Earlier this year, Florida State University Assistant Professor of Computational Linguistics Tom Juzek helped uncover why ChatGPT overuses certain words. His latest work examines how the AI chatbot could be changing the way people speak.
Through an analysis of some of the nation’s most popular tech industry podcasts, Juzek and his team compared the language used in the podcasts before and after the release of ChatGPT. They found a significant shift in the use of AI-related buzzwords such as “surpass,” “meticulous,” “strategically,” and “intricate,” indicating their increased usage.
“The buzzwords saw a significant shift between pre- and post-Chat GPT release, and the synonyms didn’t,” Juzek said. “The synonyms moved very little. It was quite a shift for such a short period of time when it comes to changes in spoken language.”
Juzek is part of FSU’s Department of Modern Languages and Linguistics in the College of Arts and Sciences. His research interests include the application of computational methods to linguistic questions, with a current focus on the features of AI-generated language, its impact on human language usage, and the causes that influence model behavior.
Research into how much AI is influencing language and behavior — such as how the technology might be accelerating language change or acting as an innovator — is still in its infancy, Juzek said.
“These models have some clear upsides and are genuinely useful tools, but open questions exist about unintended negative consequences for beliefs and behavior,” he said “With AI, it is premature to claim causation. Our current results are correlational. Next comes careful testing to see what is truly changing and why.”
Juzek noted a parallel between AI and social media: both have meaningful benefits alongside specific harms documented over time.
“Social media taught us that both can be true, but that good causal research takes time,” he added.
Juzek is available to speak on the following topics:
- The latest insights into AI and language: What the newest analyses show about AI-associated word choices in modern speech.
- A balanced view of impact: Clear benefits alongside open questions about unintended effects on language and behavior situated within the wider discourse on AI and society.
- Why models behave this way: The technical mechanisms behind word overuse and how data and objectives shape style.
- The human side of improvement, including the global annotation and content moderation workforce and concerns about working conditions.
- State of the discourse and what’s next for the industry.
Media interested in discussing Juzek’s research can reach out to him at tjuzek@fsu.edu.