The issue, as I see it, is that the inclusion of AI material violates Harlan Ellison’s guiding ethos of “Pay the writer!”. As you state, AI can’t generate its own opinions. It recognizes patterns from large datasets. To create the response OpenAI had to steal the work of human beings and ChatGPT had to synthesize their work into an answer to your query. Maybe there is some merit to the question, but is how you achieved the result something Ellison would have endorsed? Profiting a company that devalues every person it can extract wealth from? I personally don’t believe so, but, my opinion is simply that. An opinion.
“So with that in mind, and recalling Harlan’s tendency to include commentary from the other side…”
Herein lies the issue, though: the output from ChatGPT is not another “side”, it is an uncredited amalgam of analysis others have already provided. How did you verify that the output isn’t simply plagiarized?
I concur that it's a fine line, to be sure...when we say, in a history, "country X came in on the side of country Y", is the country actually a thing, or an amalgam? Does the "country" actually want? Or is "side" not tied to human specificity? As to the latter point...as noted in the essay, the concerns about AI were almost certainly derived from other sources. The important part, at least to me, is that those arguments have achieved enough visibility that they can now turn up in this format, and popped up here.
Joe, I wonder if you've read any of David Gerrold's posts on AI. He had a series of them on his Facebook and possibly elsewhere, delving into various aspects of how it worked and how he could use it as a tool. It was quite interesting, and his approach is very similar to yours.
On another note, I've noticed sometimes that two people making exactly the same query of the same AI engine end up with different results and I wonder if that particular model may be taking into account their own past queries or other content it scrapes from their devices? It would be an interesting experiment to pose to your readers here if you shared exactly the prompt that you gave about Harlan's work and had others submit it both to ChatGPT and other LLM engines, and see how the results vary.
Of course an AI didn't write this of course an AI didn't write this of course an AI didn't write this of course an AI didn't write this of course an AI didn't write this of course an AI didn't write this of course an AI didn't write this
I hope this finds you well. My name is Roger Terry. I’m a big fan of yours, and I’m big fan of Harlan Ellison and his work. I’m a writer, and he is one of my biggest influences. I fully understand where you are coming from in adding the essay to the book.
But, for the sake of sales, I think it would be a positive to also publish a separate version of the book that does not include the essay. With how heated everything is when it comes to AI, especially within literary communities, I have a fear that many people will be turned away from picking up a copy because of the AI essay alone. I want to see the book succeed; I want to see Harlan’s work reach as many people as possible, but the majority of people who read do not want to see AI in any way, no matter the intention. Thank you for all the work you have put in to help preserve Harlan’s legacy, but a separate AI-free edition I feel needs to come out to give the book its best chance to build his readership. Either, way, I hope Harlan’s legacy continues to grow, and I wish you well.
I think this is actually quite clever. It invites us to consider the products of current-generation AIs as a form of historical artifact, in the context of a much longer conversation about AI. Obviously what's happening now is very important to us, because we live in the present, but it's also important to zoom out and look at how our perspectives change as the technologies we imagined become the technologies we live with.
As a digital content creator and a union actor, I have a lot of strong opinions about AI and its use for creative purposes. Generally pretty gosh-darned against it, I have to say I'm no longer among the "This will be the death of creativity" mindset. This revelation came not from a brilliant essay or deep, philosophical discussion, but of all places, a mildly vulgar but funny meme that said "Your AI generated flyer looks like poop from a butt."
It got me to thinking that a general statement like that is valid in this case, because all evidence shows that AI generated images, graphic design, videos, etc. do all look like they came from the same place. And that's because, as you point out, an AI generated product is only the sum of the human who guided it. The more AI has to fill in the gaps itself, the more generic it looks.
AI is capable of good graphic design, but only when prompted by someone versed in the techniques, artistry, and best practices of graphic design and with an artistic eye developed over years of experience and training. If someone just sits down, doesn't know what they want to see, and just types in "Create a magazine cover with a kitten on it and the following information:" they're going to get a result that looks like every other layperson that tries to skip the expertise part and do it themselves.
Same with video. A prompt of "A man in cool tech armor walks down a futuristic street and something blows up behind him" is going to get a generic video that is unremarkable. But an experienced filmmaker will be able to give it exact shots, focal lengths, camera and lens looks, blocking, etc. and make it something more distinctive.
In short, while I still oppose the use of AI in these ways (actors and cameramen gotta eat and will always be better than even the most expertly guided AI), it's clear that expertise is still needed even when using AI.
