AI, Technology, and Equality

Technological change has always had huge impacts on societies and economies that were impossible to predict in advance, and hard to appreciate even in retrospect. The big technologies of the industrial age—fossil fuels, mass production, radio and TV, and nuclear power, required large scale and hence tended to concentrate power. They played a role in the rise of the 20^th century’s big dictatorships and large industrial combines that were able to develop and make use of them, and needed to be tamed and regulated by democracies before their benefits spread.

The rise of digital technologies in the later part of the 20^thcentury were, from the beginning, thought to be democratizing since they distributed information and thus power broadly and cheaply. They indeed did that a little too well, upsetting the hierarchies that certified and distributed information like legacy media companies, governments, universities, and other elite institutions. In an internet-enabled world where anyone could say and publish anything, power spread out a little too far and undermined trust in authority across the board. In this period it was authoritarian states that gradually learned how to make use of the opportunities provided by the digital world to weaken their opponents.___STEADY_PAYWALL___

It is a fool’s errand to try to predict the long-term social impacts of new technologies like artificial intelligence. There has been a huge amount of chatter over generative AI programs like OpenAI’s ChatGPT, and outright fear of what is to come. People’s most recent experiences with digital technology have made them wary of disruptive innovations, and they try to predict the future analogizing from past waves of change. The spread of computers and digital technology is widely blamed for the large increases in economic inequality that has occurred across the developed world over the past couple of generations, as the rewards to education and cognitive ability have increased. Contemporary polarization revolves around an educational divide, with highly educated professionals clustering in big urban conglomerations and voting for more progressive parties, and less educated working people living in lower-density regions voting for conservative and populist parties. Automation, as well as technological advances in transportation and communications, have permitted the emergence of a globalized economy where low-skill jobs are easily shipped to low labor cost parts of the world. While this has equalized outcomes internationally, it has decreased equality in rich countries.

Many people assume that the advent of AI will simply be part of this broader wave, and that the technology will further exacerbate socio-economic divisions by automating away the livelihoods of people with lesser skills and education. But the historical account of the social impact of the transition into the digital age was more complex than the current cartoon version has it, and the likely consequences of AI for future equality even more unpredictable. It is entirely possible that future AI will increase rather that decrease equality in certain key respects.

While it is doubtlessly true that the transition from an industrial to an information economy increased rewards going to education, there was another respect in which it greatly increased equality, which had to do with gender. I made this case originally in the least read of all of my books, The Great Disruption: Human Nature and the Reconstitution of Social Order, published in 1999. Beginning in the late 1960s, there was a massive increase in female labor force participation across the developed world, as hundreds of millions of women entered labor markets. A contemporaneous technological innovation, the birth control pill, separated sex from procreation and reduced the risks of sex for women. The rise of feminism in this period is often thought of as an autonomous cultural event, but it would not have been possible for women to assert their independence were they not able to support themselves economically in new ways.

Technology had a massive impact on gender relations. Mechanical energy began substituting for upper body strength across the board. Many people have noted the impact of labor-saving devices like washing machines and dishwashers for women in the home, but more important changes were going on in the workplace. Typical working-class occupations shifted from digging ditches and mining coal to sitting in front of computer screens, working in retail, providing health care, and other service activities.

In an information-empowered economy, women had great advantages over men: they were more diligent, reliable, and unlikely to take crazy risks at younger ages. They were also much more willing to sit through lectures and acquire educations; as Richard Reeves has noted, they now attend colleges and universities in much larger numbers than do men across the world.

This shift in gender relations was of course hugely disruptive socially and politically. Families began to break down as women asserted their independence and increasingly had the financial means to support themselves and their children. Women began to displace men as the principal breadwinners in many households, leaving men highly insecure about their social status and roles. We should not underestimate the degree to which the current populist backlash has a gender dimension, with working class men becoming the core constituency for populist politicians around the world. Womens’ roles in the economy and society more generally are far from being accepted and secure, but there has nonetheless been a massive increase in gender equality, made possible by the shift into an information economy.

This brings us to the question of the likely social impacts of AI. Obviously, the people who design and make AI systems will be richly rewarded for their cognitive skills and educations. But will AI systems destroy more jobs than they create?

It seems to me that programs like ChatGPT may have a much more benevolent effect. Since they simply regurgitate content that is on the internet already, these large language models will be full of errors and will need to be constantly checked and certified by their human users for any high-end application. They will obviously be a big problem for professors like me who have to grade papers and read application essays. But in other respects, they may greatly facilitate the ability of people with lower levels of skill and education to write competently.

We tend to underestimate the importance of writing skills in our current focus on STEM. A modern economy requires a huge amount of ordinary writing: performance reviews, product descriptions, manuals, ad copy, proposals, after-action reports, accident descriptions, thank you letters, and so on. And many people have a great deal of trouble writing. One of the great advantages of ChatGPT is to eliminate the terror induced by a blank sheet of paper for such people, producing a first draft that can then be easily edited and refined. Just as the universal availability of Wikipedia has eliminated the need to remember dates and facts, so generative AI may allow people poorly served by the current education system to do useful work in an information economy.

This doesn’t mean that the advent of generative AI will not have big unanticipated and unpleasant consequences. Among the most threatening are deep fakes, which are spreading rapidly and seem poised to undermine people's trust in virtually everything they see and hear. We will need regulatory systems to restore trust in their provenance, and perhaps to slow down their advance overall so that society can adjust to the disruption. Ezra Klein has suggested not shielding AI companies from liability for the harms they produce the way we shielded internet platforms under Section 230 of the CDA. Whether this the right approach is impossible to know at this juncture, but we will definitely have to regulate.

Francis Fukuyama is chairman of the editorial board of American Purpose and Olivier Nomellini Senior Fellow and director of the Ford Dorsey Master’s in International Policy program at Stanford University’s Freeman Spogli Institute for International Studies.

Image: Rolf van Root / Unsplash