Deep Learning Indaba 2024:Day 3, A Day of Insights, Innovation, and Inspiration

deeplearning indaba
Author

Lukman Aliyu Jibril

Published

September 3, 2024

Attending the Deep Learning Indaba 2024 in Dakar has been a whirlwind of learning, networking, and inspiration. Day 3 was particularly packed with enriching activities and thought-provoking sessions that left me energized and inspired.

The day began with a compelling keynote by Samy Bengio on “Learning to Reason.” Samy delved deep into the challenges that transformer models still face when it comes to reasoning. While transformers have revolutionized natural language processing, their ability to reason like humans remains limited. Samy presented some innovative methods to enhance this reasoning ability, sparking a lively discussion among attendees. It was fascinating to see how research continues to push the boundaries of what AI can achieve.

After the keynote, I attended a hands-on practical session on Federated Learning. Having spent the last two days tutoring, it was refreshing to be on the learning side again. We explored how federated learning can play a crucial role in fostering responsible AI, particularly by ensuring data privacy and security. We also got hands-on with Flower, an open-source framework for building federated learning systems, and discussed the added complexities that arise when working with heterogeneous datasets. Thanks to my prior experience with the DeepLearning.AI Federated Learning short course, I could dive deeper into the challenges and opportunities in this space.

Next, I joined a career mentorship session with Sekou Remy from IBM. This was one of the highlights of my day, as Sekou shared invaluable advice on navigating one’s career in AI. Some key takeaways included:

This session was a reminder of the importance of being strategic, adaptable, and proactive in our career journeys.

After that, I attended a fascinating session on “Multimodal Perception for Human and Animal Behavior Understanding,” led by Marwa Mahmoud from the University of Glasgow’s Behavioural AI Laboratory. Marwa discussed her work on the early detection of neurodevelopmental conditions in children using multimodal features, highlighting the critical need for explainable and interpretable models. Given that these models are designed to assist healthcare providers, transparency in AI decisions is paramount. She also touched on the use of probabilistic models as a first line of support in decision-making.

What struck me was the complexity and subjectivity involved in applying AI to psychology and behavior analysis, where getting well-labeled datasets can be incredibly challenging. Marwa also provided a glimpse into her work on modeling animal behavior, which can be even more daunting than human behavior. She referenced her paper, “Going Deeper than Tracking: A Survey of Computer Vision-Based Recognition of Animal Pain and Affective States,” which further explores these complexities.

To wrap up the day, I joined discussions on community programs, featuring inspiring talks from various AI communities like GhanaAI, Masakhane, and others. They shared their initiatives and the tangible impacts they’ve made in their regions. It was a powerful reminder of the importance of community in driving AI forward, especially in underrepresented regions. The discussions were filled with passion and a shared vision of using AI for good.

Throughout the day, I also had the opportunity to network with fellow attendees, exchange ideas, and work on our hackathon submission. The energy and camaraderie in the room were palpable, and it was exciting to be part of a community so dedicated to pushing the boundaries of AI.

Day 3 at Deep Learning Indaba 2024 was truly a day of insights, innovation, and inspiration—a perfect blend of learning, sharing, and growing together. I am looking forward to what the rest of the week will bring!