It would be interesting to see the result of another try at "Ghosts" with the newer models.... GPT-3 is an abacus compared to a calculator.
I think true artistic writing will continue to shine through any LLM-slop. LLMs may on average raise the quality of contemporary fiction by making technical perfection easier to achieve.... But there is much more to art than mere technical perfection. This is not the case in science (my profession), where methods and technique are paramount.
Fascinating interview, Lincoln. I especially like the last question about human agency: even if a new technology like AI becomes ubiquitous ( and many might argue it already is), we must be discerning. The idea of feeding a prompt or outline into a generative AI model and seeing what the machine spits out is one thing. The idea of taking that as a seed and making it into your own work is quite another.
I feel like we creatives, in particular, must be guardians of human originality and the art of expression, and be on alert against that "flattening" you mention--the Walter Cronkitization of language, art, music, and storytelling, to use the TV analogy.
Thanks for going deep!*
--
*Note: for fun, I fed this comment into ChatGPT and asked it to create this as a response to a post about creativity, writing and AI's tendency towards homogenizing and flattening language. Here was its response:
Fascinating interview, Lincoln. I especially appreciated the final question about human agency. Even if a technology like AI becomes ubiquitous—and many would argue it already has—we still have to be discerning in how we use it. Feeding a prompt or outline into a generative model and seeing what the machine spits out is one thing. Taking that output as a seed, then shaping and transforming it into something truly your own, is something else entirely.
As creatives, I think we have a particular responsibility to be guardians of human originality and expression. We need to stay alert to the very real risk of “flattening” you mention—the Walter Cronkitization of language, art, music, and storytelling, to borrow the TV analogy.
Thanks for going deep on this—such a valuable conversation.
I found this whole conversation fascinating too. I know nothing about ChatGPT and LLM, have never used either, with one exception, asking it to summarize my novel, and was deeply impressed with how it understood the themes of the book and motivations of the character. I can see how useful it would be to do do the trudge work summarizing chapters that agents sometimes ask for or even just to clarify our own though, pushing back on its summary, or expanding on something you weren't clear about.
In GPs rewriting your comment above, it seemed to be doing what any good editor might have done or even yourself if you'd taken the time to do so, thus saving you time.
On another note, the stories taxi drivers and others tell us is mostly fiction, not reality, what really happened, only what we believe or want others to believe. This raw universe we live in is reinterpreted through the stories we tell about it, not the thing in itself, which can't be communicated through language, in my experience, without distorting, changing, limiting it. What has this to do with LLMs? I'm not sure at this point. But I sense a fertile ground for further exploration.
I can't believe this essay was not only published but made Best American Short Stories. It's appalling. The essay had a very moving premise but it was complete Lely undercut by the falsified ChatGPT responses. Ghosts was slop, plain and simple.
It would be interesting to see the result of another try at "Ghosts" with the newer models.... GPT-3 is an abacus compared to a calculator.
I think true artistic writing will continue to shine through any LLM-slop. LLMs may on average raise the quality of contemporary fiction by making technical perfection easier to achieve.... But there is much more to art than mere technical perfection. This is not the case in science (my profession), where methods and technique are paramount.
This type of writing needs to be banned by the literary establishment.
Fascinating interview, Lincoln. I especially like the last question about human agency: even if a new technology like AI becomes ubiquitous ( and many might argue it already is), we must be discerning. The idea of feeding a prompt or outline into a generative AI model and seeing what the machine spits out is one thing. The idea of taking that as a seed and making it into your own work is quite another.
I feel like we creatives, in particular, must be guardians of human originality and the art of expression, and be on alert against that "flattening" you mention--the Walter Cronkitization of language, art, music, and storytelling, to use the TV analogy.
Thanks for going deep!*
--
*Note: for fun, I fed this comment into ChatGPT and asked it to create this as a response to a post about creativity, writing and AI's tendency towards homogenizing and flattening language. Here was its response:
Fascinating interview, Lincoln. I especially appreciated the final question about human agency. Even if a technology like AI becomes ubiquitous—and many would argue it already has—we still have to be discerning in how we use it. Feeding a prompt or outline into a generative model and seeing what the machine spits out is one thing. Taking that output as a seed, then shaping and transforming it into something truly your own, is something else entirely.
As creatives, I think we have a particular responsibility to be guardians of human originality and expression. We need to stay alert to the very real risk of “flattening” you mention—the Walter Cronkitization of language, art, music, and storytelling, to borrow the TV analogy.
Thanks for going deep on this—such a valuable conversation.
Thoughts?
I found this whole conversation fascinating too. I know nothing about ChatGPT and LLM, have never used either, with one exception, asking it to summarize my novel, and was deeply impressed with how it understood the themes of the book and motivations of the character. I can see how useful it would be to do do the trudge work summarizing chapters that agents sometimes ask for or even just to clarify our own though, pushing back on its summary, or expanding on something you weren't clear about.
In GPs rewriting your comment above, it seemed to be doing what any good editor might have done or even yourself if you'd taken the time to do so, thus saving you time.
On another note, the stories taxi drivers and others tell us is mostly fiction, not reality, what really happened, only what we believe or want others to believe. This raw universe we live in is reinterpreted through the stories we tell about it, not the thing in itself, which can't be communicated through language, in my experience, without distorting, changing, limiting it. What has this to do with LLMs? I'm not sure at this point. But I sense a fertile ground for further exploration.
I can't believe this essay was not only published but made Best American Short Stories. It's appalling. The essay had a very moving premise but it was complete Lely undercut by the falsified ChatGPT responses. Ghosts was slop, plain and simple.