The brain needs to work through problems on it's own to make neural connections. Yes it is painful when you cannot think of something right away, but eventually it will click. You are much better off talking to another person than a computer, we do not want to lose our human to human connection.
You know when you are thinking of a movie and you are trying to remember who that actor was......most people grab for their phones and google it. I refrain. I think and think......and I will eventually get it. A.I. will make our brains lazy. Don't let it. Use your noggin ❤️
Great article. I'm curious what your thoughts about using AI as a sounding board are? I've sometimes used AI as a sort of "rubber duck" to kind of suss out the ideas I have bouncing around inside my head. Sometimes I'll have to chastise it if it starts trying to write for me or starts going off the rails, but usually with the correct prompts I've found it helps me flesh out my ideas a bit.
Generally I'm of the 'if it helps, then it helps' mentality. The specific use case suggested by Newton that I wrote about in the article I don't think does help.
Using AI as a sounding board/rubber duck is quite different, and far more interesting to me.
There's a key difference between it and usual rubber duck debugging, though: the rubber duck (or an unfortunate colleague!) never talks back. All the thought processes remain your own.
A conversation with an AI is quite different, in that it actively responds. But unlike talking to, say, an unwitting colleague or another writer, you've got no real way of knowing where those responses are coming from. e.g., I might seek out particular writers or experts in specific fields to get precise feedback, and the value there is in having a basic idea of their viewpoint (though that is still highly unpredictable, of course).
As I said to Piotr elsewhere in the comments, my question when it comes to conversations with an AI is where the viewpoint is coming from. If it's a homogenised, generic AI response, sourced and evened out across its entire dataset, then I'm unsure whether I'd find that interesting or useful.
I think I'd prefer talking to an actual rubber duck, or an actual writer. But if it helps you write more, then that's all good. My suspicion around Newton's suggestion in the article is quite specific to his use case ("use AI to write the first chapter"), which is a very different thing to what you're describing.
I think? Give it another 12 months and we'll all have to re-evaluate all of this stuff again. :)
That's a good point. I've had a few instances where the responses from the AI have been pretty terrible and generic. But it does get the creative juices flowing, which to me is the most important part. If I ask a writer friend (presumably one who at least is knowledgeable or skilled enough to provide reasonable feedback) I can expect that they have at least a tentative understanding of what I'm going for and infer a lot. With the AI, I have to spell it out more, which can be good or bad.
"In fact, the situation there is closer to a celebrity hiring a ghost writer to write their ‘biography’ or airport thriller. It’s cosplaying as an author."
The American author David Ritz has been in this situation for the autobiographies some of my favorite musicians have written (Smokey Robinson, Ray Charles, Buddy Guy) or biographies based on his relations with said musicians (Marvin Gaye, Aretha Franklin). But he has also written and published novels on his own, so I can't really call him a "cosplayer".
I think you've hit this spot on. I read a story to connect with the author, and all the wacky things that happen to us as part of life, the universe, and everything -- not just to read the words, even if well written. If there was no creative spark, just a machine-generated one, then not only is the author getting a bit of a disservice, but so are we. It would be like going to a concert and there are four musicians on stage and a machine playing lead guitar. Gimmick works for a bit, but would not be filling stadiums in ten years. Now a dinosaur with a jetpack? That I'd follow (or at least buy the t-shirt). I'll also add that I used to be a media officer for AI at a federal agency in the U.S., and the way AI is being pitched as a replacement for creativity is never going to match human creativity -- that's just not how LLMs and transformers work. Will some publishers stop paying for fiction and get a machine (or a guy with a machine) to write instead? Absolutely. But we will not be getting Tolkein out of it no matter how advanced the models progress
If you are going to have AI write a paragraph, consider a free upgrade: read the book(s) from which that paragraph was sampled. Shocking though it seems, existing books can yield inspiration.
Most risable is the implication that any publisher is keenly interested in the possibly thwarted creative urges of 8 billion people, or even eight people.
Here are two planks of wood. I've hammered them together with six nails. Now you can move ahead with becoming a carpenter!
I spent all year experimenting with LLMs to see if they could help somehow as an aid in both my creative writing and investigative journalism. My tentative conclusion now is that they just are not reliable enough to be as helpful as we hope. I’ve now gone more the opposite direction writing-wise and creatively. I am writing more by hand and filming myself doing it.
As always, thank you! I just love the heartfelt common sense of this article. Well done. Helps me wrestle some things down and ultimately write more, which is the point!
