When You Need the AI to Explain Itself Before You Can Act…
12 Sep 2019
Innovation Hub Theatre
Commercial AI has the advantage of large amounts of quantitative historic data (Amazon builds its algorithms based on over 1bn transactions with over 180million users per year), and decisions made using commercial AI rarely have severe consequences if a mistake is made. In defence, particularly for command and control (C2), historic data is either sparse or largely qualitative. For instance, there are fewer than a dozen instances of anti-ship missiles being fired at air-defence capable platforms, but hundreds of trained air-warfare offices with experiences of deployments and exercises. DIEM has developed a range of Augmented Intelligence applications (the other AI) for a range of domains that either use narrative training data, or generate military-language explanations of the outputs of other neural-net and deep-reinforcement learning based AI algorithms. In this presentation we shall describe a range of innovative AI application DIEM has developed internally and under Dstl and DASA innovation funding.