Will A.I. Be a Creator or a Destroyer of Worlds?

The advent of A.I. — artificial intelligence — is spurring curiosity and fear. Will A.I. be a creator or a destroyer of worlds?

In “Can We Have Pro-Worker A.I.? Choosing a Path of Machines in Service of Minds,” three economists at M.I.T., Daron Acemoglu, David Autor and Simon Johnson, look at this epochal innovation:

Now, Acemoglu, Autor and Johnson write, A.I. presents a direct threat to those high skill jobs: “A major focus of A.I. research is to attain human parity in a vast range of cognitive tasks and, more generally, to achieve ‘artificial general intelligence’ that fully mimics and then surpasses capabilities of the human mind.”

The three economists make the case that

Tall is an understatement.

In an email elaborating on the A.I. paper, Acemoglu contended that artificial intelligence has the potential to improve employment prospects rather than undermine them:

This, however, “is not where we are heading,” Acemoglu continued:

Acemoglu pointed out that unlike the regional trade shock after China entered the World Trade Association in 2001 that decimated manufacturing employment, “The kinds of tasks impacted by A.I. are much more broadly distributed in the population and also across regions.” In other words, A.I. threatens employment at virtually all levels of the economy, including well-paid jobs requiring complex cognitive capabilities.

Four technology specialists — Tyna Eloundou and Pamela Mishkin, both on the staff of OpenAI, together with Sam Manning, a research fellow at the Centre for the Governance of A.I., and Daniel Rock at the University of Pennsylvania — have provided a detailed case study on the employment effects of artificial intelligence in their 2023 paper “GPTs Are GPTs: an Early Look at the Labor Market Impact Potential of Large Language Models.”

“Around 80 percent of the U.S. work force could have at least 10 percent of their work tasks affected by the introduction of large language models,” Eloundou and her co-authors write, and “approximately 19 percent of workers may see at least 50 percent of their tasks impacted.”

Large language models have multiple and diverse uses, according to Eloundou and her colleagues, and “can process and produce various forms of sequential data, including assembly language, protein sequences and chess games, extending beyond natural.” In addition, these models “excel in diverse applications like translation, classification, creative writing, and code generation — capabilities that previously demanded specialized, task-specific models developed by expert engineers using domain-specific data.”