Old Wilsonian delivers final Maths Circle Lecture of 2026
We recently hosted a mathematics lecture titled “Stochastic Parrots and Bitter Lessons: A Tale of Machines and Language,” delivered by Shavindra Jayasekera - an Old Wilsonian (Class of 2018) who is currently a third-year PhD student at Imperial College London, specialising in Statistics and Machine Learning. During the talk, Shavindra explored key questions about artificial intelligence, including how machine learning works, whether machines can ever match human performance, and how tokenisation allows Large Language Models (LLMs) to function.
The lecture introduced the concept of deep neural networks, which are made up of multiple layers that data passes through, with each layer performing its own calculations. At every layer, the network has numerous parameters - called weights and biases - that can be adjusted to help the model better fit the data. By fine-tuning these parameters, an AI system learns to recognise and replicate patterns and trends in the data it was trained on. Shavindra explained that functions are applied to matrices to adjust these weights and biases, allowing the model to capture complex relationships and improve its predictions over time.
To make these ideas more concrete, Shavindra shared an example of an early language translation algorithm. This case study highlighted both the strengths and limitations of AI showing us that when trained on enough data, and when that data is used effectively, AI can outperform humans in specific tasks.
The lecture then moved on to Large Language Models and the role of tokenisation. Shavindra explained that sentences are broken down into words or subwords, with each segment being assigned a unique token. The model then predicts which tokens are most likely to come next in the response based on the patterns it has learned from its training data and matching these to the tokens in the prompt. To add an element of randomness and creativity to the output, these probabilities are expressed on a logarithmic scale, allowing the model to introduce some variation while still favouring the most likely response. This process of analysing probabilities and specific words allows LLMs to give the impression of understanding what has been said to them and generate relevant responses.
Overall, the lecture was both fascinating and thought-provoking, introducing students to the mathematical intricacies of AI algorithms and providing insight into recent developments in the field. Shavindra’s examples and detailed explanations made the topic much more accessible, leaving us with a better understanding of how the AI around us works. The session sparked curiosity and discussion, giving students a new appreciation for the role of mathematics in technology. We hope Shavindra enjoyed the session as much as we did and would love for him to return in the future.
Article written by Soham (Year 12)
