Nope. The best tell isn't any single given character, but the general flow of the prose. I find "it's not X - it's Y" or "No X. No Y. Just Z and W." to be stronger tells than the mere presence of emdashes.
Hell, overuse of emoji is more of a tell than the mere presence of emdashes. The problem here is really "AI detection" services that themselves use AI to perform the service. More than one person has fed pre-AI text to such a detector only to be told "this is 75% AI" or some shit. Students who didn't use AI have been near harassed by their teachers because the teacher used some kind of AI-based AI detector, and it hallucinated. People want to believe the computer because they think "oh the computer is objective" and throw out all logic afterwards. Unfortunately, the rise of AI has brought subjectivity to computers, in the worst possible way.
You see the same thing with the family who got accosted in California by multiple police vehicles coming out of a retail store, because somebody in New Jersey lost a dealer tag and entered a partial tag number in the report, which ended up being fed to Flock's search engine, which then provided the partial search results it was asked to. Everyone along the way trusted the computer as if it was the single source of truth, but that "source of truth" was tainted by incorrect input. Literally GIGO on a life-altering scale. Other people were already being watched. It was just probability, to the point where that family probably had equal probability to win the multi-state lottery jackpot.
Go watch the Computer Chronicles episode on AI. It's very prescient. The main issue is that currently, people treat AI as if the problems brought up in that episode have been resolved, when they haven't at all.
Use of AI is fine as long as there is human review along every step of the process. Anyone using AI as a shortcut to actually doing the work they're paid to do needs to be removed from whatever situation they're involved in at the earliest convenience. Otherwise it's just an expose on the fact that humans are somehow intrinsically lazy, despite the outliers of the hardworking person (which does exist, I've known people who would move mountains to make what you need done possible, and the rest are just pivoting to AI to fake it).
Fake it until you make it, I feel, is the core of modern "AI". People also forget that AI as a field is much older than modern generative AI.
This type of AI usage is quite a good tell on personal laziness - whatever we can do to minimize the human fault state, even if that fault state is also baked into generative AI because it was invented by humans.
People have been running from their origins for decades. To me, this desire to "take human fault" out of the picture via AI is no different than centuries of people insisting that humans are separate from and not part of nature, even though our bodies run under the same physical and biological rules that all of nature does, and that is unavoidable. Objective reality only exists as an underpinning. Nothing that lives on this planet can ever perceive that reality without involving their own personal bias, and so it exists with the same relevance as an Earth-style planet even five lightyears away with an active civilization (if such a thing exists). I think that value might even be generous. There could be such a civilization a single lightyear away, and we'd never be able to reach them without a generational vessel designed to support at least one generation of explorers dying, and relying on their children (or grandchildren) to see it home.
On more than one occasion I've told an LLM "Do not use em dashes". It's one of the easiest tells.
I ended up adding this to every one of my `claude.md` files across all my projects, really annoying tbh
Nope. The best tell isn't any single given character, but the general flow of the prose. I find "it's not X - it's Y" or "No X. No Y. Just Z and W." to be stronger tells than the mere presence of emdashes.
Hell, overuse of emoji is more of a tell than the mere presence of emdashes. The problem here is really "AI detection" services that themselves use AI to perform the service. More than one person has fed pre-AI text to such a detector only to be told "this is 75% AI" or some shit. Students who didn't use AI have been near harassed by their teachers because the teacher used some kind of AI-based AI detector, and it hallucinated. People want to believe the computer because they think "oh the computer is objective" and throw out all logic afterwards. Unfortunately, the rise of AI has brought subjectivity to computers, in the worst possible way.
You see the same thing with the family who got accosted in California by multiple police vehicles coming out of a retail store, because somebody in New Jersey lost a dealer tag and entered a partial tag number in the report, which ended up being fed to Flock's search engine, which then provided the partial search results it was asked to. Everyone along the way trusted the computer as if it was the single source of truth, but that "source of truth" was tainted by incorrect input. Literally GIGO on a life-altering scale. Other people were already being watched. It was just probability, to the point where that family probably had equal probability to win the multi-state lottery jackpot.
Go watch the Computer Chronicles episode on AI. It's very prescient. The main issue is that currently, people treat AI as if the problems brought up in that episode have been resolved, when they haven't at all.
Use of AI is fine as long as there is human review along every step of the process. Anyone using AI as a shortcut to actually doing the work they're paid to do needs to be removed from whatever situation they're involved in at the earliest convenience. Otherwise it's just an expose on the fact that humans are somehow intrinsically lazy, despite the outliers of the hardworking person (which does exist, I've known people who would move mountains to make what you need done possible, and the rest are just pivoting to AI to fake it).
Fake it until you make it, I feel, is the core of modern "AI". People also forget that AI as a field is much older than modern generative AI.
using AI to detect AI - doesn't make sense to me
This type of AI usage is quite a good tell on personal laziness - whatever we can do to minimize the human fault state, even if that fault state is also baked into generative AI because it was invented by humans.
People have been running from their origins for decades. To me, this desire to "take human fault" out of the picture via AI is no different than centuries of people insisting that humans are separate from and not part of nature, even though our bodies run under the same physical and biological rules that all of nature does, and that is unavoidable. Objective reality only exists as an underpinning. Nothing that lives on this planet can ever perceive that reality without involving their own personal bias, and so it exists with the same relevance as an Earth-style planet even five lightyears away with an active civilization (if such a thing exists). I think that value might even be generous. There could be such a civilization a single lightyear away, and we'd never be able to reach them without a generational vessel designed to support at least one generation of explorers dying, and relying on their children (or grandchildren) to see it home.