Hello!
I have recently begun some research on the potential impact of widespread LLM use by the population on our writing, speech and ideas.
This will become a series synthesizing my research and findings. I have just published the introductory post which is a better explanation of my intent, plan, and reasons.
While anecdotal and unusable for research, I’m curious, did anyone hear notice some change in phrasing, words of choice, in someone using LLM heavily?
I would gently argue the impact of LLMs on large swaths of society has been incredibly overblown, but that’s less a reflection on hype-fueled reporting on LLMs and more a reflection on how the impact of the Internet is often overblown. Time and again we see that software is eating a lot less of the world than we think. People are still highly influenced by other real people in real places. At least where I live, you could go to an event and hang out with people and listen to live music or play a sport or do [whatever] and not even realize LLMs exist at all.
P.S. Gen Z hates AI and might be much less susceptible to changing how they talk/think based on LLMs than older people, which is a fascinating inversion of every past technology shift. Folks who are pretending the AI rollout is following the same adoption curve as the web browser or the smartphone simply don’t have the facts to back it up.
Yes, and even among other generations, AI isn’t acclaimed and well received by everyone, made evident by the number or sabotage news I’ve seen recently.
I do touch on that later on in a still WIP article. I initially wondered if LLMs have the potential to interfere with our words and meaning, and I’d argue they do. At what time scale, I do not know yet, I’m still trying to gather evidence for that. And I’m not arguing either in the subsequent articles that this will absolutely happen. Nobody knows how things will pan out exactly. The thesis is simply can AI interfere with language and meaning, and what could it look like?
You argue that people are still influenced by people and that is true, and I also write later on that part of the population rejects AI entirely and will thus resist the potential effects I discuss. But there is plenty of existing currents like anti-vaccines that show how a part of a population can have consequence on the entire population.
Then again, I decided to write those articles because I think it is worth exploring. Not because I want to spread AI doom stories or worry people but simply because that’s where my curiosity leads me.
100% of my workplace (where I spend the majority of my waking time) is heavily affected by this.
Maybe by population the vast majority of people aren’t affected; it’s a bit different ratio when looking at where the wealth is being focused.
It has the potential to fuck up everything, which is why there is so much alarm.
Agree on the potential, and it could be very insidious. Just tweak the frequency of words, tweak one down in favor of another and suddenly millions of people read a word more frequently than normal while the other is subtly less frequent. You might manage to erase a word like that without anyone crying about it if you tune it out slowly enough.
Oh believe me, there’s plenty I’m alarmed about…I’m just wondering why LLM-speak in corporate environments today would be much different than the corpospeak of past eras which affects a lot less of the general populace other than to serve as a fun target of derision & mockery (see 90s Dilbert).
Just as the “neutral/midwest dialect” tempered a lot of regional accents with the emergence of national broadcast tv and radio, using the same AI tools might cause us to lose some of our regionally/age specific idioms/etc language and give rise to an “AI dialect”. I haven’t gone looking for any hard evidence of this happening yet but Ada Palmer and Bruce Schneier have an interesting thought piece on the subject.
