Policies*

Use of AI systems

In this class, students are allowed and indeed expected to use AI systems, including coding agents, while building AI systems.* However, you remain responsible for the reports and systems you submit. Each submission will carry a requirement to describe which systems you used and how. Quizzes, on the other hand, are closed book and closed notes and will not allow the use of AI systems.

*Though we rely on free Prime Intellect credits and MIT's Parley system for serving LLMs inside AI systems, we unfortunately cannot provide additional usage specifically for coding agents. Nonetheless, verified university students in the United States can claim $100 in ChatGPT credits to use in Codex and there might be other similar opportunities from other providers.

Collaboration

Homeworks are individual. You may discuss course concepts and infrastructure concerns (e.g., how to make the most out of models, services, or coding agents), but you may not share implementations or experimental results. You must acknowledge any external literature or open source software that you build on.

We will discuss project collaboration rules when the project specification is released. The projects will be in small groups but every team member must be able to represent the team and explain the complete system.

Models and compute

The course will supply the model access and compute required for graded work. You may use personal AI subscriptions for development or additional exploration (recall that the assignments are quite open-ended, as are AI systems in practice), but you will not strictly need them or any additional funds beyond what the course supplies for successful and high-scoring completion of the assignments.