The AI Awakening: Economic Implications of Automation, Remote Work, and the Productivity Paradox

This podcast episode features Erik Brynjolfsson discussing the profound economic implications of the accelerating "AI awakening," particularly amplified by the COVID-19 crisis. He highlights the breathtaking advancements in machine learning, such as image and speech recognition and machine translation, driven by deep learning techniques, increased computational power, and vast digital data availability. Brynjolfsson argues that we are entering a "second machine age" where mental work, not just physical labor, is increasingly augmented and automated. While these technologies promise immense wealth creation, a central concern is the distribution of this wealth and the potential for job displacement, drawing parallels to historical events like the Luddite movement.

Brynjolfsson distinguishes between technology acting as a "substitute" for human labor and as a "complement" that enhances human productivity. Historically, complementarities led to broad wage growth, but this trend has stalled or reversed in recent decades, suggesting a shift towards substitution. Using a machine learning rubric applied to O*NET tasks, his research indicates that while many tasks are suitable for AI, no single occupation is fully automatable. This implies a future of work characterized by restructuring and reallocation of tasks, rather than widespread unemployment, with humans retaining crucial roles even in highly automated fields like radiology.

Practical insights are offered for both individuals and organizations. Individuals are encouraged to focus on tasks less susceptible to automation and to develop skills that complement AI, advocating for a strategy of "racing with the machines." For companies, tools like the O*NET analysis and "job to vec" can help identify vulnerable tasks, assess AI capabilities, and plan for workforce transitions, such as reinventing existing roles or upskilling employees for new, growing occupations. The example of personal bankers transitioning to leadership or HR management illustrates this adaptive approach to automation.

The broader context includes the dramatic acceleration of remote work, with approximately 50% of Americans working from home during the pandemic, leading to uneven effects across regions and occupations, favoring information workers. Brynjolfsson also addresses the "modern productivity paradox" – the puzzling slowdown in productivity growth despite rapid technological advancements. He outlines four potential explanations: overestimation of technology's power, mismeasurement of benefits, unequal capture of benefits by a small group, and the inherent time lag for benefits to fully materialize across the economy, leaning towards the latter as a significant factor.

"there are decades where nothing happens and there are weeks where decades happen"

"we are now crossing a really important threshold if you look at the use of machine learning for many many tasks"

"just recently was in the past eight years or so we've had enough computer power and some improvements in the algorithms and most importantly massive increases in data availability digital data availability and those three things put together have allowed us to use neural nets much much more effectively"

"it seems like hardly a a week goes by when i don't see a new article in in nature or science describing how a new machine learning system has outperformed humans in a task similar to that"

"there's no economic law that says that everyone's going to benefit it's possible for some people uh to be hurt even as others benefit"

"what's being augmented and automated isn't just physical muscle power but brains and mental work as well and that's a much broader scope of set of tasks"

"we shouldn't just only think about substitution technology can and should be used to complement and help people what i said in my ted talk was we should learn how to race with the machines helping having them help us get our work done rather than racing against the machines where we see it as either or"

"although we looked at all the 950 occupations we've done not find a single one where machine learning ran the table and was able to do everything in each case there was still scope for humans having to do some things"

"i don't think we see massive end of work or unemployment it's really more of a restructuring as parts of tasks get done by machines and reallocated"

"there hasn't been a productivity boom and i think that's actually the mystery because these tools at least in my view have been quite quite remarkable but we're not seeing it show up in the productivity data"