Beyond the noise: AI, augmentation, and harnessing tech power for good

While headlines related to generative AI tend to focus these days on its existential threats or utopian fantasies, an urgent challenge we must address is the test it will present to labor markets. The data is stark. Researchers foresee that by 2030 automation without generative AI could take over tasks comprising 21.5% of the hours worked in the US economy. With it, that share rises to 29.5%. We cannot foresee the long-term effects of AI on the job market but, undoubtedly, these figures point  to the potential exclusion of certain groups from future employment—particularly in the absence of guiding policies.

Technology in and of itself has resulted in greater global wealth. However, it is a wealth that is hindered by continuing–and at times, deepening–distributional inequality. Indeed, inequality has expanded dramatically within countries: over the last two decades the gap between average income of the top 10% and the bottom 50% of individuals has almost doubled—from 8.5x to 15x. 

Why is this particularly important now? At the societal level, job displacement and rising inequality are impediments to healthy democracy and social harmony. The greater the inequality, the greater the likelihood of grievance that can be weaponized and turned into polarized electoral results and social unrest. On the household and individual level, there are countless social and health consequences, including a rise in “deaths of despair” and subsequent decrease in life expectancy. Therefore, it is critical to pinpoint which jobs have been and will subsequently be lost, identify strategies to replace them, and retrain workers so that they are not left in the open. 

If we act soon, we can minimize generative AI’s potentially negative effects while simultaneously benefiting from its immense promise. While technology is a disruptor, with the right approach and focused policies, it also holds the promise to transform society. AI will undoubtedly allow us to better tackle key global challenges, including mitigating climate change, fighting disease, and promoting education. It is a unique moment in history. We can leverage technology to not just solve some of our global challenges but to help our societies become more equitable and humane. 

Hollowing out of the middle

The socioeconomic models of the last century are buckling under the pressure of geopolitics, technology diffusion, and economic instability. At the very heart of these developments are the prolonged weakening of the middle class, which, historically, had constituted the backbone of not only modern Western society but increasingly that of developing countries as well. Emerging technologies have played a role in the loss of jobs belonging to the middle-skill range: since the mid-1990s, the share of middle-skill jobs in total employment fell by about 9.5% in OECD economies, while the shares of high-skill and low-skill jobs rose by about 7.5 and 2% respectively. This trend of course has subsequent effects on income and inequality. For example, American adults in middle-income households declined from 61% in 1971 to 51% in 2019, while from 1970 to 2018, the share of aggregate income going to middle-class households fell from 62% to 43%. In contrast, the share held by upper-income households increased from 29% to 48% for the same period.

Many may be left behind 

Job and income loss is certainly not limited to the middle class. Geography, too, matters greatly. The digital divide is shockingly wide at the global scale: approximately 2.7 billion, or one-third of the world’s population, still do not have Internet access. Indeed much of the majority that do have access are limited by cost and quality of service, which could hinder their use of GPT and other consumer-facing AI technologies.

Blue Owl Group team members at the signing of the Biden Administration’s Executive Order on AI this week in Washington, DC.

Meanwhile, the rise in automation has diminished opportunities for developing countries to utilize manufacturing-led growth; the model that had historically enabled East Asian countries to make enormous leaps in economic and social development. While labor is in abundance and its relative low cost a comparative advantage for developing countries, automation is eroding that advantage. As such, even if per capita income increases in these countries, large segments of society will likely be left out. 

Discrepancies in income and job availability vary in-country as much as between countries. The US demonstrates great regional divergency regarding technology jobs. Research from The Brookings Institution has found that over 60% of generative AI job postings in the year ending July 2023 were clustered in ten US metro areas. Geographic divides aside, vulnerable groups will also be overly affected by AI. Its effect on women is an illustrative example. The ILO has determined that 3.7% of global female employment may be automatable with generative AI technology, compared with only 1.4% of male employment. In high-income countries, the share of potentially affected female jobs is 7.8%, more than double the 2.9% of male jobs for the same income group. 

Automation or augmentation

Daron Acemoglu’s research on inequality and automation finds that over half of the increase in inequality in the U.S. since 1980 is at least attributable–to some degree–to automation. This is largely stemming from wage pressure on jobs that could be automated. Furthermore, new jobs are not being created at the same rate they used to in part due to “so-so technology”, such as self-checkout kiosks, which neither improve productivity nor create new opportunities.

Businesses have focused primarily on leveraging automation as the best way to improve productivity and to increase cost savings. But in addition to automation’s negative externalities on employment and inequality, some argue it is not the most effective road to productivity gains. Erik Brynjolfsson, for example, suggests that “augmentation” has, in fact, been far more effective for productivity throughout most of the past two centuries. He cites that 60% of workers are now employed in occupations that did not exist in 1940 and concludes that, in short, automating labor ultimately unlocks less value than augmenting it to create something new. Brynjolfsson also notes that if human-replicating AI continues to focus on automating rather than augmenting labor, then it will risk creating even greater concentrations of wealth and power.  He warns of a phenomenon he calls “The Turing Trap,” a condition where those without power have no way to improve their outcomes. The avoidance of the Turing Trap–while benefiting fully from AI–is therefore “the grand challenge” of the coming era.

