Wayne State University
Warrior strong
Login
Search
C&IT Computing & Information Technology

WarriorGPT generative AI

Overview

WarriorGPT is a standalone generative AI chatbot tool designed specifically for use at Wayne State and to mitigate the challenges posed by using institutional data with external platforms. It works just like ChatGPT and other chatbots, and faculty and staff are encouraged to use it instead of such tools because they may store and train off of user-inputted information. 

WarriorGPT includes an easy-to-use chat interface and various general purpose open-source models to support both instructional and administrative functions. Faculty interested in using WarriorGPT in their term courses can follow the instructions in the Knowledge Base to enable the platform in Canvas and grant their student's access. 

FAQ

  1. How does WarriorGPT work? 
    • WarriorGPT is a generative AI platform that hosts various open-source large language models (LLM). These models undergo resource-intensive training with their service providers on vast sets of texts that include their own guardrails. Using pattern and structure recognition against the training datasets, LLMs are capable of understanding and generating human-like text. LLM responses are not drawn from a repository of sources but are instead generated by mirroring characteristics seen during training. LLMs excel at text generation, translation, text summarization, dialogue, sentiment analysis, question answering, and content creation. 
  2. How does WarriorGPT differ from other generative AI tools? 
    • WarriorGPT is housed in C&IT’s private infrastructure 
    • WarriorGPT does not save user conversations 
    • WarriorGPT does not learn from user input 
    • WarriorGPT utilizes large language models pre-trained for ethical engagement  
  3. How was WarriorGPT trained? 
    • WarriorGPT’s hosted models are trained by their service providers on publicly available online data. C&IT does not perform any additional training or finetuning on these models. 
  4. Can I trust the responses from WarriorGPT? 
    • Large language models (LLM) are generative in nature and will attempt to respond regardless of if an accurate answer exists in its training data. They do not fact-check or verify response accuracy. Important information should be verified for accuracy against known sources. 
  5. Can WarriorGPT be accessed through an API? 
    • No, access is currently limited to the visual chat interface. API access is planned for a future release. 

Related university policies and guidelines  

All use of WarriorGPT should be related to Wayne State instructional and administrative goals. Before getting started, please review the guidelines for using AI at Wayne State. 

Specific policies governing the use of AI tools in coursework are left to the instructor to establish and communicate with their students. All instructors are strongly encouraged to talk with their students about the responsible use of AI, course policies, and best practices as they relate to the discipline being taught. The Academic Senate ad hoc Subcommittee for Artificial Intelligence has created sample syllabus language for both including and excluding AI from the classroom.

Resources