Why be so black-and-white about using AI? While I would NEVER ask it to write a single sentence for me, I use it often to discuss my ideas or when I’m stuck. The thing is, I’m a conversational writer - I work best when I see my ideas bouncing off from somebody else. I must see them outside of me. And sometimes, my mind is blank and I need anything to get me going. And then I ask AI for a comment on my problem. And this comment gets me going. It’s no different from reading a random newspaper clip to get inspired. But a newspaper clip cannot be a sparring partner. Sometimes I ask AI to criticize my idea. To tell me why it’s stupid. This sparks a process too.
I like the sparring, the bouncing of ideas, because I see them like a ball that gets tossed around outside of my head.
And then I sometimes take something and use it. The selection process is mine (the same one if I chose a bit from a book or an article) and also the very act of writing and how I write is also uniquely mine.
AI has never proposed even one sentence that would sound the way I would write.
Good questions! You're describing quite a different use case to the one proposed by Newton, though. You have quite a specific use here, with clear limits and boundaries. A bit like some of my other suggestions for overcoming writer's block, what you're describing here also does not actually do the writing. Ultimately, that's down to you.
A few questions back at you, as I find the 'rubber duck' use of AI fascinating:
1. Why not just talk through your thoughts out loud (or to a rubber duck!), and interrogate them that way?
2. Why not talk to another writer? I've been mentoring a writer this last month and those conversations have been really exciting and provocative for both of us.
3. In terms of sparring, and getting feedback from the AI, I'm curious about where you imagine that 'opinion' to be coming from. Have you deliberately built a persona there, somehow, or is this off-the-shelf ChatGPT? If the latter, is there a risk that the feedback and challenges you're getting are going to be a sort of homogenised, generic 'hive mind' take? ie, not an actual viewpoint, but a sort of generalised response based on all literature. I don't know whether that would be a helpful or unhelpful situation.
I suppose for me, I struggle to see an interaction with generative AI as 'a conversation'. So that's probably a sticking point for me personally.
I must admit I was quite astonished at Newton's approach, as it would never occur to me to discuss overcoming writer's block with AI by way of actually letting AI write.
But anyway, your questions:
1. That wouldn't work. I need my thoughts bounced off a third party. I need someone (even if it's a chatbot) reflect them back to me. Usually, the go back rephrased, or with a different angle, a new point of view, something added. AI sometimes systematizes my doubts in a way that I look at them with a fresh eye. If I talk through my thoughts out loud, they stay the same, it's like a hamster wheel. I need fresh input.
2. That's a great option, sure, but the problem is, another writer is not as available as AI is ;) Besides, AI has a very deep and broad knowledge base (you just need to dig deep enough and know how to extract it by way of a right prompt). Talking with living people is great for decisions that relate to taste and aesthetics, while talking is AI is great when you need a research companion and a knowledgeable scholar.
3. I talk mostly with Claude, and, depending on the problem I have, I ask him to take on a specific persona. For example, I had a problem with a showdown in an abandoned harbor. Something with my setup didn't feel right. So I asked Claude to take on a persona of an expert in espionage, special forces operations etc., described the situation and told him why am I feeling this is stupid and how would real agents behave. His responses were logical, highly probable and he was able to explain why something is so and so. The best thing about AI is that you can design the most specific persona you want and it's not that difficult to avoid homogenized answers. Also, with AI you receive bland, homogenized answers when you ask bland, homogenized questions (aka "garbage in, garbage out"). I ask him about very specific problems, or start with highly specific idea (it's not "what should I write my next short story about?" but rather "why would anybody use demonic possession instead of just wiretapping their phone to watch their movement?").
A few rule of thumbs of using AI for brainstorming/research/editing:
When sceptical about an idea, ask for chain of reasoning, ask to explain why. Or express your doubts.
Be ready to reach a satisfying response only after a few iterations.
Ultimately, use your taste. 90% of ideas or suggestions are useless. I just use them as a starting point for something else. Claude presents me with an idea I don't like (or is downright silly), but it is enough to spark a creative process in me. I take this idea, wrangle with it, come up with something else (and much better), give it to him, he builds on that idea, which sparks another idea in me, and so on.
I think, for me, the general concept is that I am an extroverted thinker - I need to get my thoughts outside of my brain, have it verbalized by someone else, that work with this outsider's input. When I stay in my head, I block.
I hope my answer was comprehensive enough. Feel free to ask if you have more questions! And I encourage you to give it a try. It's a process, and it requires learning how to use it well and intelligently, but I believe it's totally worth it.
The brain needs to work through problems on it's own to make neural connections. Yes it is painful when you cannot think of something right away, but eventually it will click. You are much better off talking to another person than a computer, we do not want to lose our human to human connection.
You know when you are thinking of a movie and you are trying to remember who that actor was......most people grab for their phones and google it. I refrain. I think and think......and I will eventually get it. A.I. will make our brains lazy. Don't let it. Use your noggin ❤️
Great article. I'm curious what your thoughts about using AI as a sounding board are? I've sometimes used AI as a sort of "rubber duck" to kind of suss out the ideas I have bouncing around inside my head. Sometimes I'll have to chastise it if it starts trying to write for me or starts going off the rails, but usually with the correct prompts I've found it helps me flesh out my ideas a bit.
