Human in the Machine and the Challenge of Keeping AI Under Control
In the episode ‘Human in the Machine’, VPRO Tegenlicht highlights a fundamental shift. According to the program, software is no longer written but raised through testing. Artificial neural networks are inspired by the human brain and learn complex tasks through trial and error, similar to how children learn.
According to the episode, the problem with controlling AI arises when a smart machine is given a command. This is not because the machine malfunctions, but because we are poor at describing exactly what we want. Without something like common sense, a super-intelligent machine can misinterpret a simple instruction, with catastrophic consequences.
The question, therefore, is how machines learn to understand who we are and what our norms and values are. In this way, they can contribute to a society as we envision it. The broadcast features, among others, Stuart Russell, Professor of Computer Science at UC Berkeley and founder of the Center for Human-Compatible AI, and Catelijne Muller, EESC rapporteur on AI and founder of ALLAI Netherlands.
According to Tegenlicht, the industry can now profit from big data collections, while scientists, civil society organisations, and the European Union are hastily developing ethical guidelines. The goal is to maintain control over the profound changes awaiting society. Therefore, evaluate AI in your organisation not only on functionality but also on whether the system truly does what you intend.
Read the tegenlicht.vpro.nl article for more information.