After her speech at the 2024 Go Digital Energy Summit in Amsterdam, the Go Digital Energy team had a chance to ask Iulia Maes, Senior Account Executive at Cognite, about her opinion on the reliability and transparency of AI systems.

Source: Go Digital Energy

How can industries ensure reliability in their AI systems?

• Ensure data is correctly contextualized and leveraged.
• Provide users with information about data sources.
• Allow users to give feedback and control over systems.
• Address risk aversion and change management challenges.
• Emphasize transparency of data to build trust.

The steps to gain this start with ensuring that the data on which the outputs are based is correctly contextualized and correctly leveraged. It's also very important to provide the user with information about the data source or document on which the response is based. For example, ChatGPT has started providing the website from which data is coming. Similarly, in the industry, it would be valuable for users to see the data source.

Additionally, giving users the ability to provide feedback and some control over the systems is crucial. For instance, when users see the generative AI output, they should be able to give feedback on its accuracy or suggest different classifications. This builds trust in the output and makes users feel valued in the process, countering the risk of the process becoming too automatic and not human-centred. The importance of human-centred AI was a significant topic in yesterday's sessions, and it is seen as key to success in various industries.

Risk aversion is another critical issue, as people are very cautious about outputs from generative AI, especially when safety is involved. High stakes make people reluctant to adopt new technologies. Change management is also a real challenge, as people have established workflows and may find it difficult to adapt to new ones. Trust in the output is a recurring theme, with many strategies discussed at the conference.