As artificial intelligence (AI) technology continues to become integrated into our daily lives, concerns about bias are cropping up in increasing volumes, driven by stories of wrongful arrests and prosecutions, hiring discrimination, and diminished access to healthcare. While these discrete examples are the most attention-getting, according to a Deloitte expert, there’s more to making AI trustworthy than eliminating bias.
Beena Ammanath is a computer scientist who held data science and AI-related roles at General Electric, Bank of America, and Hewlett Packard before joining Deloitte in 2019, where she currently serves as the Executive Director of the Global Deloitte AI Institute and the head of Trustworthy AI and Ethical Tech. Her new book, Trustworthy AI, is an attempt to move the conversation beyond concerns about the very public examples of problems with AI to how AI can be developed with ethical considerations built into the very foundation of the technology.
Initially reluctant to get involved with the flurry of discussion surrounding the potential problems of AI, Ammanath realized that much of the rhetoric was focused on only one side of the story. While fairness and bias are certainly important in building trustworthy AI, safety, security, reliability, and robustness are also crucial factors that cannot be overlooked as the technology continues to evolve.
Additionally, how AI is utilized becomes increasingly important, as the technology can drive problematic outcomes – such as facial recognition AI misidentifying criminal suspects – as well as positive ones, such as the same technology identifying potential human trafficking victims. According to Ammanath, a key step to building trustworthy AI is building ethical considerations into any organization that builds and utilizes the technology, at every level of the organization. By recognizing and agreeing upon ethical principles and training stakeholders on their application, companies can start conversations about ethics early in the development or deployment process.
Ultimately, AI will only grow in use in our society, and with government regulation typically lagging behind new technology by decades, the most important conversations about its ethical use will be happening at the water coolers, boardrooms, and client meetings of companies who choose to build trust into their technology. Ammanath’s new book is an important part of starting those conversations now.