Humans And AI in Manufacturing

28 September 2021

The publication of this article has been kindly granted by STAR-AI, an ICT-38 Project Cluster partner.

The debate about the risks associated with deploying AI in industrial workplaces deserves the attention it gets and even more. But it often masks another fundamental one: how human-centric AI can deliver high adding value jobs?

AI-enabled solutions radically reshape the “affordances” of the interaction between human and non-human actors. Affordances are action possibilities provided to an actor by its environment. AI upscales the affordances of both human and non-human (e.g. physical, cyber-physical, or digital entities) actors, operating in production environments. But AI-enabled systems can also have vastly superior affordances when effectively integrating the human in the loop.

Despite massive leaps in the advancement of AI, the superiority of human cognitive capabilities over AI in performing across domains is undisputed. The deficiency of AI-enabled systems in “cognitive switching” can be partly mitigated by integrating the human cognitive capabilities in the AI loop. This goes beyond “transfer learning”, aiming to transfer the learned AI capabilities from one original domain to a new one; and it can build on “active learning” approaches, wherein there is an evolving interaction between AI-agents and humans. So why not seek to benefit from integrating rather than eliminating the human from the AI loop? This is not about the human in the AI loop just for the sake of doing so, but for its added value.