Are you worried about the impact of generative AI on your writing or other creative expression? Delighted by ChatGPT et al.? Scared? Dismissive? All opinions welcome!
In particular, please note whether or not you’ve used AI. If you’ve tried one of the chatbots, say which one, for what type of writing, and at what point in the writing process.
One reason I’m asking this now is that Apple, at its developers’ conference last week, announced it’s incorporating generative AI into Siri and other apps. So, it’s coming soon to iPhones – a chatbot that feels more conversational than the old Siri and can draw on information about you.
My knee-jerk response to Siri searching my emails or text messages to respond to questions is Yuck! Yet according to a Hardfork interview with Tripp Mickle of the New York Times, Apple appears to be focusing on privacy of customer data. It also seems its new apps will allow you to use AI to rewrite text you’ve written on your phone but not to generate anything new. That sounds like a promising guardrail other AI companies have ignored.
I have concerns about generative AI for writers, but I’m also looking for other viewpoints to challenge me.
I think almost everyone has concerns. My chief concern is that fear will triumph over reason. Because AI mirrors human thought patterns as revealed in our language, AI is weakest in determining truth-functions (weaker than humans I’ve found when I experimented with Claude 3 Opus to find logical truth in truth tables as Wittgenstein explained in his 1923 book). In the end, it’s not AI I fear. It’s human nature.
I agree, Terry, it's not the technology I fear so much as the way it's used thoughtlessly by humans with little feel for writing. The technology is also controlled by just a very few for-profit companies. Apple does it's own version of profit-chiseling, but at least the company knows its brand replies on designing an effective user experience. Yet once AI is embedded in our phones and email, it will be all too easy to give up thinking for ourselves. That's the existential threat troubling me.
You may be giving AI more credit than it deserves. If we continue to produce graduates with a right answer mentality and an automaton disposition toward epistemology, AI could be a tool of oppression. But schools made on the factory model can’t be expected to equip human beings for a life of confident, rational, ethical thought. Apple and AI are complicit, motivated by the goal of profit, but I fear a school system motivated by efficiency even more. The longterm solution rests with educators, not lawyers and corporations. Let’s face it: We have a population of people who learned to think in a factory where parroting an answer, spelling it right and saying it right, gets you a good grade. This is why AI is such a threat right now to the status quo.
I have not touched the stuff, mainly because of Cory Doctorow's take on it. I forget the exact wording, but the gist of what he said is that AI is not good enough to do your job but it is good enough to be used to scam your boss into firing you and hiring it to fail to do your job. I was a software professional. I know just how stupid these supposedly intelligent machines really are. The people who call AI "enhanced autocorrect" have got it right, and we've all been betrayed by autocorrect.
Has autocorrect improved the world one iota? Sometimes it's nice to have my errors magically fixed, except most of the time they aren't really fixed or the wrong "errors" get changed — and then I'm irritated. If I counted up all those momentary flashes of irritation, it might well look like a waste of time as well emotion. I like Cory Doctorow's take, too, and I think he's especially on point in talking about the inevitable "enshitification" of digital platforms (a decline that will come to Substack, too).
I'm going to agree with you except for that word "inevitable". When we were students, the process (basically, buy a business with a reputation for quality, drop quality while maintaining prices, rake in the bucks for a little while, and then shut it down or sell out to an ignoramus just before the bottom drops out) was not unheard of but it was a lot less common. That's because there were laws in place to minimize that sort of thing and those laws were enforced. Let's not fall into the trap of mistaking current conditions for the laws of nature.
I have had 2 experiences with using AI deliberately, and while they were successful, I still say hang the rest.
1) I used ChatGPT with my college freshmen. It is much easier to teach a group of college kids how to workshop a piece of writing when the example piece you're using was spit out by a computer. There's no emotional connection there, nothing to worry about in terms of emotional investment. And, because it's written by AI, there were obviously things wrong with it, so they got to practice giving critiques and feedback.
2) My current day job uses AI to diagnose what's wrong with computers, and to diagnose what was going on with mine last week. As far as I'm concerned, this is what AI should be used for and nothing else.
Thanks, Marissa. I've considered doing what you say with students, workshopping AI pieces to surface obvious problems. Interesting that you say this helps undercut some of the emotional baggage of workshopping personal pieces – that could be a benefit. But then it seems you come down on only using AI to debug computers. I'd be happy to hear more about your experience of those student workshops with AI pieces.
