Fake Books and Imaginary Stats: AI at Work 🤖

Content transparency: This blog post includes a 100% AI-generated graphic, partially AI-generated headline, and 100% human-generated blog copy. 🙂

This is not a criticism of artificial intelligence. I use AI assistants in my everyday work. It’s a reminder that there is no replacement for human oversight.

Is AI trying to make humans look stupid? Despite the numerous warnings about AI hallucinations, AI tricks even the most trusted sources, spitting out inaccurate or simply made up content.

Last month, the Chicago Sun-Times published a list of 15 must-read summer books. Great! Except 10 of them don’t exist.

The list was generated by AI, but no one took the final step: Verifying and fact-checking.

Have the hallucinations gotten worse over time? Or are we more aware because more people are using generative AI? I can think of three instances where AI has produced false or inaccurate information for me. Two of them happened in the past two weeks.

1. Nonexistent statistics: I asked to provide specific industry statistics and it produced several legit sounding statistics and included links to where it found the information. The provided links are legit industry sources. The problem? The stats did not exist on those websites.
2. Made-up people: I wanted to summarize webinar session information into 1-2 sentences. I included the link to where the information was housed online, and asked AI for help. While the webinar session description was accurate, the presenter’s information was wrong. AI provided a totally different person I had never heard of. I asked AI about the inaccurate information. When I questioned the source, the AI bot provided a new response with the correct presenter name, but it pulled the information from a different website (and not the link I provided).
3. Outdated information: We’re exploring options for media monitoring, and we asked AI to provide a list of companies for us to contact. While it produced a legit list of companies, the information was about five years old. We found this out after asking one of the company reps about the results during our call.

Experts warn that while AI can replace some duties, be super helpful, and save time, it still requires human oversight. We must verify information, fact-check sources, and review AI-generated content.

Haven’t we been saying this all along?