are Chatbots Making Us Boring?

Boring people don’t listen. Thay tell their own stories, over and over, and never make any attempt to engage in our story, our lives.And so we avoid them.

AI, conversely, is a superb listener.So much so that people, particularly teens, are turning to chatbots for companionship. But in doing so, do we run the risk of all becoming the same kind of person, wanting the same kinds of friendships, with the same kinds of interactions? In a word, boring.

Maik Bieleke and Wanja Wolff capture this well in what they call “the boredom trap.” They point out that AI large language models (LLMs)-what we use for both ChatGPT and companion bots-are, by their nature, regressing towards sameness. These models are based on predictive algorithms. They vacuum up enormous amounts of data (e.g., things actual people have written), run them through a black box to determine the likelihood that one word will follow another, and use those probabilities to churn out coherent speech. You see this in yoru text completion on your phone, and in general, it works pretty well.

But as Bieleke and Wolff point out, this will also trend towards monotony. Always choosing the high probability option narrows our expressive range. Language is as valuable and amazing as it is indeed in part as of its capacity for nuance. One of us had a French colleague who once asked, “Qu’est que c’est la différence par “perhaps” et “maybe?” It’s tough to formulate a good answer, and maybe there isn’t one. But perhaps having the different options is what matters.

Relying on LLMs for writing in science risks shaping our corpus of outputs in a narrow (and we would argue, boring) way. But there may be even bigger risks. Relying too heavily on LLMs may make us more boring.

Indeed, recent evidence suggests we might be in a feedback loop with the LLMs we’re starting to rely on more and more. Hiromu yakura and colleagues analyzed outputs of these LLMs and compared them to natural speech. It seems we’re starting to incorporate into our own speech many of the buzzwords favored by AI language models. Words like “delve,” “boast,” and “swift,” for some reason, are in vogue for ChatGPT right now, and we’ve started using them more. We’re becoming more like chatbots. As Yakura points out, this “closed cultural feedback loop” runs the risk not merely of being boring, but of eroding our vast array of unique, diverse linguistic expressions. The move to monotonous, boring sameness could be a culture killer.

But surely what we get from chatbot companions can’t be all bad? In these times