AI Ethics: A View Into the Economics of AI

We often think about artificial intelligence (AI) ethics from the perspective of the technology sector. There’s nothing wrong with that; at Genesys, we encourage discussions on our AI ethics guidelines. But it’s important to seek new perspectives outside the technology-only mindset to gain a full understanding of where AI is headed — and how it will affect workers.

In the book, Prediction Machines: The Simple Economics of Artificial Intelligence, authors Ajay Agrawal, Joshua Gans and Avi Goldfarb, look at AI from an economist’s perspective. Following is my review of the book — a great read to gain an outlook on the future of AI in the workforce.

“If economists are good at one thing, it is cutting through the hype,” notes the authors, adding that prediction technology is becoming a commodity because AI makes it less expensive. And, as costs for this technology decrease, our uses for it change and advance. We use prediction capabilities daily to improve the effectiveness of certain applications. Services like Waze, Lyft and Uber all use prediction solutions to find the fastest route from one location to another — considering traffic patterns and alternative routes in real time. We don’t need someone tied to a map constantly recommending new routes based on knowledge of that area or periodical calculation of the best route.

But as predictive technologies within these services improve, workers’ knowledge for such things can actually decrease. For example, London cab drivers possess a wealth of knowledge and can navigate the city with remarkable familiarity. Before being licensed these drivers must pass a daunting test titled “The Knowledge” to prove they possess an expert-level understanding of locations, streets and how to travel the city in various situations without consulting any kind of map. The test is very difficult, and it takes, on average, three years of studying to pass. But today, we have GPS.

And again, services like Waze, Lyft and Uber turn the ordinary driver into an expert London cab driver — without the need for extensive studying. And it gives them real-time traffic data that no human could possibly have. The authors note these original London cabbies aren’t suddenly worse at their jobs; however, millions of other regular cab drivers became better at their jobs because of cell phones and an app.

Locating the AI and Human Divide

The book highlights two major trends: where human input will still be essential and which part of a job AI technology will overtake. The authors break down jobs into tasks. If you were to think of a job as a house made out of Legos, then the tasks of a job are the individual pieces that comprise the house. While the Lego house itself is still a house, some of the pieces within it can change. Jobs won’t disappear completely, but inexpensive prediction technologies will affect certain tasks performed within that role.

In the 1980s, banks unveiled the widespread rollout of the automatic teller machine (ATM). With a machine titled explicitly to automate the teller or human employee, it wouldn’t be far fetched to think that banks would employ fewer workers at physical branches. But the exact opposite happened.

The ATM automated routine teller tasks, but their roles and numbers actually grew as time went on. Instead, these tellers took on different tasks, such as loan advisement, credit card inquiries and other complex jobs in which customers wanted human engagement. Certain tasks like bank telling were easily automated while others grew in value as a result of automation. The idea that humans have more bandwidth to focus on tasks that are innately human is exciting. But the future won’t always have this positive tilt.

The book points to warehouse workers who have been replaced by robots to eliminate costs and safety restrictions, such as heating and cooling costs, wide lanes requirements between aisles for workers and lighting needs. The robots are mostly autonomous, navigating through the warehouse with ease. But they struggle to pick up packages because of limitations in machine vision technology. Unlike robots, humans are uniquely capable of knowing the best angle and amount of pressure to use when picking up packages.

To solve this, the warehouse workers remotely control the machines, which adds the human touch. But over time, the machines will improve and learn; every package pickup executed by a human is a learning moment for a machine. At some point, the robots will handle the entire job.

Get a Vision Into the Future of AI

Prediction machines look at AI from a perspective that isn’t born in technical understanding. As a former support engineer, I tend to fawn over new technology and technical architecture that can overshadow that most vital point — purpose.

Prediction can supercharge the application of technology and free up workers to focus on being people — not resources. But we can’t operate under the notion that there won’t be redundancy and negative consequences to the workforce as these technologies become even more affordable.

The authors tackle this subject with honesty and make useful recommendations for those in leadership positions. This book is an important read for those wanting to gain a higher-level view of the effects of AI on the workforce. Its combination of real-world examples, unique perspectives, and future recommendations are essential to gaining a better understanding of AI.

Join the discussion on AI ethics.