ChatGPT Is Changing the Words We Use in Conversation
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 ChatGPT came out. 'I realized I was using 'delve' more,' he says. 'I wanted to see if this was happening not only to me but to other people.' Researchers had previously found that use of large language models (LLMs), such as those that power ChatGPT, was changing vocabulary choices in written communication, and Yakura and his colleagues wanted to know whether spoken communication was being affected, too.
The researchers first used ChatGPT to edit millions of pages of e-mails, essays, and academic and news articles using typical prompts such as to 'polish' the text or 'improve its clarity.' Next, they extracted words that ChatGPT repeatedly added while editing, such as 'delve,' 'realm' and 'meticulous,' dubbing these 'GPT words.' The team then analyzed more than 360,000 YouTube videos and 771,000 podcast episodes from before and after ChatGPT's release to track the use of GPT words over time. They compared the GPT words with 'synthetic controls,' which were formed by mathematically weighting synonyms that weren't frequently used by the chatbot—such synonyms for 'delve,' for example, could include 'examine' and 'explore.'
On supporting science journalism
If you're enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.
The team's results, posted on the preprint server arXiv.org last week, show a surge in GPT words in the 18 months after ChatGPT's release. The words didn't just appear in formal, scripted videos or podcast episodes; they were peppered into spontaneous conversation, too.
'The patterns that are stored in AI technology seem to be transmitting back to the human mind,' says study co-author Levin Brinkmann, also at the Max Planck Institute for Human Development. In other words, a sort of cultural feedback loop is forming between humans and AI: we train AI on written text, it parrots a statistically remixed version of that text back to us, and we pick up on its patterns and unconsciously start to mimic them.
'AI is not a special technology in terms of influencing our behavior,' Yakura says. 'But the speed and scale at which AI is being introduced is different.'
It may seem harmless—if a bit comical—for people to start talking like ChatGPT. But the trend carries deeper risks. 'It's natural for humans to imitate one another, but we don't imitate everyone around us equally,' Brinkmann says. 'We're more likely to copy what someone else is doing if we perceive them as being knowledgeable or important.' As more people look to AI as a cultural authority, they may rely on and imitate it over other sources, narrowing diversity in language.
This makes it critical to track and study LLMs' influence on culture, according to James Evans, a professor of sociology and data science at the University of Chicago, who was not involved in the study. 'In this moment in the evolution of LLMs, looking at word distribution is the right methodology' to understand how the technology is affecting the way we communicate, he says. 'As the models mature, these distributions are going to be harder to discriminate.' Scientists may need to look at broader linguistic trends beyond word choice, such as sentence structure and how ideas are presented.
Given that ChatGPT has changed how people talk just two and a half years into its adoption, the question becomes not whether AI is going to reshape our culture, but how profoundly it will do so.
'Word frequency can shape our discourse or arguments about situations,' Yakura says. 'That carries the possibility of changing our culture.'
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
New York Times
a minute ago
- New York Times
The Unnerving Future of A.I.-Fueled Video Games
'I need to find my way out of this simulation and back to my wife,' a man said. 'Can't you see I'm in distress?' 'I need to find my way out of this simulation and back to my wife,' a man said. 'Can't you see I'm in distress?' 'I am not just lines of code!' a man in business attire exclaimed. 'I am Liam, a real person enjoying this city.' 'I am not just lines of code!' a man in business attire exclaimed. 'I am Liam, a real person enjoying this city.' Characters in a video game version of 'The Matrix' seemed to be gaining sentience thanks to an A.I. program. Characters in a video game version of 'The Matrix' seemed to be gaining sentience thanks to an A.I. program. Some players were disturbed as these characters, told they would disappear at the edges of their virtual city, cried out for help. Some players were disturbed as these characters, told they would disappear at the edges of their virtual city, cried out for help. Replica Studios Supported by It sounds like a thought experiment conjured by René Descartes for the 21st century. The citizens of a simulated city inside a video game based on 'The Matrix' franchise were being awakened to a grim reality. Everything was fake, a player told them through a microphone, and they were simply lines of code meant to embellish a virtual world. Empowered by generative artificial intelligence like ChatGPT, the characters responded in panicked disbelief. 'What does that mean,' said one woman in a gray sweater. 'Am I real or not?' The unnerving demo, released two years ago by an Australian tech company named Replica Studios, showed both the potential power and the consequences of enhancing gameplay with artificial intelligence. The risk goes far beyond unsettling scenes inside a virtual world. As video game studios become more comfortable with outsourcing the jobs of voice actors, writers and others to artificial intelligence, what will become of the industry? At the pace the technology is improving, large tech companies like Google, Microsoft and Amazon are counting on their A.I. programs to revolutionize how games are made within the next few years. 'Everybody is trying to race toward A.G.I.,' said the tech founder Kylan Gibbs, using an acronym for artificial generalized intelligence, which describes the turning point at which computers have the same cognitive abilities as humans. 'There's this belief that once you do, you'll basically monopolize all other industries.' In the earliest months after the rollout of ChatGPT in 2022, the conversation about artificial intelligence's role in gaming was largely about how it could help studios quickly generate concept art or write basic dialogue. Its applications have accelerated quickly. This spring at the Game Developers Conference in San Francisco, thousands of eager professionals looking for employment opportunities were greeted with an eerie glimpse into the future of video games. Want all of The Times? Subscribe.
