On the development of dependable intelligent systems

With the increasing presence of intelligent systems such as robots in our everyday lives, the risk of a failure in one of these machines causing injury to someone is growing. No longer can we rely on training and fenced-off work cells to provide safety. Instead, we must consider other approaches to producing dependable intelligent systems. This includes revising how we design intelligent systems and considering the adoption of certification schemes. In this article, we propose using a modeling language, dubbed SafeML, for the design of safety-conscious and dependable intelligent systems.