Generally I'm of the 'if it helps, then it helps' mentality. The specific use case suggested by Newton that I wrote about in the article I don't think does help.
Using AI as a sounding board/rubber duck is quite different, and far more interesting to me.
There's a key difference between it and usual rubber duck debugging, though: the rubber duck (or an unfortunate colleague!) never talks back. All the thought processes remain your own.
A conversation with an AI is quite different, in that it actively responds. But unlike talking to, say, an unwitting colleague or another writer, you've got no real way of knowing where those responses are coming from. e.g., I might seek out particular writers or experts in specific fields to get precise feedback, and the value there is in having a basic idea of their viewpoint (though that is still highly unpredictable, of course).
As I said to Piotr elsewhere in the comments, my question when it comes to conversations with an AI is where the viewpoint is coming from. If it's a homogenised, generic AI response, sourced and evened out across its entire dataset, then I'm unsure whether I'd find that interesting or useful.
I think I'd prefer talking to an actual rubber duck, or an actual writer. But if it helps you write more, then that's all good. My suspicion around Newton's suggestion in the article is quite specific to his use case ("use AI to write the first chapter"), which is a very different thing to what you're describing.
I think? Give it another 12 months and we'll all have to re-evaluate all of this stuff again. :)
That's a good point. I've had a few instances where the responses from the AI have been pretty terrible and generic. But it does get the creative juices flowing, which to me is the most important part. If I ask a writer friend (presumably one who at least is knowledgeable or skilled enough to provide reasonable feedback) I can expect that they have at least a tentative understanding of what I'm going for and infer a lot. With the AI, I have to spell it out more, which can be good or bad.
Same here, as I said in my comment below.
"In fact, the situation there is closer to a celebrity hiring a ghost writer to write their ‘biography’ or airport thriller. It’s cosplaying as an author."
The American author David Ritz has been in this situation for the autobiographies some of my favorite musicians have written (Smokey Robinson, Ray Charles, Buddy Guy) or biographies based on his relations with said musicians (Marvin Gaye, Aretha Franklin). But he has also written and published novels on his own, so I can't really call him a "cosplayer".
In that comment I was thinking of the celebrity being the one cosplaying, not the writer.
Oh, I see. But some of them take the position of being an "author" more seriously than others.
I think you've hit this spot on. I read a story to connect with the author, and all the wacky things that happen to us as part of life, the universe, and everything -- not just to read the words, even if well written. If there was no creative spark, just a machine-generated one, then not only is the author getting a bit of a disservice, but so are we. It would be like going to a concert and there are four musicians on stage and a machine playing lead guitar. Gimmick works for a bit, but would not be filling stadiums in ten years. Now a dinosaur with a jetpack? That I'd follow (or at least buy the t-shirt). I'll also add that I used to be a media officer for AI at a federal agency in the U.S., and the way AI is being pitched as a replacement for creativity is never going to match human creativity -- that's just not how LLMs and transformers work. Will some publishers stop paying for fiction and get a machine (or a guy with a machine) to write instead? Absolutely. But we will not be getting Tolkein out of it no matter how advanced the models progress
(As a newbie, realizing I cannot edit Substack posts, so apologies for the grammar fails I mixed in to the comment.)
I think you should be able to edit - check the 3 dot menu at the top-right of your comment!
Weirdly, the only options are to share or delete... I'll try on my computer, maybe it's a phone app issue... thank you!!!!
If you are going to have AI write a paragraph, consider a free upgrade: read the book(s) from which that paragraph was sampled. Shocking though it seems, existing books can yield inspiration.
Most risable is the implication that any publisher is keenly interested in the possibly thwarted creative urges of 8 billion people, or even eight people.
Here are two planks of wood. I've hammered them together with six nails. Now you can move ahead with becoming a carpenter!
It's dumb.
I spent all year experimenting with LLMs to see if they could help somehow as an aid in both my creative writing and investigative journalism. My tentative conclusion now is that they just are not reliable enough to be as helpful as we hope. I’ve now gone more the opposite direction writing-wise and creatively. I am writing more by hand and filming myself doing it.
I like to write. It's fun. It's one of my favorite things to do. Having AI write for me would be like having a stand-in for sex. No thanks.
How cold is your chin? I shaved clean for Halloween for the first time in two years, and, yeah, my face is cold
Also, a long, nuanced article with many valid points when you could have typed "ni," and called it a day.
