Guest speakers

Guest Speaker 1 - Karim Douieb on Data Visualization

1. Purpose of Data Visualization
Karim emphasized that data visualization should make complex information accessible, engaging, and meaningful. The goal isn’t just to display data, but to tell stories and make people understand key insights at a glance.

2. Importance of Storytelling in Data
He introduced the concept of scrollytelling, which is a storytelling technique used in online interactive visualizations. It guides the viewer step by step through the narrative while showing the data.

3. Notable Project – Equal Street Names
He talked about his project visualizing gender inequality in Brussels street names, showing that only 6% of streets were named after women. This project was powerful because it made an abstract societal issue tangible and easy to grasp.

4. Use of Simple Design and Interactivity
Karim stressed using minimal ink (referencing Tufte’s “data-ink” principle) and simple, clean design. He also highlighted how transitions between visuals can help users understand changes and relationships in the data more intuitively.

5. Technical Tools Are Not a Barrier
He pointed out that coding skills are not required to create effective data visualizations anymore — tools are becoming more user-friendly and accessible.

6. Real-World Example – Train Maintenance Dashboard
Karim worked on a dashboard using IoT data (e.g., GPS, RPM, temperature) for SNCB/NMBS trains to monitor and optimize maintenance. It was a great example of how real-time data can be visualized to support operational decisions.

7. Key Philosophy: "The Medium is the Message"
He reminded us that how we present the data can shape the message itself — visuals can influence perception and impact.

Guest Speaker 2 – Campfire on Conversational AI

What is Conversational AI?

• AI tech (like chatbots or virtual assistants) that simulates human conversation using Natural Language Processing (NLP).

• Intent-based AI uses confidence scores to choose pre-set responses.

• Generative AI (like ChatGPT) creates responses by predicting the next word/token.

Key Lessons

• AI ≠ Human Thinking: AI processes intent and language statistically, not logically like humans.

• Start Small: Begin with simple use cases (e.g., FAQ bots) to prove value before scaling.

• Hybrid AI is Best: Mix scripted (controlled) and generated (flexible) responses to balance reliability and adaptability.

Tech Approaches

• Scripted: Controlled flows using pre-written answers.

• Generated: Flexible, using large datasets and Gen AI.

• Hybrid: Combines both, ideal for scalability and control.

Practical Applications

• Customer support bots (e.g., Argenta’s chatbot “Charlie”).

• Voice assistants for medical workers.

• WhatsApp campaign bots increasing sales engagement and automation.

Real-World Considerations

• Good use cases are frequent, recognizable, and answerable.

• Implementing AI involves message routing, data integration, and feedback loops.

• Ensure privacy and data control—Campfire emphasizes anonymization and security.

Gen AI Challenges

• Hallucinations (made-up responses).

• High computational cost (like “using a Lamborghini to deliver pizza”).

Guest lecture 3 - Sylvie Delberghe on Human side of AI

1. AI Readiness Is More Than Just Models

• Successful AI adoption requires solid data governance, strategy alignment, clear use cases, and human oversight.

• Only a minority of companies fully integrate AI into operations.

2. Data Quality Is Fundamental

• “Garbage in, garbage out” — AI depends on accurate, complete, and unbiased data.

• Humans play a key role in preparing data, checking bias, and interpreting results.

3. AI Must Solve Real Problems

• Start with clearly defined use cases and evaluation criteria.

• Involve employees early for analysis, prototyping, and feedback.

• Avoid “AI for the sake of AI.”

4. Build on Strong Technological and Ethical Foundations

• Choose the right tech and integrate it into workflows.

• Ensure ethical, responsible use with defined ownership and governance.

5. Trust & Accountability Are Crucial

• Each project needs a sponsor, clear roles, and ownership of AI outputs.

• Labeling and transparency support trust and adoption.

6. Retrieval-Augmented Generation (RAG)

• Retrieval: looks up relevant

  information from a large database to inform its response

• Generation: creates coherent and relevant text

7. Human-Centered Implementation

• AI doesn't remove personal accountability.

• Use cases in health insurance improved speed, reliability, and information access — but required governance and maintenance.

8. Transformation Comes in Stages

• From “Initial” (unmanaged use) → “Transformational” (AI as the norm).

• AI should drive value, not just automation.

9. Ethical Impact & Purpose

• AI should serve people, planet, purpose, and prosperity — not just efficiency.

10. New Skills Are Needed

• AI literacy, critical thinking, transparency, and data fluency.

• “AI doesn’t change organizations. People do — with the help of AI.”

-