On November 13, Constructor Knowledge Labs and Constructor University hosted the second event in our “TED-style” CKL Talk series, welcoming Dr. Nikolay Malkin – a Chancellor’s Fellow in Informatics at the University of Edinburgh and a Fellow of CIFAR’s Learning in Machines and Brains program.
Dr. Malkin’s talk, “Learning to Construct: Advances in Structured Inference and Bayesian Neurosymbolic AI,” turned out to be both insightful and surprisingly accessible. Students, faculty, and researchers filled the room, curious to hear how modern AI can learn to reason about complex, structured information, and why this matters for the future of machine intelligence.
Throughout the session, Dr. Malkin walked us through new ideas that combine reinforcement learning, variational inference, and Monte Carlo techniques. He showed how these methods help AI systems work with things like graphs, symbolic programs, and sequences, while still keeping a coherent sense of uncertainty – something that often gets lost in more “black-box” approaches.
A big part of the conversation focused on neurosymbolic AI, a field Dr. Malkin has helped push forward. He explained how mixing deep learning with formal reasoning isn’t just a theoretical exercise, but a practical way to build models that can generate, discover, and design in more meaningful ways.
We’re grateful to Dr. Malkin for a talk that genuinely inspired. It sparked new questions, conversations, and a sense of excitement about where structured reasoning in AI might take us next.
We look forward to seeing everyone again at the next CKL Talk!