AI Prototyping

This chapter discusses the process of AI prototyping, emphasising the importance of creating the first AI build as a key milestone in any AI program. It outlines three main approaches to creating AI solutions: using off-the-shelf AI software, exploiting capabilities from an AI platform, or creating a bespoke AI build. The document also explains different strategies for initial builds, including Software Trial, Proof of Concept (PoC), Prototyping, Minimum Viable Product (MVP), Riskiest Assumption Test (RAT), and Pilot. Each approach has its own objectives and considerations. The chapter highlights the crucial role of data in AI prototyping, stressing the need for quality, cleanliness, and accuracy. It also discusses the use of crowdsourcing for data tagging and the innovative approach of using virtual worlds or computer games for data acquisition and training. The author concludes by emphasising the need to understand all the risks that AI can raise and how to mitigate these for long-term success.