Tech Talk | AI challenges in Europe

AI is attracting headlines daily but two issues are currently top of mind in Europe – the imminent AI Act and the energy consumption of AI-focused data centres.

In a session at EU Sustainable Energy Week, the role of AI as an “enabler” in the energy sector was highlighted by speaker after speaker, with numerous examples cited, such as helping to optimise grid operations, identify faults and streamline maintenance procedures.

But even if it may not directly impact many users, the coming into force of the EU’s Artificial Intelligence Act in July and its various provisions over the following 36 months is something that all AI users need to be aware of.

Lucilla Sioli, Director for AI and Digital Industry in DG Connect – and by the time of this publication head of the new EU AI Office, which will be responsible for implementing the Act as well as other activities such as supporting research and innovation – said AI is a “force for good, but at the same time a bit of a black box” and the aim is to ensure it works well without violating fundamental rights and safety.

Have you read?
Endesa harnesses AI and 5G for smart grid worker safety
Accelerating the energy transition with Artificial Intelligence

And it is this the AI Act seeks to address by harmonising rules in Europe on the one hand and proposing a mechanism to ensure that the rights are not violated.

To this end, several categories of risk are defined – unacceptable, high, limited, minimal – each with its own rules.

Sioli cited an example of ‘unacceptable’ as the use of AI by governments or enterprises for social scoring, while the use of AI for recruitment, with its risk of discrimination or bias, is an example of the ‘high’ category.

The management and operation of critical infrastructure also has been identified as a high-risk area. In this case, all AI systems will require assessment before being put on the market and throughout their lifecycle.

People also will have the right to file complaints about such AI systems to their designated national authority.

As an example Sioli mentioned the AI systems used as safety components in the supply of electricity or gas and in this case companies will need to ensure their AI system is certified before being deployed in the market.

This certification will be based on the existing ‘CE’ marking approach and should demonstrate that the necessary risk mitigation measures have been taken.

“It’s important that there is checking about the data that is used, the design and parameters of the particular algorithm, its cybersecurity and the robustness and accuracy of the system itself,” she said.

However, probably most applications in the energy sector, such as the management of smart grids or optimisation of storage, are unlikely to be subject to the rules, she added.

Turning to generative AI, for which a specific set of rules have been developed, Sioli said she expects the technology to be very useful but that there is a need for European models trained with EU data.

To this end, the AI office is putting in place a research and innovation programme to support startups with the supercomputer network to be made available for the model’s training.

In conclusion, Sioli commented that one of the functions of the AI office is to support the integration of generative AI models into final applications, such as those of the energy sector, which should help to contain its energy consumption.

More news from European Sustainable Energy Week:
Three dimensions to finance the EU energy transition
Europe’s ‘Destination Earth’ to deliver climate data for energy planning

AI data centre challenges

Turning to the challenges of the energy consumption of data centres, Ditte Juul Jørgensen, Director-General for Energy in the EC, said the energy use of data centres is critical to address with the amount of energy required both for data processing and cooling for IT in general and AI in particular.

“We risk losing a lot of energy through the systems that are going to save us a lot of energy so to avoid a zero sum game, we need to make sure we organise them more efficiently,” she said, pointing to actions such as the development of a common energy space to integrate data.

A new reporting obligation for data centres also has been put in place with the aim to build a clear picture of data centre energy consumption and potential efficiency solutions around their planning and placement.

An example would be siting a data centre close to a district heating system into which the waste heat could be delivered for residential or industrial heating.

Paula Pinho, Director in DG Energy, also revealed “an EC secret” that in preparing the revised energy efficiency directive consideration was given to establishing data centre efficiency thresholds but at the time that was not considered possible due to the lack of consumption data.

“But once the data is available, these will come,” she warned.

Jonathan Spencer Jones

Specialist writer
Smart Energy International

Follow me on LinkedIn