Comment by low_tech_love

“We could easily be made increasingly dependent on machines for speaking and for thinking…”

One thing that bothers me about the ubiquitous encroachment of LLMs into most areas of human writing is that it helps us write faster, but does not necessarily help us read faster. Producing large amounts of human-looking text is instantaneous, but reading (and acting upon) still takes the same amount of time. I was e.g. called to read and review a report from an academic committee with hundreds of pages that mostly looks human but also not, with pages and pages of slop after slop. I felt like I was staring into the abyss, spending hours to read something that probably took seconds for someone to produce with AI. It felt like an absolute meaningless waste of my time.

The thing is that I think people have missed the fact that the act of reading is inherently connected to the act of writing; I take the time to read something because I know that someone took the time to write it. For those who live/work by writing it seems that the act of writing has become so detached and matter-of-fact that they think “words are words” regardless of whether they were written by a human or an AI. If it helps you write faster, then why not? Like someone who used to cut trees with axes and suddenly gets a chainsaw as a gift.

But the problem is that, inevitably, we will go down the road of “if you can’t bother to write it, I won’t bother to read it” and AI will also be used to read and interpret the writings that it itself generated. So we’ll have documents with thousands of pages that are never going to be read by humans, with AI processing in both ends, so the writing will basically be a payload, a protocol, and nothing more. As such processes become the norm, we’ll be entirely dependent on the AI to both produce text and read what it produced, and become enslaved by whatever hidden (or not?) ideological bias it has been fed by its masters.