BearcatGPT Pilot

The BearcatGPT Microsoft Azure OpenAI Pilot provides the UC community with a unique opportunity to explore OpenAI/GPT in a private, UC-only environment. Use cases include teaching, learning, research, and process automation. 

This limited-use pilot enables the university to gain hands-on experience with a private OpenAI environment, supporting a strategic and phased approach to adopting AI across UC.

Pilot FAQs

General

BearcatGPT is UC’s private instance of OpenAI’s ChatGPT natural language processing model.  It is a safe environment for UC faculty, staff, and students to use generative AI for a variety of tasks such as answering questions, providing information, and helping with academic work.  Currently, OpenAI’s GPT-4o and GPT-3.5 are available.  

The BearcatGPT Pilot provides the UC community with a unique opportunity to explore OpenAI/GPT in a private, UC-only environment. The data used in BearcatGPT does not contribute to the publicly available OpenAI/GPT.

BearcatGPT is in the pilot phase and currently restricted to pilot users. After the pilot phase is completed, communication regarding expanded usage will be shared. Please submit any questions to the AI Questions Submission Form.

BearcatGPT can be used in desktop or mobile web browsers by visiting https://bearcatgpt.uc.edu and logging in with your UC credentials.

There is no limitation on frequency of use; however, excessive use may result in throttling.

There is no charge to use BearcatGPT, but it is currently restricted to pilot users.

As a web application, people can use their web browser assistive technologies to use the site. The Accessibility Network has tested BearcatGPT for accessibility and has identified some areas for improvement that are currently being addressed. These enhancements should not prevent access, but users are encouraged report any issues.

BearcatGPT is capable of understanding and processing text in a wide variety of languages.  It's important to note that proficiency and accuracy can vary depending on the language and the complexity of the text.  Widely supported languages include English, Spanish, French, German, Chinese (Simplified and Traditional), Japanese, Korean, Portuguese, Russian, and Italian.

Usage and Policies

See “Acceptable Use” in UC’s AI Policies.

See “Use publicly available data” in UC’s AI Guidelines & Policies.

See “Guidelines for Students” in UC’s AI Guidelines & Policies.

See “Guidelines for Research” in UC’s AI Guidelines & Policies.

Navigation

Type into the chat. Select the paperclip icon to upload a file from your device. Or, record your voice by selecting the microphone icon. BearcatGPT is currently limited to text output; however, access to DALL-E is planned for release in BearGPT in early 2025.

View CET&L’s “How to write Generative AI Prompts” page on Bearcats Landing for more about prompting best practices.

Click on the menu with your name at the bottom of the left sidebar and click Logout.

Find the chat in the left sidebar, click the ellipsis (three dots) to the right of the chat, and select Rename, Archive, Delete, or Export from the menu.

At present, the system does not allow you to access your archived chats. The vendor has this feature on their roadmap, and once it is implemented it will be available in BearcatGPT. 

Drag and drop a file from your device to the chat.  You can also click the paperclip icon and browse files on your device. 

Click the microphone icon to record your voice.  You may need to allow your browser to access your microphone.  To enable audio output, hover your cursor over the text output and click “Play Audio” (speaker icon) at the bottom.

Use your web browser’s text controls to increase or decrease text size.

Content

The GPT-4o model’s training data is current up to October 2023.  Any events, developments, or information that emerged after this date are not included in the knowledge base. As new models are released, they will become available.

See UC Libraries’ guide on citing generative AI.

Accuracy can vary depending on the topic, context, and complexity of the question. Generally, for well-documented or widely known subjects, responses tend to be highly accurate, especially within the scope of the training data (up to October 2023).  It’s best to independently verify the accuracy of responses. 

BearcatGPT can process a multitude of file types including images, PDFs, Word, Excel, and CSV.

BearcatGPT is currently limited to text output; however, access to DALL-E is planned for release in BearGPT in early 2025.

There are limits, and they are based on tokens. Tokens do not map directly to words, so it isn’t possible to say what the limit is in words. Realistically users should not expect to see the limit in action unless they were to attempt to produce an extremely long output. For example, you will not be able to use BearcatGPT to produce a novel or dissertation.

Privacy

The data used in BearcatGPT does not contribute to the publicly available OpenAI/GPT.

Your chat history is visible only to you. Your chat history is logged in the system, and while it is not visible to other users, it is possible for administrators to view chat-related information.

Currently the chat history does not expire. Your chats will be saved until you delete them.

You cannot turn off your chat history, but you are able to delete your chats at any time.

Your chat history will not be used to improve BearcatGPT.

Support

UC regularly holds workshops on AI and teaching. To sign up for a generative AI workshop, please visit Faculty OneStop.  You can also view resources on the Teaching & Learning page of the AI website

Please submit questions, suggestions, and issues related to the BearcatGPT Pilot to the AI Questions Submission Form.

Basic guidelines are available within ChatGPT. A series of tutorials are being developed and will be available for use in the near future.

Please submit questions, suggestions, and issues related to the BearcatGPT Pilot to the AI Questions Submission Form

Updates to BearcatGPT will be published on the AI website and within BearcatGPT.