AI Debate Should Prioritise Pro-Worker Systems to Prevent Widening Inequality: Expert

The current direction of artificial intelligence development could deepen inequality and accelerate job losses unless the industry shifts towards systems designed to complement workers rather than replace them, according to a paper by MIT economist Daron Acemoglu.

In his paper, *The Humanist Review of AI*, Acemoglu, an Institute Professor at the Massachusetts Institute of Technology, argued that the AI debate should move beyond speculation about Artificial General Intelligence (AGI) and focus more closely on practical policies that encourage the development of pro-worker AI.

AGI refers to a hypothetical form of artificial intelligence capable of matching or exceeding human cognitive abilities across a broad range of intellectual tasks and domains.

Acemoglu argued that steering AI towards a more beneficial path would require a broader change in perspective rather than relying on a single policy measure. He said AI should be developed to complement human capabilities and improve people's lives rather than contribute to their economic and social disadvantage.

The paper called for policy measures aimed at creating stronger demand for pro-worker AI tools, demonstrating their effectiveness, encouraging new business models and investing in infrastructure that supports their development.

Acemoglu also criticised the prevailing business models of major technology companies, which largely depend on selling software to corporations and generating revenue through digital advertising. He argued that these models have not created sufficient incentives to develop AI systems focused on supporting workers.

The concentration of the technology industry could further discourage new companies from experimenting with alternative approaches and business models, the paper said.

Acemoglu also pointed to what he described as a practical problem within businesses. Companies often assume that technology providers will primarily offer automation tools and consequently structure their future plans around replacing or reducing human labour.

At the same time, technology companies expect businesses to continue demanding automation-focused products and therefore see limited commercial incentives to invest in pro-worker AI, creating a cycle that reinforces the existing direction of development.

The paper noted that around 52 per cent of US citizens are concerned about the impact AI could have on their jobs. Public concerns also extend to the potential impact of AI on the information environment and democracy, while some experts and commentators have raised much broader fears about existential risks from advanced AI.

However, Acemoglu argued that the economic benefits of AI may fall short of some expectations within the technology industry. He pointed to studies and reports that have found limited evidence of significant productivity gains from AI adoption across many companies.