My gut tells me part of the emotional-investment problem has to do with outmoded models for writing workshops, in which students end up judged on a hot seat. The process is very demotivating. In my journalism courses, I generally opt to have students spend much more class time "workshopping" published pieces by well known writers. I find it gets them used to focusing on specific criticisms and underscores that even the professionals make decisions (or mistakes) that not all readers like.
I finally have time to respond to this! I have tried a couple of the bots, including Claude, ChatGPT, and Sudowrite. Ethically, I don't like ChatGPT or Sudowrite, so I don't use those. I only use Claude, but during the editing stages. Since I'm a discovery writer and writing is how I process, it makes no sense to use AI to "help" me write.
One of these days, I want to write another article on generative AI, but my personal thought is that while it may raise the baseline (as in, help those who have little/no talent in writing "write"), it won't do much more for seasoned or more gifted writers who have truly unique voices than, Grammerly, for example.
Interesting observations, Tiffany - I’m not convinced these tools can raise the baseline level for anyone unless they already know how to write. Chatbots can generate polished-looking prose, and in some circumstances that may be enough, but for any kind of writing that’s meant to engage readers or dig into a topic, I don’t see it yet.
But that’s me. I’d love to hear more about how you use Claude - I agree about not using this stuff for first drafts, because the writing process is about figuring out what you want to say. So, at what point does Claude become useful to you? And ethical issues, yes, I’ve been thinking hard about those, and I wonder why ChatGPT and Sudowrite don’t meet your standards there. I think I know, but it would be good to hear more about that.
What I mean by "raising the baseline" isn't that their skill will get better, but that it'll allow people who had no skill in writing to be able to tell stories that they otherwise wouldn't have been able to tell. I don't know if that is a good or bad thing. In some ways, I struggle to see how using generative AI to write stories differs from using a ghostwriter.
I share a bit about how I use Claude here: https://tiffanychu.substack.com/p/the-case-for-ai-in-the-creative-life, but I don't use it for plot direction anymore like I mention in the article. I've found that my own ideas are still better, and not having the freedom to come up with them myself is stifling. However, I do use it to comment on things like how an article might land with readers, pacing issues, incomplete threads, etc. I definitely use it for SEO-related things haha. As for fiction, I ask it to point out pacing issues, unnatural dialogue, potentially missing scenes to complete story/character arcs, and inconsistencies/plot holes. I always specify that I don't want it to fix anything for me, only point things out so I can do it myself. I'm often an underwriter, so it's helpful for me to see where something is underdeveloped.
I comment on the ChatGPT issue in the post I linked, but I admit I haven't done much follow-up research since then. I've heard tell that OpenAI is working on licensing deals with authors and publishers, so if things start looking more above board with them, my stance might change? We'll see. I'm still iffy about their privacy policy and how they might use my input/prompting if I use ChatGPT. With Sudowrite, I think just the principle behind it grates on me, since it's solely an AI story generator and not much else..
So I'm a copywriter. I'm currently in the process of applying and interviewing for a new full time job. Understanding how to use AI and being able to explain how I use it in my writing process is now essential. That happened so fast. It wasn't like that a year ago.
For one company I just applied to they asked on the application - "Do you know how to use Chat GPT and please explain how you use it in your copy process."
I know a copywriter at that company so we had chatted before I filled out the application and I had asked her about what they use, do they have guidelines or does her team openly talk about it. So I knew I had to show I had that skill.
With another company I had to do a standard writing assignment, and then they asked me to make a video talking about my copy process (videos pre-interview stage are pretty standard now too.) And again, I have connections to other copywriters at that company and we had talked about what they use, how they talk about it. So I definitely used an AI app called Claude and then talked about how I used Claude for the assignment in my video. Claude is very nuanced with tone and voice so I used it almost like a copy editor at the end of my writing process to tighten things up.
So there you go. When I write for my newsletter or I write a poem or I'm writing something personal. No, not usually. But I will run some of those pieces through Claude (never the poems - those are sacred) and lift a few sentences that make things clearer and flow better. It depends on my mood, the time I have, and what it's for.
This is a very loooong answer -- so the short answer is for work - yes. That skill is now essential. For my own creative writing - kinda sorta maybe sometimes.
