24th Belgium NLP Meetup

There is no such thing as a slow news season in NLP. People who follow the LLM leaderboards as closely as the Olympic medal table will agree that it's time for a new NLP Meetup. Therefore I'm happy to announce our 24th edition will take place on Tuesday September 24th, at the offices of data powerhouse Collibra in Brussels.

We'll zoom in on three hot issues in NLP. Ben Burtenshaw will explore the evolving role of human feedback collection in AI projects, while Evelien Schellekens will discuss LLM Observability. Finally, Alexandre t'Kint will introduce a four-step framework to ensure AI systems are compliant, transparent and trustworthy.

As always, doors open at 7pm for pizza and drinks, kindly sponsored by Collibra. The first talk starts around 7.30pm, and after 9pm there's time for drinks and networking. I hope to see you all there!

The Annotator is Dead, Long Live the Annotator
Ben Burtenshaw (Argilla)
In this talk, we will explore the evolving role of human feedback collection in AI projects and how it is being transformed by the integration of synthetic data and generative models. We'll discuss the importance of prioritizing data quality over quantity, and how focusing on relevance can enhance model performance. We will cover practical methods for incorporating human feedback into synthetic datasets, comparing the strengths and weaknesses of human versus synthetic data preferences. Additionally, we'll show examples of community-driven datasets that leverage collective expertise to build great datasets.

UnLLMited Insights: Observing GenAI Applications
Evelien Schellekens (Elastic)
In this talk, we'll dive into the world of LLM Observability and explore why it's essential for understanding and optimizing the performance of GenAI applications. We'll explore the growing range of OpenTelemetry-based tools designed for these systems, comparing their differences and advantages. We'll also cover OpenTelemetry and its benefits for unified, flexible monitoring. To wrap up, we'll do a demo showing these concepts in action using Elastic.

Mastering AI Governance: A Practical Guide to Trusted AI
Alexandre t'Kint (Collibra)
With the recent implementation of the EU AI Act, ensuring your AI systems are compliant, transparent, and trustworthy is more critical than ever. This session will explore the importance of AI governance and provide a practical four-step framework to help you navigate these complexities:
1. Define the Use Case: Clearly outline the specific problem your AI aims to solve.
2. Identify and Understand Data: Ensure your data is relevant, unbiased, and compliant with regulations.
3. Document Models and Results: Maintain transparent records of AI models and their outcomes to ensure traceability and accountability.
4. Verify and Monitor: Continuously check and adjust AI applications for accuracy, fairness, and compliance.
Implementing this framework will help your organization stay ahead of regulations like the EU AI Act, avoiding legal consequences and reputational damage. Effective AI governance fosters innovation, enhances decision-making, and drives business success.