It also occurs to me that, the more AI generated content is produced and published to the internet, the more AI will be scraping and copying other AI generated content. One can easily see a time when the xerox-of-a-xerox problem starts to play a role in what the average layperson is able to produce with AI. I'm sure there are ways to mitigate this, but it's an interesting thought.
Perhaps we have the wrong metric. I am interested in the kind of questions an AI might ask. Not the answers. The answers may be interesting or helpful, but they come from a human somewhere. Maybe a lonely soul who wrote something on an obscure bulletin board, and has now, thanks to AI, found an audience. But it is the questions that are at the core of creativity. One good question, like, "how can we do better at ...?"
The missing component in many arguments from artists about A.I. is a fundamental misreading of the technology and its use in creative spheres. Hi, I’ve sold millions of dollars in software to the biggest companies in the world. I work with Generative Artificial Intelligence, Machine Learning, and Deterministic Automation every day.
And I’m an artist, a writer, a designer of my own books, and an incorporated publisher.
The world of personal use AI is rapidly evolving beyond the simplistic frontier models (OpenAI, Anthropic, etc.). Now, it is possible for an author to host private LLMs in their homes, on their networks, with open-source models, grounded on every sentence they’ve ever written, published, or taken as notes.
The question becomes a local one. What can an artist accomplish with full, complete, and ethical use of their entire life’s work stream? What will an artist’s next work be when then they build it from concept based on everything they’ve done before but haven’t yet accomplished?
The game incidentally is on steam and gog for 9.99./ Gog has offline installers that can download. Since I never had the chance to play I have No Mouth I must Scream. partly because I never thought to look in 1996
I think that this is fascinating, and I can't help but have a feeling of "welp, we're doomed" when something as thoughtful as this gets such a knee-jerk response that spreads like wildfire. People are scared, and they have good reasons to be, even if -- as is usually the case in the human condition -- the thing they think they're scared of is not the thing that is actually scary.
I can't help but think that Ellison would have been amused by the conflagration, whatever his signature hyperbolic public response would have been.
The issue, as I see it, is that the inclusion of AI material violates Harlan Ellison’s guiding ethos of “Pay the writer!”. As you state, AI can’t generate its own opinions. It recognizes patterns from large datasets. To create the response OpenAI had to steal the work of human beings and ChatGPT had to synthesize their work into an answer to your query. Maybe there is some merit to the question, but is how you achieved the result something Ellison would have endorsed? Profiting a company that devalues every person it can extract wealth from? I personally don’t believe so, but, my opinion is simply that. An opinion.
I can’t claim to have known HE as JMS did (or at all), but from what I know OF him, I think he would have appreciated the meta-ness of the exercise.
“So with that in mind, and recalling Harlan’s tendency to include commentary from the other side…”
Herein lies the issue, though: the output from ChatGPT is not another “side”, it is an uncredited amalgam of analysis others have already provided. How did you verify that the output isn’t simply plagiarized?
I concur that it's a fine line, to be sure...when we say, in a history, "country X came in on the side of country Y", is the country actually a thing, or an amalgam? Does the "country" actually want? Or is "side" not tied to human specificity? As to the latter point...as noted in the essay, the concerns about AI were almost certainly derived from other sources. The important part, at least to me, is that those arguments have achieved enough visibility that they can now turn up in this format, and popped up here.
This comment isn’t your original thought. It’s something that everybody is saying these days.
Joe, I wonder if you've read any of David Gerrold's posts on AI. He had a series of them on his Facebook and possibly elsewhere, delving into various aspects of how it worked and how he could use it as a tool. It was quite interesting, and his approach is very similar to yours.
On another note, I've noticed sometimes that two people making exactly the same query of the same AI engine end up with different results and I wonder if that particular model may be taking into account their own past queries or other content it scrapes from their devices? It would be an interesting experiment to pose to your readers here if you shared exactly the prompt that you gave about Harlan's work and had others submit it both to ChatGPT and other LLM engines, and see how the results vary.
don't tell me an A.I. wrote this.
Of course an AI didn't write this of course an AI didn't write this of course an AI didn't write this of course an AI didn't write this of course an AI didn't write this of course an AI didn't write this of course an AI didn't write this
Dear Mr. Straczynski,
I hope this finds you well. My name is Roger Terry. I’m a big fan of yours, and I’m big fan of Harlan Ellison and his work. I’m a writer, and he is one of my biggest influences. I fully understand where you are coming from in adding the essay to the book.
