I have never seen AI-lovers and (some) uncompromising anti-AI people agree so much about something before, and actually liking and resharing each other’s posts. It’s surreal.
I'm so glad to see someone looking at it from this perspective (and not afraid to share it!). I’m actually writing an essay that shows how AI has severed the millennia-old link between good writing and good thinking, and with catastrophic repercussions. I did some due diligence to see if anyone had made a similar argument.
I found three articles. And all three were 100% AI-generated.
Please, someone tell me a better definition of irony.
As promised, I read the post. You bring forth very interesting arguments in favor. Some of those I agree with, some of those made me reflect. I would bring more nuance into the whole idea that most people opt out because they have something to hide though. I think reality isn't as conspiratorial. I might be too naive, who knows? Are some AI writers hiding behind the opt out? Sure. But others simply see this feature as being a step in the direction of outsourcing our judgment to another AI or making another tech bro even richer. That being said, your article was well written. Sometimes a little incendiary, but I assume it was for effect. Thanks for the read.
You have every right to choose to opt out. They did give you the button. But on a systemic level this is why I’m against the option. I don’t think that things like Pangram increase suspicion, I think that the cumulative effect of not knowing whether anything is AI is what makes people paranoid. Pangram, while imperfect, provides evidence. It also exonerates people at a higher rate than it incriminates them because it’s purposely engineered to have more false negatives than false positives.
But so long as the button is there, I can’t blame you for using it.
I definitely see your point. I agree that not knowing increases paranoia. I just wish people were able to make a decision for themselves without the need for such imperfect tools. I see Pangram as a lie detector. It works sometimes, but it's not admissible in court for a reason. If everyone had the sensibility you have to only use it as a way to make a decision for themselves instead of persecuting people (even if it's a minority), it wouldn't even be a problem to begin with. That said, thank you for the refreshing debate.
Thanks for writing this, really interesting take that came at just the right time. I have only today really got up to date on everything that is going on with the Pangram integration, so I am still in the process of forming my opinion on it.
I must admit that my initial instinct on reading the news was to feel dismay. I have previously seen some pretty poor uses of automated detection elsewhere, where the results are unquestioningly taken as gospel by the powers that be, even when the supporting evidence was scant. It was thus interesting to hear about your previous experience of using Pangram specifically. I do wonder if your previous use of Pangram has contributed to you being more confident regarding it's reliability than those of us who are yet to use it.
I make a point of not using AI as part of my writing process due to wider personal beliefs on the topic. However, I would be lying if I said I didn't currently have some concerns that my articles could get falsely flagged anyway. It would feel like a real kick in the teeth.
That said, I have just published a Note committing myself to giving the tool a test run; using it on my next batch of articles and then posting the results. I definitely agree that it would be great to have greater transparency in this area, even if we have to keep an eye on what else the tech bros may be using the tech for. Hopefully, after publicly testing it on my own material I will come to have as much confidence in Pangram's reliability as yourself.
I appreciate it. One thing that I keep running into with people who are scared of the false flags is that they seem to think that it will be the end of the world. That, with one AI checker declaring their work to be AI, witch hunters will suddenly jump out of the bushes and burn them.
Given that we all understand that the tools are imperfect, I highly, highly doubt that to be the case. Even metaporically.
On my feed, I restacked a post from the Pangram creator actually positively responding to a post criticizing the imperfection of the program, and talking about the ways in which they’re trying to improve it. It’s pretty educational about how to view this and I recommend taking a look.
The trouble with your final suggestion is that developing an intuitive recognition of AI text appears to be beyond a shocking number of people, more than can be explained by a lack of experience with the style (nobody lacks experience with ai text on this modern internet, especially not on here)
Last night I copied the first chapter of a Mandarin version of Mary Shelly's Frankenstein, from a 1950's physical book and pasted it into substack and ran it through the Pangram detector.
Apparently the Chinese were using AI in the 1950's and Frankenstein is 100% AI written.
People's lives being effected by something that wrong is...
I’m in category three but trying to be open-minded. I plan to turn Pangram off for now and let other bloggers be the Guinea pigs / canary in the coal mine. ;)
This is the article I wanted to write and am glad someone took it up!
It’s really helpful because the different arguments against Pangram all run cover for each other.
I have never seen AI-lovers and (some) uncompromising anti-AI people agree so much about something before, and actually liking and resharing each other’s posts. It’s surreal.