Six Policy approaches for alleviating job losses and inequality

Some of the catastrophic employment turbulence and displacements that pundits describe are not foregone conclusions. Effective policies can mitigate the disruptive power of AI on jobs and inequality and, ultimately, on society. Here are some options:

1. Solution-focused multistakeholder initiatives 

Brynjolfsson believes that the incentives that encourage automation over augmentation should be reversed and suggests the Turing Test itself as a starting point. He suggests the creation of a new set of practical benchmarks that consider augmentation. The Partnership on AI is a multistakeholder group composed of private sector, civil society, and academics, which is focused on creating guidance related to AI issues. It recently released Guidelines for AI and Shared Prosperity with tools on labor risk assessment and mitigation for AI governance frameworks. Multistakeholder initiatives like this one, which benefit from bringing together diverse perspectives, are particularly important as they can assist product designers in gaining a more holistic view of the various social and economic consequences their products may have. This is not just good for society but good for business, too.

2. Government-led strategies

Customarily, a wide range of public policies have been used in the past to subsidize private companies willing to invest in developing or deploying new technologies in the hope that their growth leads to higher tax receipts, high-wage employment, and other economic benefits. In recent times, the revival of industrial policy has controversially come to attention after years of neoliberal policies that took a more standoffish approach. Although industrial policy has not traditionally focused on job creation per se, seeing it rather as a by-product, it can be leveraged as such. 

In this vein, Dani Rodrik proposes a two-level “Industrial Policy for Good Jobs” that includes both federal and local aspects. Starting from the premise that innovations that complement rather than displace workers are feasible and currently undersupplied, the federal program would promote early-stage investments in nascent technologies that enhance worker skills and create good jobs. The local approach would build on the existing framework of development and business assistance programs, and create more jobs in the service sector.

3. Engagement with developing countries

Considering increasing automation, countries must broaden their focus beyond manufacturing; developing the service sector where there is potential for new areas of comparative advantage will be crucial for economic development. For example, middle-income countries are more exposed to the automating effects of GPT technologies, but their digital infrastructure and skilled workforce could also allow them to create complementary industries. Specifically, India and the Philippines may lose call centre work but their experience in business process outsourcing could help them to develop ancillary, high-growth industries. Developing countries are also starting to benefit from basic AI which can be used as a service via mobile phone. For example, Nuru is an app that East African farmers use to share pictures of leaves with authorities via mobile to monitor for pests and thus help decrease the risk of food insecurity. But much more investment in infrastructure is needed to ensure developing countries will not fall behind. Programs to improve connectivity, like the EU’s Global Gateway, are vital, too. 

4. Fostering innovation ecosystems

At the regional level, rather than traditional innovation “hubs” whose outcomes are somewhat mixed and more limited, broader strategies can be employed. For example, by ensuring solid, high-performing Internet infrastructure and the presence of a highly qualified workforce through ample training opportunities, remote work can become a tool to empower rural areas. These developments will have to be secured through increased investment and well-coordinated, comprehensive regional development policies. They also require political will. 

5. Targeted education and skills investment

Hereafter, education will truly be lifelong as automation  will require evolving skill sets. Manual labor skills will continue to be used heavily, but demand for these skills will likely decrease in proportion to technical, social-emotional, and cognitive skills. McKinsey research has found that in the coming years, soft skills will be even more important as cross-functional and team work will increase—requiring excellent communication skills and empathy. Likewise, workers will have to be entrepreneurial and responsive in the face of dynamic change. Higher cognitive skills such as analytical thinking and creativity will be needed more than ever, while more basic cognitive skills (for example, data inputting) are especially susceptible to automation. And while everyone will need basic digital skills, those with advanced technical skills will be the ones continuing to innovate. 

In addition to formal education, employers themselves will increasingly provide more training as workers retrain and upskill over the course of their careers. One innovation introduced in some countries, such as France and Singapore, is “Lifelong Learning Accounts,” whereby workers accumulate rights to training that are transferable across jobs and over which they have direct control.

6. Establishing global fora and institutions for coordination 

The international arena with regard to digital regulation and policies is currently dominated by national and geopolitical considerations. The US, EU and China have their own sets of interest and all three are pursuing alternative models. The UK is charting its own path, too. Furthermore, it is not clear in which direction the BRICS, a bloc of the world’s fastest growing emerging markets and who are increasingly exerting influence on other developing countries, will move. Benjamin Larsen referred to this current trend as a reemergence of “neo-mercantilism”, which is being manifested in many ways, including data, open-source software, and cloud computing strategies. Differing standards, tax regimes, and lack of interoperability all ultimately result in lost opportunities for economic development, job creation, and business growth. Furthermore, this degree of fragmentation and protectionism is not only expensive but potentially dangerous. Although national governments have their own interests, they should return to some degree of collaboration—both through global fora for debate, as well as institutions for international coordination. It is unrealistic to expect uniform views or global policies across the full gamut of issues, but we should hope–at the bare minimum– that frameworks can help improve employment and income potential. Tax and patent reform might be a starting point. 

Conclusion

It is clear that less inequality is better for society, better for democracy, and ultimately better for innovation and economic growth. Encouragingly, with the right strategies and policies, AI presents extraordinary opportunities for more inclusive growth and more equitable societies. Technology in and of itself cannot deliver this promise, but sound technology regulation–coupled with close coordination of economic and social policies–may yet enable us to turn our societies around. The public, too, has agency and we should use it—both thoughtfully and strategically. 

We have seen time and time again that technology can be a force for good with the right policy approach. Without a plan, it can lead to harm. Hopefully, policymakers and companies alike have learned from the lessons of the social media age to develop a nuanced approach to technology policies. We are now at a point where we have technology that can–if effectively harnessed–build a more human-centric policy discourse, and promote more inclusive societies and equitable prosperity. Technology policies, as such, will have to be iterative, coordinated, and multifaceted to both decrease harm and ensure sustainable growth. The premise and the promise of the early Internet era was democratization, participation, and openness. Let us apply those same principles now.