I started a new role at a multi national corporation a couple months ago. My “onboarding” to try to get caught up on a new project involved shifting through thousands of pages and slides of AI generated content that had been created on this project… Which ironically is about creating systems to try to make AI return more relevant results. The company have paid me a small fortune in the hours I spent reading material that held no value but to make someone somewhere look like they were being productive at some point in time.
I’ve been dealing with this at my job too, I’ve been trying to review technical documents “written” by another developer who is 100% using AI to generate the technical docs. I find the docs totally unreadable. I don’t have much in the way of review comments except “can you re-write this part in a more human-readable way”? * 1000. Then the re-writes happen but again it’s clear the prompt was just “AI, re-write this in a more human-readable way” and it’s still just as unreadable just in a different format.
.
But then again, I’m always unsure, are we at the point where we aren’t expected to be able to read these docs, or are they just meant to be ingested by a different AI agent to then use them to do something else ?
or are they just meant to be ingested by a different AI agent
Given that these models are (ideally) trained on good, well-documented human code, I’m guessing that poor documentation will also lead to poor results, so they’re probably not meant for LLM input, but of course, they’ll probably end up as it anyway.
I read a good article on this sort of thing a month or two ago… One executive got an several page email from a worker that was obviously ai generated. So he had his own ai read it and write a reply. This created a sort of table flip moment for him as he realized the company was paying for him and the employee and the tokens to generate both emails and nothing of value had been produced from any of it.
I started a new role at a multi national corporation a couple months ago. My “onboarding” to try to get caught up on a new project involved shifting through thousands of pages and slides of AI generated content that had been created on this project… Which ironically is about creating systems to try to make AI return more relevant results. The company have paid me a small fortune in the hours I spent reading material that held no value but to make someone somewhere look like they were being productive at some point in time.
I’ve been dealing with this at my job too, I’ve been trying to review technical documents “written” by another developer who is 100% using AI to generate the technical docs. I find the docs totally unreadable. I don’t have much in the way of review comments except “can you re-write this part in a more human-readable way”? * 1000. Then the re-writes happen but again it’s clear the prompt was just “AI, re-write this in a more human-readable way” and it’s still just as unreadable just in a different format.
.
But then again, I’m always unsure, are we at the point where we aren’t expected to be able to read these docs, or are they just meant to be ingested by a different AI agent to then use them to do something else ?
Given that these models are (ideally) trained on good, well-documented human code, I’m guessing that poor documentation will also lead to poor results, so they’re probably not meant for LLM input, but of course, they’ll probably end up as it anyway.
I read a good article on this sort of thing a month or two ago… One executive got an several page email from a worker that was obviously ai generated. So he had his own ai read it and write a reply. This created a sort of table flip moment for him as he realized the company was paying for him and the employee and the tokens to generate both emails and nothing of value had been produced from any of it.