AI's Impact on rapid releasing, a decreased ability to find category arbitrage, and the shift in the balance of Author Ecosystems dominance in publishing.
I’m not at all surprised by any of this, including your use of AI to cut your writing time. What happens when everyone is doing this? How many substacks will go out a day? How many blog posts? Let’s hope we all take a breath and stop crowding the marketplace. I find it ironic that now that the marketplace has been flooded with “fast” books in order to gain algorithmic dominance, people are talking about, gasp, quality. The flooders are about to be out-flooded and NOW all of a sudden quality matters. Sigh. I’m tired. I’ve also been working on my craft for 40 years. Maybe my era is coming?
I think there’s a huge opportunity in that, though. I wrote about it in this article so I won’t rehash it. I also wrote about it in this recent article.
Joanna Penn calls herself the artisan ai author so I do think there is space for it even in most author businesses. I especially think that is true if you edit heavily and only use it for drafting ideas.
Finally, AI is at its best when you the user knows exactly what you want and it just has to generate it. I spend most of my time with AI trying to get it to do the thing so already know I want, but can’t generate myself for some reason, and that knowledge only comes which skill and that only comes from learning the craft.
Unquestionably, AI could be a valuable tool. But I think it might be worth tapping the breaks a time or two before committing too heavily to it. To be clear, I don't mean the next part as in any way a criticism of authors who are already making use of AI in one way or another. We all have different circumstances and use AI in a variety of ways. Please don't interpret my criticism of AI as condemnation of people who use it.
That said, the large language models on which various forms of AI are built require enormous amounts of data, much of which was obtained on the assumption that using any intellectual property the developers could get their hands on was fair use. But at best, it's a gray area. It's true AI training isn't prohibited by copyright law--because such training didn't exist when the law was written. But AI doesn't fall within any of the traditional examples of fair use and is, in fact, radically different from anything claimed as fair use in the past. Several lawsuits are currently in progress on that specific issue. If AI companies lose even one of them, the result could be a considerable restriction in the use of new material for training. And even AI developers admit that AI needs to be fed constantly, and it can't be fed on its own output. The latter is already becoming a concern as the internet--one of AI's feeding grounds--because of the number of people throwing up more or less unedited AI work.
In other words, AIs training method may run afoul of legal challenges, and AI may decline in quality as a result. In this respect, the EU's AI Act is relevant. Besides prohibiting some uses of AI entirely, it requires transparency about the training processes and adherence to EU copyright law. That could have implications for AIs trained on the anything-we-want-to-use-is-ours theory of fair use, though companies have considerable time to make their current products compliant.
In other words, I wouldn't put too many eggs in the AI basket until legislatures and courts finish chewing on what AI is allowed to do and how it's allowed to be trained. It could a great technology ready to serve us--or it could be the dirigible industry waiting for its Hindenburg moment. I wouldn't become too dependent on it just yet.
In thinking about how I would use AI, I would try to restrict to applications that don't put someone out of a job. I've heard of people using AI as a brainstorming tool, something to bounce ideas off of, etc. None of those uses put someone else out of work. But I wouldn't use it to replace a human editor. (It isn't yet ready for that, anyway.) For images, which I have experimented with already, I use a company that at least provides compensation for artists whose work is used in the training--I want to encourage that kind of model. Anyway, I'd use it for Substack post images. I couldn't afford to hire an artist to do all those images for me, so I'm not putting anyone out of work by using AI. I don't imagine many authors on Substack have professional illustrators on call. However, I will continue to use my cover designer rather than AI. I'm not faulting anyone who feels differently. That's just what feels best to me.
I'd love to use AI to speed up my writing process. I understand the appeal of that. But I'm waiting until the legal situation becomes clearer.
Anyway, thanks for raising these important questions, Russell. Love it or hate it, AI will have an impact on us, and we need to consider how to deal with that.
I'll use AI to test my logic in a post or ask if it's hitting right. I'm working on a series where I asked it a bunch of questions on a common theme, how to turn it into a series, and then write the posts.
Now I'm going through and cleaning them up because it's sterile feeling. May even toss the bullet points.