It’s mainly my top lip that is suffering. My top lip is very confused
As always, thank you! I just love the heartfelt common sense of this article. Well done. Helps me wrestle some things down and ultimately write more, which is the point!
Why be so black-and-white about using AI? While I would NEVER ask it to write a single sentence for me, I use it often to discuss my ideas or when I’m stuck. The thing is, I’m a conversational writer - I work best when I see my ideas bouncing off from somebody else. I must see them outside of me. And sometimes, my mind is blank and I need anything to get me going. And then I ask AI for a comment on my problem. And this comment gets me going. It’s no different from reading a random newspaper clip to get inspired. But a newspaper clip cannot be a sparring partner. Sometimes I ask AI to criticize my idea. To tell me why it’s stupid. This sparks a process too.
I like the sparring, the bouncing of ideas, because I see them like a ball that gets tossed around outside of my head.
And then I sometimes take something and use it. The selection process is mine (the same one if I chose a bit from a book or an article) and also the very act of writing and how I write is also uniquely mine.
AI has never proposed even one sentence that would sound the way I would write.
But it’s great at research and other stuff.
Good questions! You're describing quite a different use case to the one proposed by Newton, though. You have quite a specific use here, with clear limits and boundaries. A bit like some of my other suggestions for overcoming writer's block, what you're describing here also does not actually do the writing. Ultimately, that's down to you.
A few questions back at you, as I find the 'rubber duck' use of AI fascinating:
1. Why not just talk through your thoughts out loud (or to a rubber duck!), and interrogate them that way?
2. Why not talk to another writer? I've been mentoring a writer this last month and those conversations have been really exciting and provocative for both of us.
3. In terms of sparring, and getting feedback from the AI, I'm curious about where you imagine that 'opinion' to be coming from. Have you deliberately built a persona there, somehow, or is this off-the-shelf ChatGPT? If the latter, is there a risk that the feedback and challenges you're getting are going to be a sort of homogenised, generic 'hive mind' take? ie, not an actual viewpoint, but a sort of generalised response based on all literature. I don't know whether that would be a helpful or unhelpful situation.
I suppose for me, I struggle to see an interaction with generative AI as 'a conversation'. So that's probably a sticking point for me personally.
I must admit I was quite astonished at Newton's approach, as it would never occur to me to discuss overcoming writer's block with AI by way of actually letting AI write.
But anyway, your questions:
1. That wouldn't work. I need my thoughts bounced off a third party. I need someone (even if it's a chatbot) reflect them back to me. Usually, the go back rephrased, or with a different angle, a new point of view, something added. AI sometimes systematizes my doubts in a way that I look at them with a fresh eye. If I talk through my thoughts out loud, they stay the same, it's like a hamster wheel. I need fresh input.
2. That's a great option, sure, but the problem is, another writer is not as available as AI is ;) Besides, AI has a very deep and broad knowledge base (you just need to dig deep enough and know how to extract it by way of a right prompt). Talking with living people is great for decisions that relate to taste and aesthetics, while talking is AI is great when you need a research companion and a knowledgeable scholar.
3. I talk mostly with Claude, and, depending on the problem I have, I ask him to take on a specific persona. For example, I had a problem with a showdown in an abandoned harbor. Something with my setup didn't feel right. So I asked Claude to take on a persona of an expert in espionage, special forces operations etc., described the situation and told him why am I feeling this is stupid and how would real agents behave. His responses were logical, highly probable and he was able to explain why something is so and so. The best thing about AI is that you can design the most specific persona you want and it's not that difficult to avoid homogenized answers. Also, with AI you receive bland, homogenized answers when you ask bland, homogenized questions (aka "garbage in, garbage out"). I ask him about very specific problems, or start with highly specific idea (it's not "what should I write my next short story about?" but rather "why would anybody use demonic possession instead of just wiretapping their phone to watch their movement?").
A few rule of thumbs of using AI for brainstorming/research/editing:
When sceptical about an idea, ask for chain of reasoning, ask to explain why. Or express your doubts.
Be ready to reach a satisfying response only after a few iterations.
Ultimately, use your taste. 90% of ideas or suggestions are useless. I just use them as a starting point for something else. Claude presents me with an idea I don't like (or is downright silly), but it is enough to spark a creative process in me. I take this idea, wrangle with it, come up with something else (and much better), give it to him, he builds on that idea, which sparks another idea in me, and so on.
I think, for me, the general concept is that I am an extroverted thinker - I need to get my thoughts outside of my brain, have it verbalized by someone else, that work with this outsider's input. When I stay in my head, I block.
I hope my answer was comprehensive enough. Feel free to ask if you have more questions! And I encourage you to give it a try. It's a process, and it requires learning how to use it well and intelligently, but I believe it's totally worth it.
The local buses all have mustaches on the front of them. It’s kind of cute!