While the church feared people interpreting information on their own, with LLMs it's the opposite: we fear that most interpretation of information will be done through a singular bland AI extruder. Tech companies running LLMs become the pre-press churches, with individuals depending on them to analyze and interpret information on their behalf.

This analogy is going places.

Who is "the Church" in this analogy?

refers to the gutenberg press, and mass production of printed works, threatening the siloed, ivory towers of knowledge at the time.

https://en.wikipedia.org/wiki/Printing_press#Gutenberg.27s_p...

if everyone has a bible, then who needs the church to tell you what it says.

Relying on an AI oracle to think for you is just as bad as relying on a priestly one.

> if everyone has a bible, then who needs the church to tell you what it says.

Clearly, all the protestants who burned more witches than the catholics ever did, and kept at it for centuries after the inquisition had stopped. But that's just my opinion here.

People who consider themselves exceptionally smart, who are well educated and write well, who only ever need to communicate in their native tongue ye, and who have the luxury of investing time in developing a personal writing style.

It is a good analogy. There is great concern that the unwashed masses won’t know how to handle this tool and will produce information today’s curators would not approve of.

It's an extremely poor analogy, the original point is its an information virus telling people what to think (or thinking for them). It's the exact opposite of the allowing people to think for themselves that came with the enlightenment, it's back to the days of the "church" (someone else) telling people how to think and literally writing their words for them.

The church would've LOVED everyone asking the same one-to-four sources everything. ChatGPT is literally a controllable oracle. Quite the opposite of the printing press.

"Running your own models on your own hardware" is an irrelevant rounding error here compared to the big-company models.

this would be the opposite. the llm situation may be heading back towards something similar the church age.

the church did all of the reading and understanding for us. owners of the church gobbled up as much information as it could (encouraging confessions) and then the church owners decided when, how, where and which of that information flowed to us.

This exactly is the terminal state of big AI: models that only four companies can train (on datasets only they can obtain) who also happen to own all of the ancillary services for accessing those models (because most businesses are really chatgpt in a trenchcoat). Yet the world is begging and kicking down the doors for this, just like they did for social media.