1. Review the University’s AI Guidelines
Before getting started with the chatbot, please make sure to read the University’s AI guidelines based on your role at Yale.
- For teaching and learning, please refer to the Poorvu Center’s guidelines.
- For staff, please refer to Yale University AI guidelines for staff.
2. Understand your responsibility
Yale University operates under the shared responsibility model which means that all users have a responsibility for using technology provided by the University in a responsible and appropriate manner. This means that you are responsible for adhering to current university policies on academic integrity and ensuring that AI application complies with the Yale University Policy Against Discrimination and Harassment and University Sexual Misconduct Policies
You are responsible for any content that you produce or publish that includes AI-generated material: AI-generated content can be inaccurate, misleading, or entirely fabricated (sometimes called “hallucinations”) or may contain copyrighted material. Review your AI-generated content before publication. You are responsible for helping to prevent AI hallucinations and the accuracies of responses.
You are responsible for complying with Microsoft’s Code of conduct for Azure OpenAI Service
3. Understand the capabilities and limitations of the chatbots
The Yale secure AI chatbot is currently available as a limited pilot, with access coordinated by Yale ITS. The secure chatbot can answer general questions based on information available on the web or data passed into the prompt. The AI chatbot will not save your conversations and data entered will not be used to train any public AI tools. It offers a single interface that enables access to a limited number of Large Language Models (LLMs): currently Azure OpenAI GPT-3.5, GPT-3.5 Turbo, and GPT-4.0. It is the expectation that more models will be added in the future.
For chatbots utilizing GPT-4, please note that if you inquire about the version, it is expected that the response will indicate GPT-3. This is because the model is trained solely on data up to early 2023, and that data may not contain information about GPT-4, if it hadn’t been released at that point in time.
Capabilities of the chatbot
- You can copy text and insert it into the chatbot. The chatbot will then perform analysis on the content if instructed properly. Please refer to the section named “Copy your own text into the AI chatbot” for instructions on how to do this.
Limitations of the chatbot
-
For the current versions of the chatbot, it is not possible to upload files of your own materials into the chatbot. This means that the chatbot cannot be custom trained on your own material or content. Future versions may allow for the uploading of materials for the model to reference.
-
The chatbot uses LLMs that are subject to rate-limiting by the vendors we are using, meaning that there is a restriction on the number of interactions that can be made with the chatbot within a specific time frame. As a result, the responses by the chatbot may at times be slower than expected.
-
The chatbot cannot be interacted with through an API as Yale ITS does not currently offer Azure OpenAI API access. This means that you as a user can only interact with the chatbot through the website interface and not connect to and interact with it programmatically.
4. Login to the chatbot
-
Log in using CAS.
-
For teaching and learning, the link to the chatbot can be found in your course in Canvas given that your course is using a chatbot. Please consult your teacher if you are in doubt.
-
For staff members and others who are a part of this pilot chatbot effort, you will have been informed by your ITS contact person with the link to the website with the chatbot that you have been granted access to.