Sarah, I can't thank you enough for your warm and wise words and I always feel you are taking a real interest in us writers. I did some inner work, as you suggested, and am drafting this for my AI Policy. I'm not sure where to put it as you mentioned a Pop Up menu. Will go look. Thank you!
My AI Policy
I come to Psychology for the People with ideas that come from 30 years of experience as gay activist, journalist, performer, professor and author. In attempting to create a bridge between the isolated world of depth psychology and everyday life, which is an anti-fascist approach, all ideas belong to my experience.
So I do not use AI to generate those ideas in part because the model may water down my rhetoric or invite me to create a more balanced argument, in some ways less, "gay," which I often find quite irritating, although I do take the suggestions into account.
However, I will use AI to work as I would an editor. I ask if I am being repetitive, if there are ways I can get to my message faster, and if my logic is holding up. I will ask Claude, my preferred assistant, if it might be best to break up my long statements into essays of several parts. I will also ask it to organize my footnotes and number them and to make sure I am using Chicago or APA. I will brainstorm how to deepen an idea, with needed sensation, or historical context, and then go back to the drawing board.
I use AI to explore ideas for visuals, and will use AI to take an idea of a Rainbow Field of Jacob Wrestling with the Angel based on some of my own art.
Sarah, I read the original article "Against Claudefishing" on The Substack Post shortly after it was published. I added a comment based on my own personal experience with Pangram.
For a little background: I have been a college instructor for thirteen years. When I first heard about Pangram, I decided to check it out because at the university where I teach, we use TurnItIn, and I wanted to see if Pangram would be a good second option to check a student's work for when TurnItIn indicated that a student had a high Similarity Score (compared to another student's work). Just in case TurnItIn gave the student a false positive, I was considering using Pangram, too.
I do not think Pangram is 'against writers', and I saw an interview of Max Spero (co-founder of Pangram). He is aware of false positives, and is actively working to minimize the occurrence of them. But I had it happen to me, on Pangram.
Here is my comment on The Substack Post's "Against Claudefishing" article:
"I am a college instructor. When I first heard about Pangram, I wanted to test it before deciding to use it in any official capacity. So, I copied text from a blog post that I wrote in 2009 to see how Pangram would assess it. It said it was 90% sure that it was written by AI. I was shocked!
That was enough of a test for me.
I would never choose to use Pangram as a way to assess text. My students would never deserve to have that done to them. My own writing from 2009 was a false positive!"
Thanks for this, Sarah, and for your plan to leave it up for a while so we have a chance to digest the info.
One element of Substack and AI that really bothers me is the automatic creation of an AI audio of any post. And when you want to substitute your own human reading of your post for the bot-voice, you lose the “listen” button. People who want to listen to a post instead of read it have to look to see that it’s there.
As someone for whom the human voice has been the primary focus of my work throughout my life, this is not trivial. I do plan to create new posts with my own voice as the speaker, and to also go back through all my prior posts and substitute my own reading of them. But it’s slow going, and I don’t like the loss of the simple listen button that people can easily see. It’s put up a real obstacle to my productivity on this platform. I’ve barely made any new posts in a month. I’d like to be able to get to the decision-makers at Substack to see what can be done about this.
In the meantime, I leave the artificially assigned (very poor male) voice that reads my posts in place, because I at least want people to understand immediately that it’s not me reading them that way.
Danielle, when you publish a post, you have the option of uploading an mp3 file of your own voice narration. I added my own voice narration to my first post, published on July 9th. It has a working "play" button. I have a second post that I published on July 19th, but I have not created an audio file for that one yet. I use Audacity to create, edit, and save my audio files.
I do not use AI to generate prose. However, I do use it for research. Case in point: I have a scene where the story arc needs the protagonist to get a sonogram in their hospital room with a leg in traction. Claude.ai helped me with exhaustive research on portable sonogram equipment in 2004. In case you are interested, they were the size of a refrigerator back then. The scene actually improved once the tech had to shove that beast in and out of the room. This saved me hours researching a detail that is important. I also used it to help with air bag technology, first aid protocols, medical protocols. It let me focus on the emotions, and helped me weave accurate elements into the creative heart of the story.
I don't see Substack's new AI detector as negative or being "against writers." In fact, it's a way to prevent Substack from going the way of Medium and LinkedIn, where it's just rampant with AI slop.
It's great if it makes writers pause before using AI. I would much rather have the writer's ideas and writing than anything AI could produce or polish.
It may also remind all of us that writing is hard. It's supposed to be. That's the magic. And if it takes everyone longer to produce a post or note, great! There's too much out there anyway. Let's take our time. Let's slow down and work hard!
I have been using Pangram to review each scene's first draft. There has been a ton of press about AI detection tools, and nothing is perfect. However, it does feel good when Pangram returns "100% of this text is Human Written"
Sarah, I can't thank you enough for your warm and wise words and I always feel you are taking a real interest in us writers. I did some inner work, as you suggested, and am drafting this for my AI Policy. I'm not sure where to put it as you mentioned a Pop Up menu. Will go look. Thank you!
