I think the authors have some good grounds for their concerns here, the large pusblishin houses and the whole field of AI-literature risks stagnation if we gatekeep young and talented authors from earning a living in their field of choice.
The reason for this is since AI models trained on AI will eventually experience a so-called model collapse, if future AIs are trained on literature (and coding) based on previous AI outputs, this could potentially cause the entire AI field to collapse. The current technology of AI, based on transformer architecture, does have a slight advantage in pattern matching compared to humans, but in their entirety, AIs are very prone to entropy. So, my fear is that future AI "co-writers" may suddenly turn out to be a lot less useful than current versions.There is a need for a large and growing corpus of human-produced knowledge and art, and possibly for AIs to summarize and expand on that knowledge.
Basically, AIs in their current form are very dependent on human talent, so if publishers reduce the number of young, aspiring authors, future AIs will stagnate. I would guess any technology that goes beyond this bottleneck would, in practice, be equivalent to AGI, and in that case, humanity would have a whole new set of problems to address.
There is no danger of that. There are far more books being written today than there were 100 years ago. Granted, most of them are terrible, but that has been the case for the last 150+ years.
I get where people are coming from on this one, but they're looking at the whole thing entirely wrong.
It's not so different from the digital camera revolution; in the early days, film enthusiasts were convinced that digital could never match the quality. Then they started to panic. Today, everyone has multiple high-quality digital cameras to use whenever they desire. Even so, what effect has the proliferation of digital cameras and simple editing tools had on the numbers of professional photographers? It still takes skill, talent, and mastery to craft a good photo, and there are still a lot of people making a decent living doing wedding shoots and publicity shots for businesses. The smart ones are learning how to incorporate AI into their businesses now, too.
AI may generate books, but it still takes someone who has mastery over the craft of storytelling to turn that basic AI material into something truly worth reading.
And if the 'no-AI' authors band together and open their own business, they will achieve their goals relatively rapidly. Otherwise: Zero chance.
At this point, I would pay Castalia for a deep revision that rationalized Gravity's Rainbow or Finnegan's Wake for me. I have a sneaking suspicion they are crap (by my own personal barometer, not really relevant to anyone else) and I refuse to give them a 3rd, and 2nd chance, respectively, because time is too valuable.
Bonus points if (business idea incoming....) Castalia takes *audience prompts* and re-writes/condenses/explain/cartoonifies/etc such books and prints them. I don't mean "a summary" like Cliff/Spark Notes. I mean "hey Castalia, re-cast Moby Dick as a VC/founder tale, add in fintech, make it no longer than 150 pages and write with a hint of Colman McCarthy and Bret Easton Ellis. That's a service I'd pay for. And, I could still kinda/sorta talk about Moby Dick at my Hampton's cocktail party with the hedge fund bros. Wins all around.
Anyway: Good luck to all.
PS Monster's story is going well. Light fun. Looking forward to more.
Has anyone seen AI provide an original plot outline?
Claude at least could not come up with a major crisis that wasn't a trope for a post collapse 2033 Idaho. If all your writing is derivative, then yes, AI is terrifying.
It'll be interesting to see if AI ever can start actually coming up or synthesizing new ideas from its trained data.
I think that will require a specific plot-generation AI engine that the more general writing engines that we presently have cannot do. It's not really within their pattern-recognition purview.
Say the subject is chess; you need one bot to play, one that only generates images of chess, one that writes about chess. There is no connection between the 3. Only a human can do that.
I think the authors have some good grounds for their concerns here, the large pusblishin houses and the whole field of AI-literature risks stagnation if we gatekeep young and talented authors from earning a living in their field of choice.
The reason for this is since AI models trained on AI will eventually experience a so-called model collapse, if future AIs are trained on literature (and coding) based on previous AI outputs, this could potentially cause the entire AI field to collapse. The current technology of AI, based on transformer architecture, does have a slight advantage in pattern matching compared to humans, but in their entirety, AIs are very prone to entropy. So, my fear is that future AI "co-writers" may suddenly turn out to be a lot less useful than current versions.There is a need for a large and growing corpus of human-produced knowledge and art, and possibly for AIs to summarize and expand on that knowledge.
Basically, AIs in their current form are very dependent on human talent, so if publishers reduce the number of young, aspiring authors, future AIs will stagnate. I would guess any technology that goes beyond this bottleneck would, in practice, be equivalent to AGI, and in that case, humanity would have a whole new set of problems to address.
There is no danger of that. There are far more books being written today than there were 100 years ago. Granted, most of them are terrible, but that has been the case for the last 150+ years.
I get where people are coming from on this one, but they're looking at the whole thing entirely wrong.
It's not so different from the digital camera revolution; in the early days, film enthusiasts were convinced that digital could never match the quality. Then they started to panic. Today, everyone has multiple high-quality digital cameras to use whenever they desire. Even so, what effect has the proliferation of digital cameras and simple editing tools had on the numbers of professional photographers? It still takes skill, talent, and mastery to craft a good photo, and there are still a lot of people making a decent living doing wedding shoots and publicity shots for businesses. The smart ones are learning how to incorporate AI into their businesses now, too.
AI may generate books, but it still takes someone who has mastery over the craft of storytelling to turn that basic AI material into something truly worth reading.
Oddly, I appreciate both stands presented here.
And if the 'no-AI' authors band together and open their own business, they will achieve their goals relatively rapidly. Otherwise: Zero chance.
At this point, I would pay Castalia for a deep revision that rationalized Gravity's Rainbow or Finnegan's Wake for me. I have a sneaking suspicion they are crap (by my own personal barometer, not really relevant to anyone else) and I refuse to give them a 3rd, and 2nd chance, respectively, because time is too valuable.
Bonus points if (business idea incoming....) Castalia takes *audience prompts* and re-writes/condenses/explain/cartoonifies/etc such books and prints them. I don't mean "a summary" like Cliff/Spark Notes. I mean "hey Castalia, re-cast Moby Dick as a VC/founder tale, add in fintech, make it no longer than 150 pages and write with a hint of Colman McCarthy and Bret Easton Ellis. That's a service I'd pay for. And, I could still kinda/sorta talk about Moby Dick at my Hampton's cocktail party with the hedge fund bros. Wins all around.
Anyway: Good luck to all.
PS Monster's story is going well. Light fun. Looking forward to more.
Even if they ban it, how many people can tell what is written by AI and not?
Most people already can't. Within 18 months, no one will be able to tell.
Has anyone seen AI provide an original plot outline?
Claude at least could not come up with a major crisis that wasn't a trope for a post collapse 2033 Idaho. If all your writing is derivative, then yes, AI is terrifying.
It'll be interesting to see if AI ever can start actually coming up or synthesizing new ideas from its trained data.
I think that will require a specific plot-generation AI engine that the more general writing engines that we presently have cannot do. It's not really within their pattern-recognition purview.
Agreed.
Models cannot mix domains like humans.
Say the subject is chess; you need one bot to play, one that only generates images of chess, one that writes about chess. There is no connection between the 3. Only a human can do that.
20th centurians have quite a lot of complaints that 21sters don't care about at all.
what a bunch of hand-wringing nancies. this is gonna be *fun.*