I'm so glad to see someone looking at it from this perspective (and not afraid to share it!). I’m actually writing an essay that shows how AI has severed the millennia-old link between good writing and good thinking, and with catastrophic repercussions. I did some due diligence to see if anyone had made a similar argument.
I found three articles. And all three were 100% AI-generated.
Please, someone tell me a better definition of irony.
Again, great piece.
That's insane. And dark. I'd love to read your essay on the topic.
I appreciate the interest and would very much love to hear your thoughts on it. It's here if curious: https://hereisyourbrain.substack.com/p/ai-generated-content-mind-virus
Such great stuff
As promised, I read the post. You bring forth very interesting arguments in favor. Some of those I agree with, some of those made me reflect. I would bring more nuance into the whole idea that most people opt out because they have something to hide though. I think reality isn't as conspiratorial. I might be too naive, who knows? Are some AI writers hiding behind the opt out? Sure. But others simply see this feature as being a step in the direction of outsourcing our judgment to another AI or making another tech bro even richer. That being said, your article was well written. Sometimes a little incendiary, but I assume it was for effect. Thanks for the read.
I appreciate it.
You have every right to choose to opt out. They did give you the button. But on a systemic level this is why I’m against the option. I don’t think that things like Pangram increase suspicion, I think that the cumulative effect of not knowing whether anything is AI is what makes people paranoid. Pangram, while imperfect, provides evidence. It also exonerates people at a higher rate than it incriminates them because it’s purposely engineered to have more false negatives than false positives.
But so long as the button is there, I can’t blame you for using it.
I definitely see your point. I agree that not knowing increases paranoia. I just wish people were able to make a decision for themselves without the need for such imperfect tools. I see Pangram as a lie detector. It works sometimes, but it's not admissible in court for a reason. If everyone had the sensibility you have to only use it as a way to make a decision for themselves instead of persecuting people (even if it's a minority), it wouldn't even be a problem to begin with. That said, thank you for the refreshing debate.
Thanks for writing this, really interesting take that came at just the right time. I have only today really got up to date on everything that is going on with the Pangram integration, so I am still in the process of forming my opinion on it.
I must admit that my initial instinct on reading the news was to feel dismay. I have previously seen some pretty poor uses of automated detection elsewhere, where the results are unquestioningly taken as gospel by the powers that be, even when the supporting evidence was scant. It was thus interesting to hear about your previous experience of using Pangram specifically. I do wonder if your previous use of Pangram has contributed to you being more confident regarding it's reliability than those of us who are yet to use it.
I make a point of not using AI as part of my writing process due to wider personal beliefs on the topic. However, I would be lying if I said I didn't currently have some concerns that my articles could get falsely flagged anyway. It would feel like a real kick in the teeth.
That said, I have just published a Note committing myself to giving the tool a test run; using it on my next batch of articles and then posting the results. I definitely agree that it would be great to have greater transparency in this area, even if we have to keep an eye on what else the tech bros may be using the tech for. Hopefully, after publicly testing it on my own material I will come to have as much confidence in Pangram's reliability as yourself.
I appreciate it. One thing that I keep running into with people who are scared of the false flags is that they seem to think that it will be the end of the world. That, with one AI checker declaring their work to be AI, witch hunters will suddenly jump out of the bushes and burn them.
Given that we all understand that the tools are imperfect, I highly, highly doubt that to be the case. Even metaporically.
On my feed, I restacked a post from the Pangram creator actually positively responding to a post criticizing the imperfection of the program, and talking about the ways in which they’re trying to improve it. It’s pretty educational about how to view this and I recommend taking a look.
The trouble with your final suggestion is that developing an intuitive recognition of AI text appears to be beyond a shocking number of people, more than can be explained by a lack of experience with the style (nobody lacks experience with ai text on this modern internet, especially not on here)
Nobody wins a literary agent by fracturing their prose to 160 lowercase words.
Last night I copied the first chapter of a Mandarin version of Mary Shelly's Frankenstein, from a 1950's physical book and pasted it into substack and ran it through the Pangram detector.
Apparently the Chinese were using AI in the 1950's and Frankenstein is 100% AI written.
People's lives being effected by something that wrong is...
See above to see what I said about the existence of false positives.
Pangram will get things wrong. It’s about how we choose to respond to it, as well as the repeated patterns over someone’s entire body of work.
How much of Frankenstein did you put into Substack? It couldn’t have been the entire thing?
First chapter.
Very well thought out article that has got me thinking!!
I’m in category three but trying to be open-minded. I plan to turn Pangram off for now and let other bloggers be the Guinea pigs / canary in the coal mine. ;)