The role of technology in complying with the AI Act.

Blog:

The AI Act is the new European law that sets rules for the use of AI. The aim of the European Commission is to ensure that citizens and organizations can trust that AI systems are safe, transparent and environmentally friendly. And that decisions made by AI models are traceable and non-discriminatory. Although the impact of the law is still unclear and even though each country still has its own regulator with its own opinion and therefore its own interpretation, Valentin Calomme, strategy consultant at Conclusion AI360, is enthusiastic nonetheless. "It's the first form of regulation that focuses on the entire AI chain: how is it built, how is it controlled, how is it used, who does it impact? The EU is very clear: when working with AI you must put people first. And these are the values and standards that you must adhere to. "I also endorse the fact that the AI Act takes climate and environmental impact into account."

The fact is that legislation always lags behind technology. Still, Valentin doesn't want to make too big a deal out of it. "Of course, the technology is far from fully developed and so the law will continue to develop with it. But I think the AI Act coming into effect now is a good thing, at a time when scaling up is the order of the day. When the GDPR was introduced, we witnessed the difficulty of introducing new rules in a domain that everyone had been dealing with for years. The AI Act changes this. While AI itself is not new, it has been used and developed on a large scale in only the last few years and by only few companies. Most organizations are just starting out. They are now forced by law to pay attention to responsible AI from the very start. You can no longer ignore the subject."

Sustainable value creation and responsible AI are inextricably linked. It is a broad theme that Conclusion has divided into six pillars. In order to monitor whether these six pillars are being met correctly, Conclusion developed the Responsible AI Framework. Not only does this framework assess code or mathematics, but also at the human components that are a core part of what makes technology deliver sustainable value. "After all, technology is more than code and mathematics, it also involves things like policy, governance and organizing human involvement," Valentin explains. 

The fact that the field of AI is still relatively young compared to a field such as traditional software development only increases the challenges, Valentin believes. "It's missing about 20 years of trial and error when it comes to figuring out how to build things right. Just think of automated testing, continuous integration and deployment, etc. AI uses many lessons learned from traditional software development, but it remains a different field with its own challenges, requiring its own solutions. While the field is developing rapidly, norms and standards are lagging behind."

He cites the example of monitoring energy consumption by AI. "If you're not careful, you'll be developing energy-guzzling applications. It would be great if a plug & play monitoring tool was available that you could use right off the shelf; a tool that uses a standard that everyone embraces. Now we still have to develop our own monitoring system and implement and manage it." Something that takes a lot of time but also adds value, because monitoring the ecological impact of AI models allows you to come up with more creative solutions, as Conclusion has already experienced with several customers.