Business Insider
a minute ago
- Business Insider
If you live in one of these 13 states, you might have a higher electric bill next year. Blame data centers.
Your electric bills may have shot up in recent months, and you might be tempted to blame your roommate, who never turns the lights off, or your old window air conditioner unit. It's not because of your roommate. It's not your window unit. It's actually Big Tech's fault. Customers of the biggest regional power grid operator in the US could see their bills go up next year, largely due to skyrocketing demand for electricity coming from AI data centers. Last week, PJM Interconnection closed its annual capacity auction with prices for wholesale electric capacity up 22% from 2024, another record-breaking year. As a result, monthly electric bills in PJM's territory, which covers 67 million customers, could increase up to 5% next year, the grid operator said. PJM Interconnection territory spans thirteen states from the Midwest to the East Coast— including all or parts of Delaware, Indiana, Illinois, Kentucky, Maryland, Michigan, New Jersey, North Carolina, Ohio, Pennsylvania, Tennessee, Virginia, West Virginia, and Washington, D.C. Every year, PJM's summer auction determines the cost of wholesale electricity for the following year. Though its territory doesn't cover the entire US, the energy industry looks at the PJM auction as a bellwether for electricity prices for the entire country. PJM's territory includes Data Center Alley in Northern Virginia, home to the world's biggest concentration of data centers. It also includes areas of the country where data centers are rapidly expanding, such as Columbus, Ohio. The grid operator identified data center expansion as the primary driver of demand in its territory, which caused the jump in wholesale electricity prices. After a decade of little to no growth, electricity demand in the US is expected to grow 2.5% annually through 2035, driven largely by data centers, according to the Bank of America Institute. Utility bills are climbing faster than the pace of inflation, according to the US Energy Information Administration. That trend is expected to continue through the next year. In Maryland, People's Counsel David Lapp has been urging state and federal regulators to intervene on behalf of residential utility customers and small businesses. "We are witnessing a massive transfer of wealth from residential utility customers to large corporations—data centers and large utilities and their corporate parents, which profit from building additional energy infrastructure," said Lapp. "Utility regulation is failing to protect residential customers, contributing to an energy affordability crisis."
Bloomberg
a minute ago
- Bloomberg
Boxing, Backflipping Robots Rule at China's Biggest AI Summit
From lumbering six-foot machines to nimble back-flipping dogs, robots lorded over China's most important annual AI conference in Shanghai this week. Thousands turned up to gawk at the antics of a bewildering array of droids at work: dispensing popcorn and drinks (messily), peeling eggs, sparring in a boxing ring, playing mahjong or just wandering around the cavernous exhibition hall. The more popular robots were the creations of Unitree, UBTech Robotics Corp. and Agibot, who've built up some name-recognition among the hundreds of startups and big tech firms vying to produce the world's most advanced humanoid androids. The scores of machines on display were the most visible symbol yet of China's surprisingly rapid ascent in a key arena of artificial intelligence. Hangzhou-based Unitree teased an entry-level $6,000 droid and ByteDance Ltd. posted a video of its Mini hanging up a shirt just days before the World Artificial Intelligence Conference kicked off over the weekend.
