Generative AI: ChatGPT & Other Text-Based Generative AI

Detection tools available or in development that purport to determine whether a work was generated by a human or AI, such as GPTZero or OpenAI Detector, should be utilized conscientiously, as detectors are prone to false positives as well as negatives if content generated is significantly altered by a human before submission. The college does not recommend you rely solely on detectors to claim instances of academic misconduct as they are unreliable. **Many of the available detection tools are not free and can be quite expensive**

OpenAI CEO Sam Altman said in a 2023 interview that "there may be ways we can help teachers be a little more likely to detect output of a GPT-like system. But honestly, a determined person will get around them." OpenAI is experimenting with watermarking technology to indicate when content has been AI generated (see video below) but cautioned folks against depending on these tools to detect AI outputs.

AI technology like ChatGPT is constantly evolving, but the current iteration of ChatGPT includes the following limitations: 

• Cannot integrate research or provide citations. 

• ChatGPT does not draw on data from the open internet but instead uses a data training set consisting of information from 2021 and earlier. Therefore, it knows nothing of events post-2021 and cannot respond to prompts with real-time information like current events.  OpenAI has created plug-ins to incorporate live data from the web into ChatGPT conversations (May 2023).

• Has the potential to generate biased and incorrect responses. Developers have stated it contains a heavy Western knowledge bias. 

• Has the tendency to make up information and sources.

• Cannot fact-check or engage in critical evaluation of sources. 

• Seems to have difficulty with simple math.