In the factory, the key focus will be on physical AI

However, physical AI brings not only innovation, but also new questions. Indeed, as long as the machine merely displayed data, the decision always rested with the operator. Now that the machine decides and acts, we find ourselves having to place our trust in something we do not yet fully understand. It is a new, different kind of trust: not in a colleague, but in a system that makes choices on our behalf. This raises some very practical questions: who is accountable for a decision made by the algorithm? A machine that has been substantially modified after installation must be recertified – but by whom? These are issues that the European framework is beginning to address.

And it is precisely trust that distinguishes initial enthusiasm — the so-called ‘hype’ — from a technology that is here to stay. This is confirmed, from a different perspective, by Luca Zanella, Global e-commerce Channel Manager at ABB’s Electrification division. ‘A technology remains hype as long as it amazes without solving any problems. A robot dancing in a hotel reception is of little use,’ says Zanella. In his experience, a technology becomes a genuine market when it meets a real need and an ecosystem of services — configurators, platforms, support — grows up around the product, making it usable, integrable and sustainable over time. It is at that point, says Zanella, that a technology stops making headlines and begins to generate value: “When it produces a repeatable economic return, not just an isolated demonstration.”

Trust involves another key element: data. When a product becomes connected, the relationship between the technology provider and the customer changes in nature: the provider begins to ‘see’ the data of those using the machine. Zanella interprets this in terms of shared growth: ‘Asking for data not to take it away from the customer, but to grow together.’ This approach is a winning one from a commercial perspective. However, from a governance perspective, data sharing must be carefully structured, particularly when it comes to physical AI. The issues to be addressed are the basis on which that data is collected, for what purposes, and who is permitted to do what with it. Prudent management of these issues transforms a ‘constraint’ into a genuine partnership and confirms that the real barrier, once again, is not technological, but mutual trust.

So when does a factory become ‘smart’? Not when it has a robot that mimics humans, but when both artificial intelligence and machinery support the operators. When they learn to manage the variability of the real world as effectively as possible. When that intelligence demonstrates reliability and trustworthiness in all situations. Despite the technological acceleration we are currently experiencing, both Taisch and Zanella nevertheless urge a pragmatic approach. This journey has only just begun, and it is up to us to determine the return on investment, to understand which decisions to delegate, and to safeguard those areas where responsibility must remain with humans. In short, to shape it in a way that is entirely to our benefit and not merely futuristic.