Fascinating! My assessment of the AI article is that I admittedly felt like it was watered-down in some way that I couldn’t quite pinpoint. I read the first third, felt like I got the picture and it was losing my attention, so I skipped to the end and was surprised by the reveal that AI was involved. I would normally not comment about having not read the entire Substack BUT it feels relevant as you assess if it was worth the time spent expanding a shorter article into a lengthier one with ChatGBT
I think it would have been better as a 1500 word article. I kept finding new angles, and then I didn't smooth it out as well as I wanted, but this is good feedback. I don't particularly want to do it this way. I have no aversion to doing it either if it does the thing I want it to do.
On the points of the article: I'm just getting started but so far believe grassland is a close fit for my personality. However, I've recently considered I can be a bit of a forest if I choose to put energy into it. So, I find this shift exciting and timely. Here's hoping I can implement what I am learning from you!
On the use of Chat - honestly - I skimmed. I think using AI to expand content may not be best use if you end up with material that could be improved by condensing. Hmmm.
Generally, I am finding AI useful to add more material, but not necessarily more interest.
This makes sense. I think AI is good for lists and summarizing things, but it’s very bland. I think people on Substack want non-bland, but it might work better on Medium.
I so appreciate your no-bullshit honesty, Russell. That’s one of the reasons I keep reading your stuff. I have often believed (and, yes, this is going to sound a little woo-woo and not business-minded at all), that the abundance we put out into the world comes back to us, though the avenue is not always clear. Perhaps this cycle is part of the mystery you write about here. I am also appreciating the way you take care of your health and wellbeing in your business decisions. Thanks again for your perspective.
I will tell you that the more successful I get, the more I believe in planned serendipity and all that stuff about abundence. I thought it would root me more in reality, but it has done the opposite.
The article was good except it repeated the same points over and over. But I also read a version of this on Facebook and this didn't really expand on the original post I read or add much except quotes.
Hrm. Not sure I agree with that. I added quotes and the part about content farms which wasn’t in the original. I also added several additional pieces about how to use it for each ecosystem. I think there is plenty to criticize, but I am not sure not adding from the original is one of them.
I just hear about authors saying the AI programs don’t actually work well and it’s faster to write books yourself. What are these AI novel programs that actually work? I’ve tried a few and they aren’t great.
Good only for very specific tasks that don’t really speed up the process.
Non-fiction though… def AI can help, but fiction, less convincing.
IDK. Most of my friends now use Claude, but there’s also Rexy. The people I quote in the article would be the people to ask. Kevin and Steph know the landscape better than anyone.
I tried the paid version of Claude recently, hoping to be impressed, but it wasn’t impressive and I find myself using the free version of chatGPt more, though usually for random questions.
I definitely use it for blurbs and marketing material for book launches, and for anything where I know what I want but don't have the energy to generate it.
I have it take our podcast transcriptions and create show notes. It's pretty good at that. Again, I think you have to know the outcome you want to make it have any value. If you don't, then you don't know how to manipulate it to your advantage.
I don't know. I feel like there's a lot of writing that benefits from mediocrity. Like, a book summary benefits from it being middle of the road. Movie times benefit from being bland. IF something benefits from being bland and informative, I think AI has a place there. Creating in universe names and products is another one where like...I don't care what that character's name is, and my brain doesn't want to spend the effort, or what a name of a tavern should be named.
The best use case in fiction is to create a puzzle box you can't solve, since you are always limited by what you know already.
The “twist” to this post was funny to me because I started wondering in the first few paragraphs if it had been written by AI. I don’t know why; nothing was written incorrectly. It just felt ChatGPT-ish to me, so I skimmed to the end—and voila!
And I know the piece was massaged by you, as you said. By human hands. It still felt off to me. I’ve tried the same thing (having ChatGPT generate something, and having me go over it myself, adding/substracting, etc.), and while it’s definitely a useful tool, it’s still not yet feeling totally organic. My friend wrote a piece that was helped out by ChatGPT, and it was anthologized in Best American Essays. She wrote about the experience here: https://www.wired.com/story/confessions-viral-ai-writer-chatgpt/
I’m not at all surprised by any of this, including your use of AI to cut your writing time. What happens when everyone is doing this? How many substacks will go out a day? How many blog posts? Let’s hope we all take a breath and stop crowding the marketplace. I find it ironic that now that the marketplace has been flooded with “fast” books in order to gain algorithmic dominance, people are talking about, gasp, quality. The flooders are about to be out-flooded and NOW all of a sudden quality matters. Sigh. I’m tired. I’ve also been working on my craft for 40 years. Maybe my era is coming?