Jen, this on-the-ground feedback is so excellent! I'd love to talk with you more about how you see your job of copywriter evolving. I find some hope in the process you're describing, and I've been thinking a lot about the most effective way to use AI as a tool for journalists (and how I should be preparing students). I also agree that Claude is currently the bot of choice, partly because of the ethical guardrails Anthropic has put into its language model but mostly because it doesn't serve Sam Altman of Open AI 's megalomaniac goals :-)
Glad it was helpful. This round of applications has been eye opening around AI. The landscape is changing. So Claude is my 1st choice and something I'm learning how to leverage to position as a new skill set.
Haven’t used it and have seen my last book pirated, thanks to AI. I have many concerns but the door in my mind is ajar.
I like the image of a door ajar, Rona. Really sorry to hear about your book, though. That's happening a lot.
It's endemic. We shut one down, another one pops up.
I think almost everyone has concerns. My chief concern is that fear will triumph over reason. Because AI mirrors human thought patterns as revealed in our language, AI is weakest in determining truth-functions (weaker than humans I’ve found when I experimented with Claude 3 Opus to find logical truth in truth tables as Wittgenstein explained in his 1923 book). In the end, it’s not AI I fear. It’s human nature.
I agree, Terry, it's not the technology I fear so much as the way it's used thoughtlessly by humans with little feel for writing. The technology is also controlled by just a very few for-profit companies. Apple does it's own version of profit-chiseling, but at least the company knows its brand replies on designing an effective user experience. Yet once AI is embedded in our phones and email, it will be all too easy to give up thinking for ourselves. That's the existential threat troubling me.
You may be giving AI more credit than it deserves. If we continue to produce graduates with a right answer mentality and an automaton disposition toward epistemology, AI could be a tool of oppression. But schools made on the factory model can’t be expected to equip human beings for a life of confident, rational, ethical thought. Apple and AI are complicit, motivated by the goal of profit, but I fear a school system motivated by efficiency even more. The longterm solution rests with educators, not lawyers and corporations. Let’s face it: We have a population of people who learned to think in a factory where parroting an answer, spelling it right and saying it right, gets you a good grade. This is why AI is such a threat right now to the status quo.
I have not touched the stuff, mainly because of Cory Doctorow's take on it. I forget the exact wording, but the gist of what he said is that AI is not good enough to do your job but it is good enough to be used to scam your boss into firing you and hiring it to fail to do your job. I was a software professional. I know just how stupid these supposedly intelligent machines really are. The people who call AI "enhanced autocorrect" have got it right, and we've all been betrayed by autocorrect.
Has autocorrect improved the world one iota? Sometimes it's nice to have my errors magically fixed, except most of the time they aren't really fixed or the wrong "errors" get changed — and then I'm irritated. If I counted up all those momentary flashes of irritation, it might well look like a waste of time as well emotion. I like Cory Doctorow's take, too, and I think he's especially on point in talking about the inevitable "enshitification" of digital platforms (a decline that will come to Substack, too).
I'm going to agree with you except for that word "inevitable". When we were students, the process (basically, buy a business with a reputation for quality, drop quality while maintaining prices, rake in the bucks for a little while, and then shut it down or sell out to an ignoramus just before the bottom drops out) was not unheard of but it was a lot less common. That's because there were laws in place to minimize that sort of thing and those laws were enforced. Let's not fall into the trap of mistaking current conditions for the laws of nature.
I don’t think if it as a law of nature - I think it’s a law of capitalism 😉
I have had 2 experiences with using AI deliberately, and while they were successful, I still say hang the rest.
1) I used ChatGPT with my college freshmen. It is much easier to teach a group of college kids how to workshop a piece of writing when the example piece you're using was spit out by a computer. There's no emotional connection there, nothing to worry about in terms of emotional investment. And, because it's written by AI, there were obviously things wrong with it, so they got to practice giving critiques and feedback.
2) My current day job uses AI to diagnose what's wrong with computers, and to diagnose what was going on with mine last week. As far as I'm concerned, this is what AI should be used for and nothing else.
Thanks, Marissa. I've considered doing what you say with students, workshopping AI pieces to surface obvious problems. Interesting that you say this helps undercut some of the emotional baggage of workshopping personal pieces – that could be a benefit. But then it seems you come down on only using AI to debug computers. I'd be happy to hear more about your experience of those student workshops with AI pieces.
My gut tells me part of the emotional-investment problem has to do with outmoded models for writing workshops, in which students end up judged on a hot seat. The process is very demotivating. In my journalism courses, I generally opt to have students spend much more class time "workshopping" published pieces by well known writers. I find it gets them used to focusing on specific criticisms and underscores that even the professionals make decisions (or mistakes) that not all readers like.