But, for the sake of sales, I think it would be a positive to also publish a separate version of the book that does not include the essay. With how heated everything is when it comes to AI, especially within literary communities, I have a fear that many people will be turned away from picking up a copy because of the AI essay alone. I want to see the book succeed; I want to see Harlan’s work reach as many people as possible, but the majority of people who read do not want to see AI in any way, no matter the intention. Thank you for all the work you have put in to help preserve Harlan’s legacy, but a separate AI-free edition I feel needs to come out to give the book its best chance to build his readership. Either, way, I hope Harlan’s legacy continues to grow, and I wish you well.
Sincerely, your friend,
Roger.
I think this is actually quite clever. It invites us to consider the products of current-generation AIs as a form of historical artifact, in the context of a much longer conversation about AI. Obviously what's happening now is very important to us, because we live in the present, but it's also important to zoom out and look at how our perspectives change as the technologies we imagined become the technologies we live with.
As a digital content creator and a union actor, I have a lot of strong opinions about AI and its use for creative purposes. Generally pretty gosh-darned against it, I have to say I'm no longer among the "This will be the death of creativity" mindset. This revelation came not from a brilliant essay or deep, philosophical discussion, but of all places, a mildly vulgar but funny meme that said "Your AI generated flyer looks like poop from a butt."
It got me to thinking that a general statement like that is valid in this case, because all evidence shows that AI generated images, graphic design, videos, etc. do all look like they came from the same place. And that's because, as you point out, an AI generated product is only the sum of the human who guided it. The more AI has to fill in the gaps itself, the more generic it looks.
AI is capable of good graphic design, but only when prompted by someone versed in the techniques, artistry, and best practices of graphic design and with an artistic eye developed over years of experience and training. If someone just sits down, doesn't know what they want to see, and just types in "Create a magazine cover with a kitten on it and the following information:" they're going to get a result that looks like every other layperson that tries to skip the expertise part and do it themselves.
Same with video. A prompt of "A man in cool tech armor walks down a futuristic street and something blows up behind him" is going to get a generic video that is unremarkable. But an experienced filmmaker will be able to give it exact shots, focal lengths, camera and lens looks, blocking, etc. and make it something more distinctive.
In short, while I still oppose the use of AI in these ways (actors and cameramen gotta eat and will always be better than even the most expertly guided AI), it's clear that expertise is still needed even when using AI.
It also occurs to me that, the more AI generated content is produced and published to the internet, the more AI will be scraping and copying other AI generated content. One can easily see a time when the xerox-of-a-xerox problem starts to play a role in what the average layperson is able to produce with AI. I'm sure there are ways to mitigate this, but it's an interesting thought.
That's the point where AI begins fouling its own nest....
Precisely. It will be interesting to see how that manifests and what these companies do to try to adjust for that.
I appreciate this meditation.
Where Chat GPT ended up reminds me of one way of understanding LLMs: as mirrors held up to our society, at least the digitized parts.
"art comes through passion, through feeling and life-experience of which AI has none"
A skeptic might point out that a well-trained AI has internalized billions of people's passions, feelings, and life experiences.
Perhaps we have the wrong metric. I am interested in the kind of questions an AI might ask. Not the answers. The answers may be interesting or helpful, but they come from a human somewhere. Maybe a lonely soul who wrote something on an obscure bulletin board, and has now, thanks to AI, found an audience. But it is the questions that are at the core of creativity. One good question, like, "how can we do better at ...?"
The missing component in many arguments from artists about A.I. is a fundamental misreading of the technology and its use in creative spheres. Hi, I’ve sold millions of dollars in software to the biggest companies in the world. I work with Generative Artificial Intelligence, Machine Learning, and Deterministic Automation every day.
And I’m an artist, a writer, a designer of my own books, and an incorporated publisher.
The world of personal use AI is rapidly evolving beyond the simplistic frontier models (OpenAI, Anthropic, etc.). Now, it is possible for an author to host private LLMs in their homes, on their networks, with open-source models, grounded on every sentence they’ve ever written, published, or taken as notes.
The question becomes a local one. What can an artist accomplish with full, complete, and ethical use of their entire life’s work stream? What will an artist’s next work be when then they build it from concept based on everything they’ve done before but haven’t yet accomplished?
The game incidentally is on steam and gog for 9.99./ Gog has offline installers that can download. Since I never had the chance to play I have No Mouth I must Scream. partly because I never thought to look in 1996
AI is like what communism intends, only with communicating ideas rather than money?
I think that this is fascinating, and I can't help but have a feeling of "welp, we're doomed" when something as thoughtful as this gets such a knee-jerk response that spreads like wildfire. People are scared, and they have good reasons to be, even if -- as is usually the case in the human condition -- the thing they think they're scared of is not the thing that is actually scary.
I can't help but think that Ellison would have been amused by the conflagration, whatever his signature hyperbolic public response would have been.