I think there’s a huge opportunity in that, though. I wrote about it in this article so I won’t rehash it. I also wrote about it in this recent article.
https://www.theauthorstack.com/p/the-great-medium-vs-substack-debate
I also covered it at the top of this digest.
https://www.theauthorstack.com/p/digest-cyborgs-vs-scabs?r=571jf
Joanna Penn calls herself the artisan ai author so I do think there is space for it even in most author businesses. I especially think that is true if you edit heavily and only use it for drafting ideas.
Finally, AI is at its best when you the user knows exactly what you want and it just has to generate it. I spend most of my time with AI trying to get it to do the thing so already know I want, but can’t generate myself for some reason, and that knowledge only comes which skill and that only comes from learning the craft.
Yes I’m aware of Joanna Penn. I guess I’m just agreeing with what you and ChatGPT wrote…while acknowledging the irony of this turn of events.
Yeah, I get that.
Keep up the good work.
Unquestionably, AI could be a valuable tool. But I think it might be worth tapping the breaks a time or two before committing too heavily to it. To be clear, I don't mean the next part as in any way a criticism of authors who are already making use of AI in one way or another. We all have different circumstances and use AI in a variety of ways. Please don't interpret my criticism of AI as condemnation of people who use it.
That said, the large language models on which various forms of AI are built require enormous amounts of data, much of which was obtained on the assumption that using any intellectual property the developers could get their hands on was fair use. But at best, it's a gray area. It's true AI training isn't prohibited by copyright law--because such training didn't exist when the law was written. But AI doesn't fall within any of the traditional examples of fair use and is, in fact, radically different from anything claimed as fair use in the past. Several lawsuits are currently in progress on that specific issue. If AI companies lose even one of them, the result could be a considerable restriction in the use of new material for training. And even AI developers admit that AI needs to be fed constantly, and it can't be fed on its own output. The latter is already becoming a concern as the internet--one of AI's feeding grounds--because of the number of people throwing up more or less unedited AI work.
In other words, AIs training method may run afoul of legal challenges, and AI may decline in quality as a result. In this respect, the EU's AI Act is relevant. Besides prohibiting some uses of AI entirely, it requires transparency about the training processes and adherence to EU copyright law. That could have implications for AIs trained on the anything-we-want-to-use-is-ours theory of fair use, though companies have considerable time to make their current products compliant.
In other words, I wouldn't put too many eggs in the AI basket until legislatures and courts finish chewing on what AI is allowed to do and how it's allowed to be trained. It could a great technology ready to serve us--or it could be the dirigible industry waiting for its Hindenburg moment. I wouldn't become too dependent on it just yet.
In thinking about how I would use AI, I would try to restrict to applications that don't put someone out of a job. I've heard of people using AI as a brainstorming tool, something to bounce ideas off of, etc. None of those uses put someone else out of work. But I wouldn't use it to replace a human editor. (It isn't yet ready for that, anyway.) For images, which I have experimented with already, I use a company that at least provides compensation for artists whose work is used in the training--I want to encourage that kind of model. Anyway, I'd use it for Substack post images. I couldn't afford to hire an artist to do all those images for me, so I'm not putting anyone out of work by using AI. I don't imagine many authors on Substack have professional illustrators on call. However, I will continue to use my cover designer rather than AI. I'm not faulting anyone who feels differently. That's just what feels best to me.
I'd love to use AI to speed up my writing process. I understand the appeal of that. But I'm waiting until the legal situation becomes clearer.
Anyway, thanks for raising these important questions, Russell. Love it or hate it, AI will have an impact on us, and we need to consider how to deal with that.
Thanks for your insightful comments. I really like the way you are thinking.
I’ve always liked the way you approach issues as well.
I always knew my life calling was to be a forest on the edge of grasslands :-)
Yaaas. I love this so much
I'll use AI to test my logic in a post or ask if it's hitting right. I'm working on a series where I asked it a bunch of questions on a common theme, how to turn it into a series, and then write the posts.
Now I'm going through and cleaning them up because it's sterile feeling. May even toss the bullet points.
Cool!