I finally have time to respond to this! I have tried a couple of the bots, including Claude, ChatGPT, and Sudowrite. Ethically, I don't like ChatGPT or Sudowrite, so I don't use those. I only use Claude, but during the editing stages. Since I'm a discovery writer and writing is how I process, it makes no sense to use AI to "help" me write.
One of these days, I want to write another article on generative AI, but my personal thought is that while it may raise the baseline (as in, help those who have little/no talent in writing "write"), it won't do much more for seasoned or more gifted writers who have truly unique voices than, Grammerly, for example.
Interesting observations, Tiffany - I’m not convinced these tools can raise the baseline level for anyone unless they already know how to write. Chatbots can generate polished-looking prose, and in some circumstances that may be enough, but for any kind of writing that’s meant to engage readers or dig into a topic, I don’t see it yet.
But that’s me. I’d love to hear more about how you use Claude - I agree about not using this stuff for first drafts, because the writing process is about figuring out what you want to say. So, at what point does Claude become useful to you? And ethical issues, yes, I’ve been thinking hard about those, and I wonder why ChatGPT and Sudowrite don’t meet your standards there. I think I know, but it would be good to hear more about that.
What I mean by "raising the baseline" isn't that their skill will get better, but that it'll allow people who had no skill in writing to be able to tell stories that they otherwise wouldn't have been able to tell. I don't know if that is a good or bad thing. In some ways, I struggle to see how using generative AI to write stories differs from using a ghostwriter.
I share a bit about how I use Claude here: https://tiffanychu.substack.com/p/the-case-for-ai-in-the-creative-life, but I don't use it for plot direction anymore like I mention in the article. I've found that my own ideas are still better, and not having the freedom to come up with them myself is stifling. However, I do use it to comment on things like how an article might land with readers, pacing issues, incomplete threads, etc. I definitely use it for SEO-related things haha. As for fiction, I ask it to point out pacing issues, unnatural dialogue, potentially missing scenes to complete story/character arcs, and inconsistencies/plot holes. I always specify that I don't want it to fix anything for me, only point things out so I can do it myself. I'm often an underwriter, so it's helpful for me to see where something is underdeveloped.
I comment on the ChatGPT issue in the post I linked, but I admit I haven't done much follow-up research since then. I've heard tell that OpenAI is working on licensing deals with authors and publishers, so if things start looking more above board with them, my stance might change? We'll see. I'm still iffy about their privacy policy and how they might use my input/prompting if I use ChatGPT. With Sudowrite, I think just the principle behind it grates on me, since it's solely an AI story generator and not much else..
Hey Martha,
So I'm a copywriter. I'm currently in the process of applying and interviewing for a new full time job. Understanding how to use AI and being able to explain how I use it in my writing process is now essential. That happened so fast. It wasn't like that a year ago.
For one company I just applied to they asked on the application - "Do you know how to use Chat GPT and please explain how you use it in your copy process."
I know a copywriter at that company so we had chatted before I filled out the application and I had asked her about what they use, do they have guidelines or does her team openly talk about it. So I knew I had to show I had that skill.
With another company I had to do a standard writing assignment, and then they asked me to make a video talking about my copy process (videos pre-interview stage are pretty standard now too.) And again, I have connections to other copywriters at that company and we had talked about what they use, how they talk about it. So I definitely used an AI app called Claude and then talked about how I used Claude for the assignment in my video. Claude is very nuanced with tone and voice so I used it almost like a copy editor at the end of my writing process to tighten things up.
So there you go. When I write for my newsletter or I write a poem or I'm writing something personal. No, not usually. But I will run some of those pieces through Claude (never the poems - those are sacred) and lift a few sentences that make things clearer and flow better. It depends on my mood, the time I have, and what it's for.
This is a very loooong answer -- so the short answer is for work - yes. That skill is now essential. For my own creative writing - kinda sorta maybe sometimes.
Jen, this on-the-ground feedback is so excellent! I'd love to talk with you more about how you see your job of copywriter evolving. I find some hope in the process you're describing, and I've been thinking a lot about the most effective way to use AI as a tool for journalists (and how I should be preparing students). I also agree that Claude is currently the bot of choice, partly because of the ethical guardrails Anthropic has put into its language model but mostly because it doesn't serve Sam Altman of Open AI 's megalomaniac goals :-)
Glad it was helpful. This round of applications has been eye opening around AI. The landscape is changing. So Claude is my 1st choice and something I'm learning how to leverage to position as a new skill set.