employment rate is still about a percentage point lower than it was in February 2020. On a US-sized labor force, that could equate to several million workers. .
AI could even drive productivity down to the next wave
The absence of human labor means that labor has become more expensive than capital, according to Kedrosky and Norlin. And when economies find themselves in such a situation, they turn to automation to solve their 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 the developed world as reasons to support AI innovation. The big difference with the upcoming round of automation, caused by AI, is that it mainly focuses on jobs that involve ‘tacit knowledge’. Others have made similar claims, saying AI will come first for white-collar jobs, a marked change from most previous types of innovation that mainly affect agricultural and industrial production.
The problem, however, is that some of these jobs are too complex to be automated by the current state of AI. That threatens to create a world where some workers are displaced, leaving them without a job, but by a technology that is still too rudimentary. to generate meaningful productivity gains; leaving everyone worse off as workers lose their jobs, while the economy remains unproductive. “Not all waves of automation create jobs as quickly as they displace them,” write Kedrosky and Norlin. “More importantly for our purposes, not all waves of automation provide productivity bumps that offset the displacement.”
This type of mid-market innovation, which replaces workers without increasing productivity, is known as “moderate automation,” according to a research paper cited by the two venture capital firms. A recent example is call center agents being replaced by generative AI that can answer customer service questions. The employees are often not redirected to another division of the company, they are usually just fired, but the customer ends up with a worse experience. “These products and services will inevitably displace large numbers of people, but they will not promote human flourishing,” write Kedrosky and Norlin.
Kedrosky and Norlin’s fellow venture capitalist Marc Andreesen remain convinced of the benefits of AI for productivity. “Technology is not and never will destroy jobs,” Andreesen wrote in a widely circulated blog post about the benefits of AI. And even if AI did, and somehow take every job away from a human being, that would actually be a good thing, he argues. . “It would be a straight spiral to a material utopia that neither Adam Smith nor Karl Marx ever dreamed of,” Andreesen said of a world where AI does all the work.
A report from consulting firm McKinsey seems to support the Andreessen idea that AI will lead to productivity growth. McKinsey’s research estimates that productivity will increase by 0.1% to 0.6% per year through 2040. Although the company acknowledges that some of the variance in its projections depends on the economy’s ability to move workers who lose their jobs to new positions.
Complicating Kedrosky and Norlin’s thesis is the possibility that once a generation of workers are trained in AI, they may further displace existing workers who lack these skills. Whether that possibility would boost productivity if more skilled workers took over from less skilled workers, or neutralize it as a result of mass unemployment, remains to be seen.
What the two consider certain, however, is that automation without “high productivity gains” resulting in “economic disruption” is almost a certainty, according to Kedrosky and Norlin, especially when workers are as hard to find as they are today. job market.
The next wave of AI innovation will be critical in helping the US solve the macro problems it will face due to a shrinking workforce, Kedrosky and Norlin argue. Without this risk, the US economy risks ending up with its head in the “workforce wormhole,” which risks eating it alive. “We need to look beyond the boundaries of current AI technology if we are to break free from past decades of automation and offset the gravity that has dragged the American workforce into that wormhole.”
