ChatGPT Is Changing the Words We Use in Conversation
Chatgpt modifies the words we use in conversation
Words frequently used by Chatgpt, including “Delve” and “Meticle”, become more common in the spoken language, according to an analysis of more than 700,000 hours of videos and podcasts
After its release at the end of 2022, Chatgpt reached 100 million users in just two months, which applies consumption for the fastest growth in history. Since then, the artificial intelligence tool (AI) has considerably affected the way we learn, write, work and create. But new research shows that this also influences us in a way in which we may not know – as changing the way we speak.
Hiromu Yakura, a postdoctoral researcher at the Max Planck Institute for Human Development in Berlin, first noticed differences in his own vocabulary about a year after the release of Chatgpt. “I realized that I was using” more “more,” he says. “I wanted to see if it happened not only for me but for others.” The researchers had previously discovered that the use of models of large languages (LLM), such as those which Power Chatgpt, changed vocabulary choices in written communication, and Yakura and his colleagues wanted to know if spoken communication was also affected.
Researchers first used Chatgpt to modify millions of email pages, trials and academic and press articles using typical prompts such as “polish” text or “improve its clarity”. Then they extracted the words that Chatgpt added several times during the edition, such as “Delve”, “Kingdom” and “Meticle”, doubling these “GPT words”. The team then analyzed more than 360,000 YouTube videos and 771,000 episodes of before and after the Chatgpt version to follow the use of GPT words over time. They compared the GPT words with “synthetic controls”, which were formed by mathematical weighting synonyms which were not frequently used by the chatbot – such synonyms to “dive”, for example, could include “examine” and “explore”.
On the support of scientific journalism
If you appreciate this article, plan to support our award -winning journalism by subscription. By buying a subscription, you help to ensure the future of striking stories about discoveries and ideas that shape our world today.
The results of the team, published on the Arxiv.org preparatory server last week, show a wave of GPT words in the 18 months following the release of Chatgpt. The words do not only appear in formal and scripted videos or episodes of podcast; They were also dotted with a spontaneous conversation.
“The models stored in AI technology seem to transmit to the human mind,” explains the co-author of the Levin Brinkmann study, also at the Max Planck Institute for Human Development. In other words, a kind of cultural feedback loop is formed between humans and AI: we form the AI on the written text, it parrot a statistically remixed version of this text to us, and we take back its models and we unconsciously start to imitate them.
“AI is not a special technology in terms of influence on our behavior,” says Yakura. “But the speed and scale at which AI is introduced is different.”
It may seem harmless – if a little comical – so that people start to speak as a chatgpt. But the trend has deeper risks. “It is natural that humans imitate each other, but we do not imitate everyone around us too,” explains Brinkmann. “We are more likely to copy what someone else does if we perceive it as being well informed or important.” While more and more people are turning to AI as a cultural authority, they can count and imitate it on other sources, narrowing diversity in language.
It is essential to follow and study the influence of LLMS on culture, according to James Evans, professor of sociology and data science at the University of Chicago, which was not involved in the study. “At the moment of the evolution of LLMs, looking at the distribution of words is the right methodology” to understand how technology affects the way we communicate, “he says. “As the models ripen, these distributions will be more difficult to discriminate.” Scientists may need to look at wider linguistic trends beyond the choice of words, such as the structure of sentences and how ideas are presented.
Since Chatgpt has changed the way people only speak two and a half years in its adoption, the question does not become if AI will reshape our culture, but how deep it will do.
“The frequency of words can shape our discourse or our arguments on situations,” explains Yakura. “It brings the opportunity to change our culture.”
