Scaling AI under the EU AI Act: catch up with our replay

AI is no less than reshaping the world. It has become a quintessential asset for competitiveness and there is a pressing need for businesses to fully comprehend and navigate its ethical and regulatory dimensions. Yet, it could well turn into a hazard for some if not meticulously crafted and deployed.

In our recent event in partnership with Vlerick Business School and the University of Antwerp, we tackled this exact challenge. We brought together experts to cut through the hype and dive deep into the foundational pillars of responsible artificial intelligence. From mitigating algorithmic bias and ensuring strict data privacy, to building transparent models that scale safely, this session is a must-watch for anyone looking to navigate the future of technology responsibly.

If you missed the live event, don't worry! You will find the full replay here. It will shed light on a simplified path to scale complex AI usage while keeping in line with ethical principles, regulations, and societal norms. The path optimises the potential of AI while curtailing its possible detrimental effects. We tackle areas such as compliance and responsibility, trust and robustness, as well as fairness and transparency.

Timecodes:

0:00 First Talk: The Need for Responsible AI 26:50 Q1/ Should we use tools (like SHAP or LIME) to evaluate why a model make a prediction? 29:25 Q2/ Spotting correlations between several factors can be tricky, making it difficult to be unbiased? 30:35 Q3/ What went wrong in your example on recidivism decision-making, biased by ethnicity? When could the bias be introduced in the data? 36:15 Q4/ Acceptable AI: how to, as a data scientist, build an unbiased data science model with biased data? 40:17 Q5/ If we keep adding risk levels and differencing different levels of risks, do we not give AI more paths to be unfair and make unreliable predictions? 42:20 Q6/ How to distinguish between the different levels of risks? 43:53 Q7/ Who defines what is responsible and what is not?

46:50 Second Talk: Bridging Ethical Governance & AI Efficiency 01:04:41 Q1/ There have been no published rules, according to the AI Act, how do we know if our approach is responsible? 01:05:43 Q2/ Can our ethics ever keep up with how rapidly changing AIs are? 01:07:39 Q3/ If a client wishes to be at the middle-risk level in the risk scoring, are we willing to be that flexible?