Fascinating! My assessment of the AI article is that I admittedly felt like it was watered-down in some way that I couldn’t quite pinpoint. I read the first third, felt like I got the picture and it was losing my attention, so I skipped to the end and was surprised by the reveal that AI was involved. I would normally not comment about having not read the entire Substack BUT it feels relevant as you assess if it was worth the time spent expanding a shorter article into a lengthier one with ChatGBT
I think it would have been better as a 1500 word article. I kept finding new angles, and then I didn't smooth it out as well as I wanted, but this is good feedback. I don't particularly want to do it this way. I have no aversion to doing it either if it does the thing I want it to do.
On the points of the article: I'm just getting started but so far believe grassland is a close fit for my personality. However, I've recently considered I can be a bit of a forest if I choose to put energy into it. So, I find this shift exciting and timely. Here's hoping I can implement what I am learning from you!
On the use of Chat - honestly - I skimmed. I think using AI to expand content may not be best use if you end up with material that could be improved by condensing. Hmmm.
Generally, I am finding AI useful to add more material, but not necessarily more interest.
This makes sense. I think AI is good for lists and summarizing things, but it’s very bland. I think people on Substack want non-bland, but it might work better on Medium.
Yes to non-bland.
I so appreciate your no-bullshit honesty, Russell. That’s one of the reasons I keep reading your stuff. I have often believed (and, yes, this is going to sound a little woo-woo and not business-minded at all), that the abundance we put out into the world comes back to us, though the avenue is not always clear. Perhaps this cycle is part of the mystery you write about here. I am also appreciating the way you take care of your health and wellbeing in your business decisions. Thanks again for your perspective.
You're welcome :)
I will tell you that the more successful I get, the more I believe in planned serendipity and all that stuff about abundence. I thought it would root me more in reality, but it has done the opposite.
The article was good except it repeated the same points over and over. But I also read a version of this on Facebook and this didn't really expand on the original post I read or add much except quotes.
Hrm. Not sure I agree with that. I added quotes and the part about content farms which wasn’t in the original. I also added several additional pieces about how to use it for each ecosystem. I think there is plenty to criticize, but I am not sure not adding from the original is one of them.
I just hear about authors saying the AI programs don’t actually work well and it’s faster to write books yourself. What are these AI novel programs that actually work? I’ve tried a few and they aren’t great.
Good only for very specific tasks that don’t really speed up the process.
Non-fiction though… def AI can help, but fiction, less convincing.
IDK. Most of my friends now use Claude, but there’s also Rexy. The people I quote in the article would be the people to ask. Kevin and Steph know the landscape better than anyone.
I tried the paid version of Claude recently, hoping to be impressed, but it wasn’t impressive and I find myself using the free version of chatGPt more, though usually for random questions.
I definitely use it for blurbs and marketing material for book launches, and for anything where I know what I want but don't have the energy to generate it.
Yes, it’s good for quick tasks like that, but even asking it to reorganize long notes it often messes everything up.
I have it take our podcast transcriptions and create show notes. It's pretty good at that. Again, I think you have to know the outcome you want to make it have any value. If you don't, then you don't know how to manipulate it to your advantage.
Like it’s impressive in that it exists, but it’s not able to complete tasks to a high enough standard and more just frustrates me with its mediocrity.
I don't know. I feel like there's a lot of writing that benefits from mediocrity. Like, a book summary benefits from it being middle of the road. Movie times benefit from being bland. IF something benefits from being bland and informative, I think AI has a place there. Creating in universe names and products is another one where like...I don't care what that character's name is, and my brain doesn't want to spend the effort, or what a name of a tavern should be named.
The best use case in fiction is to create a puzzle box you can't solve, since you are always limited by what you know already.
Yeah, I think I’m asking it to do the wrong tasks.
What do you mean by “puzzle box”?
any kind of puzzle or situation.
The “twist” to this post was funny to me because I started wondering in the first few paragraphs if it had been written by AI. I don’t know why; nothing was written incorrectly. It just felt ChatGPT-ish to me, so I skimmed to the end—and voila!
And I know the piece was massaged by you, as you said. By human hands. It still felt off to me. I’ve tried the same thing (having ChatGPT generate something, and having me go over it myself, adding/substracting, etc.), and while it’s definitely a useful tool, it’s still not yet feeling totally organic. My friend wrote a piece that was helped out by ChatGPT, and it was anthologized in Best American Essays. She wrote about the experience here: https://www.wired.com/story/confessions-viral-ai-writer-chatgpt/