Chapter 6. Governing AI
Technology is morally neutral until we apply it.
William Gibson
AI-powered has become the most popular adjective across organizations and products. Much of that is pure hype, certainly, but it is true that more companies are indeed trying to leverage their data and build something useful using the latest AI advancements. Just like the digital transformation, the AI transformation is real, and it is happening.
The biggest challenge, however, is to build an AI-based system that is fully governed and secure. You might have experienced this firsthand: AI projects, even simple, classic ML-based ones, can be difficult to productionize. That’s often because the typical origins of such projects trace back to the laptops of data scientists or ML/AI engineers, who are normally more focused on mathematical and statistical aspects of the project and less focused on how to deploy them. In most AI projects, security is typically an afterthought, and governance aspects are nonexistent. Moreover, with the advancements in generative AI (GenAI) offerings, notably GPT models from OpenAI, new security and governance challenges have emerged.
Some of these challenges seem comical. On January 18, 2024, a user on X (formerly Twitter) posted his conversation with an AI-powered customer service chatbot from parcel delivery company DPD. The chatbot, in addition to not being as helpful as the user expected it to be, when prompted, generated a poem on how useless it is and how terrible ...
Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.
Read now
Unlock full access