Gabriella Chiarenza is managing editor for Invested and Regional and Community Outreach at the Federal Reserve Bank of Boston.

Click play above to hear from some of those working most closely with AI in their own words about the capabilities of today’s and tomorrow’s robots, the impact of technology on workers and businesses past and present, and the challenge and promise of developing thoughtful AI-augmented assistive technology for the workplace. The audio program—which you can tune in to as you would a podcast—features an in-depth discussion of the ideas summarized below, and interviews with Daron Acemoglu, Professor of Economics at MIT; Morgan Frank, Post-doctoral Associate at MIT; Sean Murray, Director of Robotics Engineering at Realtime Robotics; Julie Shah, Associate Professor and Director of the Integrative Robotics Group at MIT; and Stefanie Tellex, Associate Professor and Director of the Humans to Robots Lab at Brown University and consultant with Realtime Robotics. To view the audio transcript, please click here.

Every now and then, a technology comes along that changes everything. These are known as general purpose technologies (GPTs), because they change many aspects of our lives in different ways, spur new innovations that flow from them, and make us more productive and prosperous. But the impacts of GPTs take time to be seen, and they don’t flow equally to every segment of society, especially without complementary investments that help the power of a GPT’s benefits reach a broader range of communities. While it took several decades for AI to get off the ground in a real way, those who know this technology best now observe it becoming a GPT. Having seen this to be the case in other eras of industrial revolution, such as when electricity or railroads fueled decades of enormous economic and societal change, how can we be best prepared for the arrival of a GPT like AI today, and what can we do to help the benefits AI might generate stream more quickly and effectively to more people across industries and incomes? The answer may lie in a thoughtful approach to a process called augmentation.

AI at its base is software that enables machines to handle tasks that would otherwise require human intelligence to complete. AI software can be used to enhance robotics to create “intelligent” machines, but not all robots are programmed with AI. AI augmentation may include automation and/or collaboration between technology and humans on the various tasks that make up work processes. Depending on the need, the immediate goals of augmentation might include making a process more efficient by automating highly repetitive tasks so that a human worker can concentrate on aspects that are more unpredictable, or that machines cannot yet handle. Or, the objective might be to make a process safer, more ergonomic, or more interesting for the human team working alongside AI and/or robot teammates. But because AI is a GPT, its augmentation powers go beyond taking over tasks. For example, AI can process a huge amount of data in a split second and draw out relevant data points on demand for humans to use in many ways.

Consider a few examples of potential AI-based worker augmentation. AI could be used to shave vital seconds off a factory-floor process by providing the right materials at the right time to the right human teammate. Or it could draw on hundreds of complex data points to help a teacher develop an ideal lesson plan and customized learning approach for each individual student, to optimize curriculum absorption and comprehension. Or it could assume some of the paperwork and data-processing burden of busy health care professionals, freeing them up to focus on unpredictable situations like emergencies and on the interactive human aspects of their job.

Developing AI that augments humans certainly involves considerable technological expertise. But taking it a step further to developing equitable, effective, and widely beneficial AI augmentation also requires an interdisciplinary approach drawing on input from many different stakeholders: human workers in a wide range of fields, experts on societal norms and values, policy and decision makers, business leaders, and community advocates and practitioners representing the full spectrum of our population. Drawing true prosperity for a wide swath of Americans out of AI’s capabilities calls for extensive collaboration, investment, and thoughtful development.

On the technological side, successful augmentation faces several challenges. Take the case of AI embedded in augmentation-oriented robots: the robot must be safe for humans to work around and have the autonomy to handle physical challenges like gripping, selecting items, and moving through spaces that may include obstacles. These are very simple things for humans to do, but for robots, they remain extremely challenging, though a lot of progress is being made in improving these capabilities. Even more challenging, robots and other forms of AI must have a basic understanding of the humans working alongside them, including those humans’ roles in their shared work, be able to communicate reasonably well, and have enough intelligence to anticipate what their other AI and human teammates might do or need.

We are quite far from robots being able to do all of these things independently, but over time, AI is becoming more and more agile at working with us in helpful ways. But robots, in particular, that have this level of spatial and logical awareness require special programming, additional components such as cameras and sensors, and skillful maintenance—all of which can come at a significant cost for a business hoping to take advantage of the technology and for workers who must learn how to work with and maintain such machines.

Roboticists are working to make the technology more responsive, flexible, and intuitive to help reduce the burden of adapting to a new machine, but it still requires that the robots’ human teammates learn how to take care of, communicate and work with the technology. Because AI technology is continuously evolving and improving, workers and firms will need to be receptive toward change. They must be agile and resilient if they are to draw the greatest benefit from the technology with the least negative disruption to workers’ lives and business flows.

On the societal, economic, and political sides, the challenges are just as great. At the moment, most AI development and testing is confined to universities, military operations, and start-up hubs in places like Silicon Valley and Boston, as well as in some manufacturing domains. If left to develop in ways that are simply the most profitable for companies or politically advantageous for those in power, AI could  contribute to greater widening of the income and wealth gaps, with little positive impact on or access for a broader range of society. If, however, AI were to be further developed with public and expert input from a broad range of stakeholders and with investment and policy goals that focus on improving the standard of living and the broader economy for everyone, this GPT’s way forward could be much more beneficial.

Those who know AI best largely believe we have reached the point where we as a society must make some decisions about which of these paths we wish to take. Taking the more equitable, beneficial path requires public understanding and contribution of ideas, investment that prioritizes societal benefit, and policy that incentivizes further AI innovation that supports the public good as well as business and economic advancement. At the moment, these conversations are very few and far between. The stakeholders we spoke with for this issue of Invested make it clear that we need to change that, and soon, if we want AI to reach its full beneficial potential for more Americans.

The views expressed are not necessarily those of the Federal Reserve Bank of Boston or the Federal Reserve System. Information about organizations, programs, and events is strictly informational and not an endorsement.