Perhaps you want another explanation of a concept, more papers related to one you have read, or help starting an email. The collections below include short descriptions to help you find the material you need.
If you have not tried a project yet, choose a starter project and come back when a question arises.
Courses and papers
- Courses, textbooks, and foundations: computing, Python, mathematics, ML/DL, and architectures, including Crash Course, 3Blue1Brown, Stanford, and MIT.
- Mu Li: courses and paper readings: if you want to move from D2L to reading papers, see how the courses and reading sessions fit together.
- Courses, papers, and projects by direction: vision, LLMs, multimodal models, generation, RL, world models, embodied AI, agents, systems, AI4X, graphs, and time series. Includes Lumina, awesome lists, and paper lists.
Doing research: methods and process
- Research methods and tools: find tools and tutorials for searching for papers, troubleshooting training, writing, and preparing a talk.
- Technical blogs and research perspectives: read fuller explanations of model mechanisms and accounts of specific training and evaluation questions.
- Research methods and communication: find examples for a first email and researchers' accounts of reading, discussion, and collaboration.
Research information and sources
- LessWrong: research judgment and AI use: ten essays on evaluating evidence, checking whether advice fits a task, and examining how AI responses shape judgment. Start with a question relevant to your work.
- Research groups and teams: what to look for, how to follow papers to people, and what to check before contact.
- Blogs, interviews, and information channels: authors, media, original releases, and perspective essays. See Lookout for following research.
- BlogrXiv · OpenEnvision: AI research blogs, lab essays, and technical notes organized by field, with links to the original writing.
- ScholarTube · OpenEnvision: Researcher interviews, video podcasts, complete courses, and research talks, searchable by field and linked to the original videos.