When Machines Learn: Understanding How and Why AI Works

What is AI really, and broadly, how does it work? What role does machine learning play and are there limitations to its capabilities? In fact, why do we even have AI? Without going into technical depths, Maarten Lamers discusses the why and how of AI.

Learning unlocks intelligence. And for long, learning was considered uniquely human, or animal, just as creativity. But machine learning artificial intelligence enables machines to create pictures more realistically than we can, to compose songs in mere seconds, and to write texts about any topic imaginable.

So, what is AI really, and broadly, how does it work? What role does machine learning play and are there limitations to its capabilities? In fact, why do we even have AI? Without going into technical depths, Maarten discusses the why and how of machine learning AI, and what it generally can and cannot do.

Maarten Lamers was elected by Vrij Nederland (2015) as one of the "most inspiring technologists, inventors and tinkerers in The Netherlands and his research typically approaches academia in a playful manner, both in subject and in method. 

He believes that playfulness can be a valuable asset in academic work. Maarten is Assistant Professor and PhD at Leiden University and has a Master's degree in computer science from Utrecht University. ". He also gave a TEDx talk entitled "Academic Freedom for the Young" in which he argues that academic playfulness should be a student's right.

Summery

Join Martin Lammers, Assistant Professor and PhD from the University of Leiden, in this comprehensive webinar exploring the fundamental principles of artificial intelligence. Rather than focusing solely on AI's societal consequences, this presentation delves into the essential question: why do we have AI and how does it actually work?

Lammers begins by distinguishing between classical algorithmic machines and modern machine learning systems, explaining how AI emerged to solve problems that are simple for humans but impossible to formalize into traditional algorithms.

Through engaging thought experiments and real-world examples, he demonstrates how machine learning systems create latent spaces—imaginary multidimensional environments where data points gain semantic meaning.

The webinar progresses from foundational concepts like algorithms and embeddings to more complex discussions about pattern recognition, bias in AI systems, and the famous fish classification study that reveals how machines learn unexpected patterns from data.

Lammers also addresses a critical concern for many professionals: can machines be creative? Drawing on Margaret Boden's dimensions of creativity, he explores exploratory creativity (like Bob Ross's paintings) and transformational creativity (like Picasso's artistic evolution), arguing that while current AI exhibits exploratory creativity, transformational creativity remains an open question.

The presentation concludes with insights into model collapse, the dangers of training AI on AI-generated data, and predictions about the future value of human-generated data. This webinar is essential viewing for anyone seeking to understand AI beyond surface-level applications.