How Machines Learn: AI from the Perceptron to GPT

Catalog description

This course covers learning, deep learning, and neural networks from the perceptron through modern architectures such as GPT. Students will build intuition for how machines learn, explore foundational neural architectures, implement learning algorithms on neural networks, learn about small-scale and large-scale hardware architectures for leaning and AI, efficiency of and energy cost of AI systems, and reflect on the ethical and societal implications of AI. More advanced topics such as recurrent neural networks (RNNs), long short-term memory (LSTMs), Transformers and Large Language Models will be introduced at a high level with emphasis on intuition and demonstrations rather than mathematical details. The course emphasizes concepts and applications with mathematical tools that freshmen can wield to connect engineering tools with AI methods. The course also discusses AI hardware topics related to power consumption, computational infrastructure requirements, and the role of large-scale data centers in enabling AI systems. Some programming experience in Python (or an equivalent programming language) will be useful in completing assignments and design projects.