Enterprises need to show their innovative side

Markt&Technik: The energy crisis will probably still be on our minds in 2023. AI requires a lot of computing resources, which means a lot of energy. How can embedded AI be implemented in an energy-efficient way?

Viacheslav Gromov: The image of AI consuming a lot of energy is based on the use of large servers, the cloud or local PC systems. However, this is not the case with embedded AI systems. By their very nature, they are based on few resources – starting with computing power and memory and ending with power consumption. In computer technology, energy efficiency is measured in terms of computing power, i.e. instruction set and parallelization per clock cycle in relation to power consumption. Energy efficiency will continue to improve in 2023 due to the constant optimization of semiconductors with neural processing units (NPUs). In addition, there is the trend towards decentralization: the use of many small decentralized and intelligent components instead of centralized processing will increase.

M&T: How can AI finally make the leap from research to application across the board? What steps do German companies need to take to achieve this?

VG: The transfer of research in Germany is a problem. Nevertheless, there is in-house research and pre-development that is pursuing AI topics in the long term. On the one hand, the will to implement must be there. On the other hand, the pressure of the crisis also seems to be a factor – as we are currently seeing across all industries. When the pressure increases, research is given a higher priority again, while other market players fall into a fear-filled lethargy. I see a split mentality in German companies. However, many in this country still have the innovation gene, which they now have to play to on a grand scale.

M&T: It can currently be seen that Hardware as a Service (HaaS) is becoming increasingly important. What do you understand by this and in which areas will HaaS be used?

VG: The first companies are already using HaaS. Due to global system competition – especially in the crisis – manufacturers are no longer winning in direct unit price competition due to inflation and supply chain problems. For this reason, they are increasingly focusing on customer orientation and service: leasing, subscription and full-service business models are a constant source of revenue with low customer entry barriers. Predictive maintenance integrated into the product is therefore very popular due to its low one-off costs instead of the cloud’s ongoing costs. This helps to prevent predictable service downtime and at the same time reduces the need for regular servicing or even eliminates it altogether. The areas of mechanical engineering, medicine and automotive are predestined for HaaS, especially when it comes to margin-sensitive or particularly price-intensive product areas.

Read the full interview here