Jack Dorsey shouldn’t scare people: Every employer needn’t deploy AI to lay human workers off

As an economist, I disagree with Dorsey’s prediction. Block is not representative of the tens of millions of US businesses, and the choice Dorsey made was not the only option. In conversations about AI, it’s important to focus less on dystopian scenarios and more on ways to bring about an AI future that fosters shared prosperity.

Widespread automation—replacing human workers and devaluing human expertise, as Dorsey predicts—is only one possible outcome of AI.

Another path is “pro-worker AI,” a phrase coined by a team of MIT economists, that enhances the value of existing human expertise and creates new tasks. Unlike AI automation, pro-worker AI that raises a worker’s productivity can increase wages and expand employment into new areas.

An existing example they give is a food delivery app in China that added a voice chatbot to support hearing-impaired delivery workers, significantly improving their performance.

A speculative example is an AI assistant that allows aircraft maintenance workers to become spaceflight maintenance workers. It may be easier to imagine the jobs that AI will destroy than the ones it will create, but job creation is common with many technologies. In fact, about 60% of jobs in 2018 were in occupational specialties that did not exist in 1940.

Whether AI is pro-worker is not principally about AI’s technological capabilities. It’s about how leaders in corporations, government agencies, schools or civic organizations deploy it. Essentially, the same AI tool could be used to support and empower workers, or to surveil and sanction them. As the authors note, it’s the intention, not the technology, that drives the effects on workers.

Speed is another dimension to consider. AI tools are developing rapidly, but we can still be deliberate in the implementation. Federal Reserve Governor Chris Waller recently spoke about the use of AI within the Fed system, including in its operations, such as payment processing. He argued that the Silicon Valley ethos of “move fast and break things,” is not appropriate for the Fed, given its responsibilities and need to maintain public trust.

“AI systems can amplify errors as quickly as they amplify efficiency,” Waller noted, laying out how the Fed built an internal AI platform and coding tools with guardrails for use across the system. The Fed doesn’t want to move so fast that errors occur, but it wants to move fast enough to benefit from the new technology. Speed is a choice.

Finally, there is a role for the government in encouraging a scenario in which AI broadly serves people. Public funding could be directed toward pro-worker AI or AI to improve public services, including competitive prize-style funding as used by the Defense Advanced Research Projects Agency or private-public partnerships like Operation Warp Speed, which rapidly developed the Covid vaccine.

The goal is not to stop AI but to encourage its empowering applications and provide examples of use cases beyond automation.

The private market, on its own, may focus too much on AI automation for a variety of reasons, as the MIT researchers suggest. It could be because these are the first AI tools being developed, and there is pressure from shareholders on companies to quickly demonstrate cost-cutting.

AI that disempowers workers could also be a way for managers to reduce workers' bargaining power and shift a firm’s economic profits away from workers. Or the longstanding vision behind artificial general intelligence may be biased toward worker-replacing outcomes.

In addition, the government will have a role in responding to the disruptions AI causes, but it should also play a role in shaping the technology in its early stages.

Despite the many challenges that AI presents, it also holds considerable promise. AI has the potential to substantially boost productivity growth—that is, increasing the size of the economic pie. And a growing pie is a much better backdrop for improving people’s lives than a stagnant or shrinking one. But a more positive future cannot simply be willed into existence. The decisions we make today will shape the AI of tomorrow. ©Bloomberg

The author is the chief economist at New Century Advisors and a former Federal Reserve economist.