Are We Becoming More Like AI?

Google’s Deepmind visualized. Courtesy | Wes Cockx

Recent studies have found that the word choice and phrasing of language is being heavily influenced by artificial intelligence models, even when the work was entirely created by a human. 

Not only is this remarkable, it could be seriously detrimental to human expression.

As a specific instance, the word ‘delve’ was found to be nearly twice as common after the release of ChatGPT, according to research published by the Max-Planck Institute for Human Development in July.

The study further suggests this is most likely due to the use of artificial intelligence and large language models being more frequently used in general. This has led to the biases and word choices seen in those models to be fresh on the mind, with a higher chance for authors to use the words in their own voice.

AI influence has not just been found in scholarly articles either, the Max-Planck Institute included thousands of hours of YouTube video transcripts and podcasts to search for these influenced words and phrases.

This is a greatly negative development because AI models have an algorithmic bias that shrinks the human vocabulary to be more objective and efficient. Additionally, models are continuously fed new pieces of information to improve which would cause a feedback loop if key phrases are found in more writing.

Florida State University published a related study specifically surrounding the trends of spoken English, in the sense of conversation rather than scripted performances to see if word choices were only influenced during the process of creating a hard copy.

After testing words that were presumed to be more associated with AI and words that were not, several words were found to be more than twice as likely to be said within spoken communication, the study said.

Words like ‘surpass’, ‘boast’ and ‘meticulous’ showcased some of the most sharp increases, the study said.

This is a unique circumstance because word choice is a form of creativity. Taking away that creativity by making it more likely for people to use the same words is a step back in society. Words do not have to be efficient or uniform for all purposes, they can be inspiring, personalized and distinctive.

The structure of speech and writing is also transforming alongside the words themselves. AI models like ChatGPT use more nouns and less adverbs to show objectivity and less emotion, according to a study published in 2023 by Biyang Guo.

This means that not only are words getting more repetitive in a feedback loop, but the formational structure of the language itself is being optimized for supposed efficiency.

The varied syntax of writing is powerful because it conveys various nonverbal indications such as formality or tone. If this structure continues to be simplified through AI models, communication may become similar to playing a game of Mad Libs.

Considering the rate at which modern AI has advanced in just under three years the possibilities of its influence on various forms of communication are unknown.

For the environment of a college campus, sophisticated and flavorful ideas are an important component to communication. Each student’s voice is more substantial than a formulaic mode of speech that sacrifices taste for perceived coherence. It is also possible that a backlash effect could happen after the spread of this worry and these studies. Words more frequently utilized by AI may decrease in usage when compared to before 2022 out of this human fear.

Human expression is under the automated threat of simplicity as ideas and creativity begin to be streamlined. The studies show that certain words and phrases are drastically becoming more common in the time period following 2022, which correlates with the release of ChatGPT and other developed AI models.

Comments

comments