Reliability
Large language models (LLMs) (like Chat GPT, natural language processing (NLP) tools and chatbots are trained on a limited set of materials that does not include research behind paywalls. They generate answers that “sound good” and statistically match what human language sounds like. There is no guarantee that what they come up with is truthful or contains reliable answers.
Transparency
AI tools are fed a large amount of information, but they do not have access to research that is behind a paywall. They are not always able to find information that supports what they are saying and could invent research articles and citations that do not exist.
Bias
These tools are only as good as the information they are provided with. The information they are provided with is created by humans and contains all of the biases, cultural, social and otherwise, that humans have.
Privacy
AI tools are not able to determine that some information should remain private (HIPAA). Additionally AI tools accept the input of private information from users who may not be aware of how the platform intends to use that information.
Equity
AI tools use the work of human authors without acknowledging or compensating them for their work. Many of the points of the current WGA and SAG/AFTRA strikes are about humans competing with AI in ways that are impossible as the AI works for no compensation.
AI tools’ information sources are dominated by english language resources and are biased toward that language and leave little room for indigenous and less common languages.
Powerful, quality AI tools often require payment and create a disparity between those who can afford access to the tools and those who cannot.
adapted from USF AI Tools and Reources