My AI Policy
I come to Psychology for the People with ideas that come from 30 years of experience as gay activist, journalist, performer, professor and author. In attempting to create a bridge between the isolated world of depth psychology and everyday life, which is an anti-fascist approach, all ideas belong to my experience.
So I do not use AI to generate those ideas in part because the model may water down my rhetoric or invite me to create a more balanced argument, in some ways less, "gay," which I often find quite irritating, although I do take the suggestions into account.
However, I will use AI to work as I would an editor. I ask if I am being repetitive, if there are ways I can get to my message faster, and if my logic is holding up. I will ask Claude, my preferred assistant, if it might be best to break up my long statements into essays of several parts. I will also ask it to organize my footnotes and number them and to make sure I am using Chicago or APA. I will brainstorm how to deepen an idea, with needed sensation, or historical context, and then go back to the drawing board.
I use AI to explore ideas for visuals, and will use AI to take an idea of a Rainbow Field of Jacob Wrestling with the Angel based on some of my own art.
Sarah, I read the original article "Against Claudefishing" on The Substack Post shortly after it was published. I added a comment based on my own personal experience with Pangram.
For a little background: I have been a college instructor for thirteen years. When I first heard about Pangram, I decided to check it out because at the university where I teach, we use TurnItIn, and I wanted to see if Pangram would be a good second option to check a student's work for when TurnItIn indicated that a student had a high Similarity Score (compared to another student's work). Just in case TurnItIn gave the student a false positive, I was considering using Pangram, too.
I do not think Pangram is 'against writers', and I saw an interview of Max Spero (co-founder of Pangram). He is aware of false positives, and is actively working to minimize the occurrence of them. But I had it happen to me, on Pangram.
Here is my comment on The Substack Post's "Against Claudefishing" article:
"I am a college instructor. When I first heard about Pangram, I wanted to test it before deciding to use it in any official capacity. So, I copied text from a blog post that I wrote in 2009 to see how Pangram would assess it. It said it was 90% sure that it was written by AI. I was shocked!
That was enough of a test for me.
I would never choose to use Pangram as a way to assess text. My students would never deserve to have that done to them. My own writing from 2009 was a false positive!"
Finally! AI detection tools are far from perfect, but what a relief that Substack finally addresses the problem. I love the How I Make This button.
Thanks for this, Sarah, and for your plan to leave it up for a while so we have a chance to digest the info.
One element of Substack and AI that really bothers me is the automatic creation of an AI audio of any post. And when you want to substitute your own human reading of your post for the bot-voice, you lose the “listen” button. People who want to listen to a post instead of read it have to look to see that it’s there.
As someone for whom the human voice has been the primary focus of my work throughout my life, this is not trivial. I do plan to create new posts with my own voice as the speaker, and to also go back through all my prior posts and substitute my own reading of them. But it’s slow going, and I don’t like the loss of the simple listen button that people can easily see. It’s put up a real obstacle to my productivity on this platform. I’ve barely made any new posts in a month. I’d like to be able to get to the decision-makers at Substack to see what can be done about this.
In the meantime, I leave the artificially assigned (very poor male) voice that reads my posts in place, because I at least want people to understand immediately that it’s not me reading them that way.
Danielle, when you publish a post, you have the option of uploading an mp3 file of your own voice narration. I added my own voice narration to my first post, published on July 9th. It has a working "play" button. I have a second post that I published on July 19th, but I have not created an audio file for that one yet. I use Audacity to create, edit, and save my audio files.
I do not use AI to generate prose. However, I do use it for research. Case in point: I have a scene where the story arc needs the protagonist to get a sonogram in their hospital room with a leg in traction. Claude.ai helped me with exhaustive research on portable sonogram equipment in 2004. In case you are interested, they were the size of a refrigerator back then. The scene actually improved once the tech had to shove that beast in and out of the room. This saved me hours researching a detail that is important. I also used it to help with air bag technology, first aid protocols, medical protocols. It let me focus on the emotions, and helped me weave accurate elements into the creative heart of the story.
This is great news, Sarah. As always, thanks for the clarity.
I appreciate Chris Best’s position on ensuring expectations of subscribers are met.
To have 100% AI-generated content behind a paywall is not good for the cred of writers generally, or the platform overall.
I don't see Substack's new AI detector as negative or being "against writers." In fact, it's a way to prevent Substack from going the way of Medium and LinkedIn, where it's just rampant with AI slop.
It's great if it makes writers pause before using AI. I would much rather have the writer's ideas and writing than anything AI could produce or polish.
It may also remind all of us that writing is hard. It's supposed to be. That's the magic. And if it takes everyone longer to produce a post or note, great! There's too much out there anyway. Let's take our time. Let's slow down and work hard!
I have been using Pangram to review each scene's first draft. There has been a ton of press about AI detection tools, and nothing is perfect. However, it does feel good when Pangram returns "100% of this text is Human Written"
Doesn't it?