AI in PR: Negative Aspects of AI

There are many great, practical uses of AI, but you must proceed with caution; it is not the answer to everything.  If you are considering using AI-generated content to aid your clients and the work you produce for them, you must understand its limitations, including: 

Accuracy – Machine learning must have a database as its foundation of information. Chat GPT uses the internet, which, as we all know, is not always an accurate source. Recently, Chat GPT was used by a reporter who wrote a story about a radio host from Georgia that was completely false. Open AI, the platform who owns Chat GPT is now being sued for defamation. This example highlights an issue with using AI for articles or other long-form writing. 

Bias – In addition to accuracy issues, bias is a major concern for many experts. One example of this is the study conducted on DALL-E 2, an AI image creating platform. An MIT article discusses a study that found the program produced images of white men 97% of the time when asked to feature a CEO, director, or other similar prompts. It is important to note this bias, to ensure that you are not using AI in a way to perpetuate stereotypes. 

Data Security – AI programs are constantly learning; they are updating through every “conversation” with users. It is not yet clear how inputs affect the program, however it is possible that all given text is used to train the technology. This means when you ask AI to use some piece of information, the program now has that data stored. The security that is protecting your stored data is not clear, which is why you must be cautious when giving AI access to information that should not be public.

Copyright – Laws surrounding AI programs are currently in development. Although not yet prohibited, training AI on copyrighted material may become problematic. This is applicable to both text producing and image creating AI programs. For example, if OpenArt was asked to draw a picture of Mickey Mouse, and you then use that image in an advertisement, it could be considered copyright infringement. 

AI programs, while often an appealing choice, may lead you down the wrong path. Whenever using AI, always: fact check the output; ensure it is not in any way reflecting a bias towards any group; avoid inputting sensitive information; and be wary of content that could be protected by copyright. 

Learn next about creating strong inputs for the best outputs.

About the author:  Devon Rudolph is a senior at Robinson Secondary in Fairfax, Va. She plans to study journalism in college. A focus of her summer internship at BWG has been artificial intelligence in public relations. This is the third in a 4-part blog series.