Generative AI Use

There has been immense progress in the last 2-3 years regarding the use of language models to assist in writing, learning, and code generation. With this assistance comes a responsibility to use it responsibly and ethically. Below are some compiled resources on how to do. The key points are to take ownership of ideas, code, and analysis generated by LLMs and double check your work by cross-referencing with other sources. Use it transparently and provide proper attribution. LLMs still exhibit phenomena such as sycophancy and hallucination, which may alter the reliability of its generated answers.

Resources

  1. Living guidelines on the responsible use of generative AI in research (European Commission)
  2. Reminder of the Importance of Research Integrity & Use of AI (University of Virginia)
  3. Effective and Responsible Use of AI in research (University of Washington, Georgia Tech)
  4. Slide from Patrick Mineault on potential pros of AI use (Dan on Bluesky)
  5. Should neuroscientists 'vibe code'?(Benjamin Dichter, The Transmitter)

GitHub Copilot Access

As university-affiliated researchers, we may have access to AI completion tools such as GitHub Copilot. The Pro tier provides a 30-day free trial for researchers/educators. Here are the steps to get access:

  1. Visit the GitHub documentation on Copilot Pro access.
  2. Click on "Apply to GitHub Education as a teacher".
  3. Navigate to Education Benefits settings
  4. Complete the form and submit. You may need to verify that you have an @yale.edu email address and provide a picture of your ID card. The application will be approved by GitHub admins.
  5. You should see a coupon for GitHub Copilot Pro populate in 48-72 hours. This provides access for 30 days. This can get renewed.