I mean its true they could have offed more context; but the link is a short read. To summarize: the author believes llm writing to be of poor quality and when he notices it on linkedin he stops reading; he suspects that other people do this too. Making the argument that if you want people to listen to you or take you seriously you shouldn’t use LLMs to write your posts.
I think both of your perspectives are valid. Yours deals with the overt consequences: “you’re a bot” should be met with indifference and blocking of those people. Thats fair, be true to yourself.
However I think it is a valid aim when talking to other people to wish to convince them of things, or to not be put in a bucket with slop.
However, I’m not sure how effective not using em dashes or having some spelling errors will be in the ling run.
Already if you ask an llm not to do the things that it stereotypically does you can make it harder to detect. You could always add a static layer that intercepts the llms output and stochastically introduces spelling errors. So for now it’s mostly, amusingly, a human social convention due to lack of specificity in prompting.
So we need more IRL community; and social spaces online must to a certain extent become smaller and be reflections of those IRL communities, if we want to have genuine human connection in the long run.
It could be as simple as a meetup in your local area so you can create a graph of “known good humans”.
This doesn’t prevent accounts from being hijacked; but if recurring, meetings, could weed out the vast majority of bots. Takes effort though and a rekindling of the atrophied social constructs for IRL gatherings. Some portion of people may simply opt out for personal preference, and thats ok.
I read the text behind the link - I just didn’t engage with it.
What’s worth keeping in mind here is that using an LLM to assist in writing could mean anything from it fixing a few spelling mistakes and adding missing punctuation to a text the author wrote themselves all the way to AI writing the entire thing from scratch. Just because someone spots an em dash doesn’t mean an LLM wrote the whole thing.
I don’t think LLMs are the issue but poor quality writing is. Generic AI output tends to just make people’s eyes glaze over so they stop reading. The same thing happens with poorly written human slop. People are only against AI use when they can detect it. It’s the good old toupee fallacy: all toupees look bad - except the ones you didn’t recognize as a toupee.
I mean its true they could have offed more context; but the link is a short read. To summarize: the author believes llm writing to be of poor quality and when he notices it on linkedin he stops reading; he suspects that other people do this too. Making the argument that if you want people to listen to you or take you seriously you shouldn’t use LLMs to write your posts.
I think both of your perspectives are valid. Yours deals with the overt consequences: “you’re a bot” should be met with indifference and blocking of those people. Thats fair, be true to yourself.
However I think it is a valid aim when talking to other people to wish to convince them of things, or to not be put in a bucket with slop.
However, I’m not sure how effective not using em dashes or having some spelling errors will be in the ling run.
Already if you ask an llm not to do the things that it stereotypically does you can make it harder to detect. You could always add a static layer that intercepts the llms output and stochastically introduces spelling errors. So for now it’s mostly, amusingly, a human social convention due to lack of specificity in prompting.
So we need more IRL community; and social spaces online must to a certain extent become smaller and be reflections of those IRL communities, if we want to have genuine human connection in the long run.
It could be as simple as a meetup in your local area so you can create a graph of “known good humans”.
This doesn’t prevent accounts from being hijacked; but if recurring, meetings, could weed out the vast majority of bots. Takes effort though and a rekindling of the atrophied social constructs for IRL gatherings. Some portion of people may simply opt out for personal preference, and thats ok.
I read the text behind the link - I just didn’t engage with it.
What’s worth keeping in mind here is that using an LLM to assist in writing could mean anything from it fixing a few spelling mistakes and adding missing punctuation to a text the author wrote themselves all the way to AI writing the entire thing from scratch. Just because someone spots an em dash doesn’t mean an LLM wrote the whole thing.
I don’t think LLMs are the issue but poor quality writing is. Generic AI output tends to just make people’s eyes glaze over so they stop reading. The same thing happens with poorly written human slop. People are only against AI use when they can detect it. It’s the good old toupee fallacy: all toupees look bad - except the ones you didn’t recognize as a toupee.