Skip to Main Content
University of Texas University of Texas Libraries

Artificial Intelligence (AI)

Benefits and Limitations

Benefits of Generative AI Tools

  • Brainstorm: Chatbot tools such as Microsoft CoPilot* and Gemini may be helpful for brainstorming ideas for research topics, organizing your thoughts, jump-starting your work and tackling writer's block.
  • Break down concepts: Chatbot tools such as Microsoft CoPilot* and Gemini may help you break down and understand a complex concept or assignment prompt.
  • Illustrate: Image generation tools such as Midjourney and Dall-e may be helpful for presenting or illustrating your work.
  • Create: Many AI tools may be useful for creative adaptations.
  • Summarize: Many AI tools, including chatbots like Microsoft CoPilot* and AI literature search, synthesis and visualization tools like ResearchRabbit and Elicit, can summarize articles.
  • Discover new research: Many literature search, syntheisis and visualization  AI tools such as Research Rabbit and Elicit may help you discover new research in your area and visualize connections between researchers and between research literature.
  • Translate: AI tools may help you translate between languages.
  • Code: AI tools such as GitHub Copilot may help you generate new code and clean up existing code.

*Microsoft CoPilot is licensed by UT.

Limitations of Generative AI Tools

  • Hallucinations: When using AI tools such as Microsoft CoPilot or ChatGPT for research, they may make up credible-sounding citations to sources that do not exist, or give inaccurate information, which is called “hallucinating.”
  • Paywalled content: Literature search tools such as Research Rabbit or Elicit do not have access to the full range or full text of articles that are behind a paywall (access you may have with your EID). They may help with literature searching and with systematic reviews, but cannot fully substitute for a human being with access to this paywalled content.
  • Scope of training data: AI tools can only produce based on the data they have been trained on, so it is important to understand what comprises the training data and the date ranges included. For example, as of 2023, ChatGPT 3.5 (free version) is only trained on content ingested from the open, non-paywalled Web through December 2021.
  • Reproducibility: Because generative AI tools such as Microsfot CoPilot and ChatGPT create new content based on their training data, the content it creates is not reproducible. In addition, because these tools create new content, you may find that multiple people using the same prompt at the same time will get different results.
  • Ethics, Privacy, etc.: There are numerous limitations related to ethics, privacy, bias, labor and environmental impact outlined on the Ethics and Privacy page.
  • Legal Context: US laws related to AI tools are evolving, and tools to identify AI-created content are not fully effective. See Copyright page for more information about copright and intellectual property.
  • The Office of Academic Technology's Addressing the Limiations of Generative AI for Learning includes additional considerations.

Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 Generic License.