Schedule

🚧 Warning: Under construction 🚧

Epoch 0. Initialization: overview basic concepts of artificial intelligence, symbolic computation, machine learning, deep learning, and neural networks, and multi-agent systems as well as AI ethics and ML ops. In lab we’ll write some basic decision trees to control simulated and real robots and tap into knowledge bases to make a trivia answering system.

Epoch 1. Machines that Learn: fundamental concepts of machine learning, including optimization and gradient descent, probablistic modeling and factor graph representation, regression, classification, clustering, regularization, and ensembles. In lab we’ll parse demographic data and build a shallow machine learning model to predict the likelihood of a user’s income.

Epoch 2. Deep Dive: basic concepts of deep learning, common architectures, layer types, and activation functions as well as backpropagation, optimization, loss functions, and training paradigms. In lab, we’ll build a neural network from custom-built layers to solve an image classification problem.

Epoch 3. The Zoo: overview of specific deep learning architectures including DNNs, AEs, CNNs, GANs, RNNs, attention-based models, GNNs, and mutlimodal systems. In lab we’ll do neural network ‘surgery’ on several popular models and try to improve their performance.

Epoch 4. Reinforcement Learning: the classical utility agent, various extrinsic and intrinsic rewards, training setups, and environments, cooperative/competitive/mixed-mode multi-agent reinforcement learning systems, and advantages and pitfalls of RL. In lab we’ll train a mixed-mode multi-agent reinforcement learning system in various simulated environments.

Epoch 5. Intelligence: transfer learning, generalization, zero/one/few-shot learning, distilation, information theory, corellators, criticality, generalization, transferability, the relationship between intelligence and learning, and the limits of AI generality. In lab we’ll play with large language models including gpt-3 ada, babbage, curie, davinci, davinci-instruct, davinci-codex. We’ll also attempt small and large examples of transfer learning all the way to transfering a pretrained language model to perform computer vision tasks.

Epoch 6. Hello World: AI safety, security, privacy, federated learning, and ethics, AI in the news, in acedemia, and in the workplace, ML ops, human-AI interaction, environmental impact, and the impact of AI on human life. In lab students will put MLops to work on a small self-guided group project of any size. Groups with reasonable metric performance will receive additional compute resource credits.

Epoch 7. Project Day! Students present their projects to the class. Students will also have a chance to connect with and other students to discuss their projects, the course, and AI.