The AI Dilemma: 7 Principles for Responsible Technology

AI holds unbelievable promise to fortify virtually every aspect of our lives, but we will be able to’t ignore its risks, mishaps and misuses. Juliette Powell and Art Kleiner offer seven principles for ensuring that machine learning supports human flourishing. They draw on Powell’s research at Columbia University and use a wealth of real-world examples.

Four principles relate to AI systems themselves. Human risk will have to be rigorously made up our minds and consciously included in any design process. AI systems will have to be understandable and transparent to any observer, not just the engineers working on them. People will have to be allowed to give protection to and manage their personal data. The biases embedded in AI will have to be confronted and reduced.

The final three principles pertain to the organizations that create AI systems. There will have to be procedures in place to hold them accountable for negative consequences. Organizations want to be loosely structured so that problems in one area can also be isolated and resolved before they spread and sabotage the whole system. In the end, there will have to be psychological safety and creative friction, so that any one involved in software development can bring problems to light without fear of reprisal.

Powell and Kleiner explore how to implement each principle, citing current best practices, promising new developments, and sobering cautionary tales. Incorporating the perspectives of engineers, businesspeople, government officials, and social activists, this book will help us realize the unprecedented benefits and opportunities AI systems can provide.