I suspect a long term consequence of AI adoption in workplaces is that document reviews become worthless.
-
I suspect a long term consequence of AI adoption in workplaces is that document reviews become worthless. An initial problem some have begun to encounter is where the author didn’t even read the document AI produced.
Some AI-forward companies have an even more pernicious problem of people not reading the document but instead asking AI what questions to raise to make it look like they did.
Over time this will become a waste of time with only the AI companies benefiting.
@carnage4life I just don't review AI documents. My autistic brain feels very uneasy when reading the slop so I detect quite quickly that it's AI.
-
I suspect a long term consequence of AI adoption in workplaces is that document reviews become worthless. An initial problem some have begun to encounter is where the author didn’t even read the document AI produced.
Some AI-forward companies have an even more pernicious problem of people not reading the document but instead asking AI what questions to raise to make it look like they did.
Over time this will become a waste of time with only the AI companies benefiting.
The way out of this mess might be adopting more efficient writing styles that are better aligned with reader needs, at least in the technical writing space.
Much of the reason why LLMs are successful is because our current style of writing is filled with repetitive and predictable fluff. The entropy of typical writing, a measurement of how efficiently information is packed in, is atrocious. Most of this fluff exists because of societal expectations for what writing is supposed to feel like.
Perhaps we just need to get a lot better about conveying information that the reader needs without all that extra fluff.
As Antoine de Saint Exupéry said, "perfection is attained not when there is nothing more to add, but when there is nothing more to remove."
Or, as another great philosopher once postulated, "why use lot word when few word do trick?"
-
@carnage4life They have been for years. A not uncommon thing is being tasked with writing a document and having to go through a weeks to months long development and review process, only to find out that your boss' boss wrote a three line email that made it unnecessary an hour after the meeting you were tasked in.
@drwho @carnage4life Yep, I think, most technical documents I‘ve written in my ten years as an engineer were never read by anyone else but me and the person I asked to proofread.
-
I suspect a long term consequence of AI adoption in workplaces is that document reviews become worthless. An initial problem some have begun to encounter is where the author didn’t even read the document AI produced.
Some AI-forward companies have an even more pernicious problem of people not reading the document but instead asking AI what questions to raise to make it look like they did.
Over time this will become a waste of time with only the AI companies benefiting.
@carnage4life In 3-5 years, skill atrophy and simulated busywork will hit those "AI-forward" companies very hard. Management will scramble to design efficient processes around ineffective (sloppy) tasks performed by employees who will be the textbook definition of "learned helplessness".
-
True, but it's just one of many examples of perverse incentives in writing. I used to work in a public sector research organization, and the internal joke was "the job isn't done until the paper weighs good." But most of the perceived need for that voluminous writing is internally driven. Most clients don't read much past the execsum.
If we can remove these perverse incentives and change societal expectations a bit then maybe we won't need all of this boilerplate that is so generic that robots can do it.
-
I suspect a long term consequence of AI adoption in workplaces is that document reviews become worthless. An initial problem some have begun to encounter is where the author didn’t even read the document AI produced.
Some AI-forward companies have an even more pernicious problem of people not reading the document but instead asking AI what questions to raise to make it look like they did.
Over time this will become a waste of time with only the AI companies benefiting.
@carnage4life For a long while there is intrinsic part of every org that mandates a noop documents to be produced. They have no value on their own for the goals of the organisation, but to fulfil the needs of some checklist process.
If a person uses [a tool] to shorten the time&resources spent on said task, it should be a net positive for said organisation. -
@carnage4life For a long while there is intrinsic part of every org that mandates a noop documents to be produced. They have no value on their own for the goals of the organisation, but to fulfil the needs of some checklist process.
If a person uses [a tool] to shorten the time&resources spent on said task, it should be a net positive for said organisation.@carnage4life Obviously #1: we can argue on what the "useless thing" is and what just suffers from Chesterton's fence-like situation.
Obviously #2: This solution does not address the root cause, and thus does not help in broader sense.
Obviously #3: Yep, capitalistic AI hypers do profit on this. But IMO correct solution is not to "ban AI", but rather critically look at the tasks at hand and have a deep thought of why they should not be eliminated / automated. -
@drwho @carnage4life Yep, I think, most technical documents I‘ve written in my ten years as an engineer were never read by anyone else but me and the person I asked to proofread.
@minus @carnage4life None of mine have been, either. Even my proofreaders rubberstamped it.
I don't know why I bother writing documentation for anyone. Nobody every reads it.
-
@carnage4life For a long while there is intrinsic part of every org that mandates a noop documents to be produced. They have no value on their own for the goals of the organisation, but to fulfil the needs of some checklist process.
If a person uses [a tool] to shorten the time&resources spent on said task, it should be a net positive for said organisation.@MichalBryxi @carnage4life The problem is that LLMs have a huge cost to produce those valueless documents. If the organization had to bear the actual costs involved, LLM usage would be *banned* so fast it'd unscrew your head from your neck.
Cheaper and simpler to just re-evaluate whether you need that particular step in the checklist or not.
-
@MichalBryxi @carnage4life The problem is that LLMs have a huge cost to produce those valueless documents. If the organization had to bear the actual costs involved, LLM usage would be *banned* so fast it'd unscrew your head from your neck.
Cheaper and simpler to just re-evaluate whether you need that particular step in the checklist or not.
@tknarr @carnage4life Agree with your second point: https://mastodon.world/@MichalBryxi/116037210000728401
Not so much with the first. The reality we live in is that that price has been already paid for current generation of models. And using them locally costs the org next to nothing.
-
I suspect a long term consequence of AI adoption in workplaces is that document reviews become worthless. An initial problem some have begun to encounter is where the author didn’t even read the document AI produced.
Some AI-forward companies have an even more pernicious problem of people not reading the document but instead asking AI what questions to raise to make it look like they did.
Over time this will become a waste of time with only the AI companies benefiting.
@carnage4life we are using less AI for writing docs in the startup I work at due to this dynamic. I find it more effective as a coach for myself to clarify thinking, but most writing is best done by the individual with context & intent directly.
-
R relay@relay.mycrowd.ca shared this topic
