Query on AI writing apps

Stephen Palmer said:
Let's reframe the question: Will AI writing apps make human beings redundant? No.
Click to expand...
Since we are all of us more alike than different, one could say human beings are largely engaged in making redundant human beings.
 
I feel that a lot of the fear of AI is due to a vast underestimation (and lack of understanding) of just how complex human activities, such as writing, truly are.

Read through any how to books and one finds core beliefs of mysticism about creative writing. Maybe we no longer have muses come and touch us, but we have to listen to our subconsciousness to spew forth ideas. Or wake up and write down our dreams. What in the world are dreams and where do they come from? Computer science does not have any clearer idea of how this is done than the myriad of writers on this topic.

Instead, AI assumes that, if we merely cram enough random text into a machine learning program, then somehow a maser algorithm will appear that will create works that appeal to popular culture (never mind that popular culture is also a mysteriously changing beast). We believe (or fear) that an average of all written works will be uniquely different and appealing. Even if this approach succeeds once. what causes the program to create anything different for a second attempt? What randomization causes the same algorithm to create a different result?

I find the human creative process fascinating and mysterious. I feel that it is quite naive to be able to replicate this process through mostly random actions in a relatively short period of time (as compared to the evolution period of life or even just man). No, I don't believe that AI programs are going to replace writers anytime soon, not even the most formulistic, hack writers need to feel threatened. We are far from understanding what creativity is, much less being able to recreate it in some manufactured format.
 
AI keeps me up at night. I think there will be hell to pay in the job market, and it's insane that our species is so dead-set on rendering itself obsolete. That said, when it comes to art---not just the creation of pretty-looking images or prose, but ideas---generative AI is very poorly suited to the task. All ChatGPT can do is remix other things people have written. It understands nothing. Sure, the technology will probably evolve to the point where it can write boilerplate SF novels, but what use is a novel without a soul?
 
>what use is a novel without a soul?

To those looking only to make a quick buck, such a novel is very much of use. Invest a penny, turn a dime. A great many people will take that trade, and if the reader can't tell the difference, then game on..

But I think a direct comparison is not quite on point. First, there are many different kinds of writing. Why not have AI write a textbook? Or a reading primer? Why not have it write an instruction manual? Lore for an RPG video game? Or, as is already being done, write journalism; feed it the facts of the story then let the AI do the rest. Let it do sports coverage or a weather report.

Lots of kinds of writing.

But even in the realm of fiction writing, maybe the AI could give us new forms of literature. After all, just in recent decades we've had fan fiction, LitRPG, manga and its many relatives. All forms of literature nearly unimaginable fifty years ago. AILit [tm] could be a whole genre complete with its own awards (granted by an Awards AI). The ideas don't have to be new in themselves; it can just be new combinations of old ideas. Plenty of humans have rolled up plenty of miles on that odometer.

I do see the threats, as do many of us. Or, not so much threats as thefts. PersonX doesn't get to make money off my writing without compensating me. Right now, there's no recourse. But artists on many fronts are already moving to the offensive. It will be a battle because those who seek advantage are the get-rich-quick crowd. Appeals to ethics won't work. Legal redress works only for the individual case, and they'll always be finding ways around the latest bulwarks. In the music business, piracy (I don't want to open up that box here!) was only really shut down when the biggest owners of music took coordinated action. That "solution" has been both good and bad. Like as not, that's where AI-gen will wind up some several years from now as well.

Meanwhile, I got a book to write.
 
sknox said:
Why not have AI write a textbook?
Click to expand...
I do not feel that using an AI program to create a textbook would be a good idea. The problem is that an AI (currently Machine Learning) does not evaluate the correctness or accuracy of what information that it is trained on. The program merely aggregates all the information it is trained with and repeats what it sees in common. This means that the AI is likely to generate views and ideas that are commonly held or held by significant minorities. In the case of advancing science, the AI would likely ignore new advancements in lieu of older accepted theories. Even the teaching of history has evolved over time.

A real-world example of this problem is shown in the article that I posted earlier. CNET used an AI routine to generate online articles. CNET is now going back and editing those articles for correctness. Just because something is done by a computer doesn't mean that it is correct.

When creating a textbook, a writer is likely relying on his or her accumulated knowledge and cross-referencing supporting texts. If, instead, one wanted to use a Machine Learning program to create something of similar quality, it would require carefully curating the texts provided to the program to include the most current interpretations and avoiding questionable or invalided information. I suspect the curation of data would require more effort than having a knowledgeable individual write it.
 
Again, I think the line is drawn to sharply. Let the AI draft the textbook, then let it be peer reviewed. That's the way monographs work anyway.

The threat there is to academia, for while the real prestige lies in publishing articles and monographs, there's credit gained for publishing or contributing to a textbook. It fills out the vita.

In addition, there are a number of specialty type publications, such as a literature review, where the aim is not furthering knowledge but simply collating knowledge. At the other end of the scale, there's editorial commentary on collected manuscripts, which is unlikely ever to have AI writing.

In short, there will be a place for both. There will also be a place for entirely new forms, and for gray areas that will make folks queasy. It's only the binary choice portrayal I'm trying to shade a bit.
 
According to a New York Times article, some universities and high schools are already making plans to change their teaching methods by making the students do more work in the classroom, including rough drafts and monitoring their online work by having students provide documentation that explains changes they make to their rough drafts.

The idea is that the use of ChatGPT can't be stopped, it is inevitable that it will continue to get better and that people will use it. The way to handle the situation is to use it as an educational tool which should outweigh the risks.

One thing that might come out of this will be a remarkable improvement in the grammar of phishing emails.
 
No. Not ever.

Today's AI is the equivalent of ransom letters created by cut-out newspaper headline characters. Some artists have noticed that and are suing for copyright infringement.

To write good fiction you need to feel pain. It works the same as writing song lyrics.

But even if we had AI that felt, I'm leaning towards us still wanting to read books by actual authors.

As I have understood it, a modern author pretty much already needs to have an audience on social media for publishers to even touch them. That you have to do talks, lectures, and guest speeches to not fall off the radar. A juicy scandal or juicy story apparently helps a lot.

I'm not saying an AI couldn't do the same, but would you, as a reader be able to connect with them the same way you connect with your favorite author? If "Spare" was written by the equivalent of the Queens iPhone, would you care?

I think asking if AI will replace flesh and blood authors is a bit like asking if cars and motorbikes will replace flesh and blood 100-meter runners, or why we still have Chess tournaments (and I'm sure Go tournaments) when there are machines that can outsmart us all.

It's not just about speed or accomplishment. It's also about the human behind it. We would lose something if we had AI-generated novels, or robotic 100-meter runners, or chess tournaments between computers. It wouldn't be about us any longer and we'd stop watching, participating and, I'm sure, generating money for someone.

However, I fear the advent of AI might put even higher pressure on us to be visible, conspicuous, and online personalities and whatnot.
 
erikalm said:
No. Not ever.

Today's AI is the equivalent of ransom letters created by cut-out newspaper headline characters. Some artists have noticed that and are suing for copyright infringement.

To write good fiction you need to feel pain. It works the same as writing song lyrics.

But even if we had AI that felt, I'm leaning towards us still wanting to read books by actual authors.

As I have understood it, a modern author pretty much already needs to have an audience on social media for publishers to even touch them. That you have to do talks, lectures, and guest speeches to not fall off the radar. A juicy scandal or juicy story apparently helps a lot.

I'm not saying an AI couldn't do the same, but would you, as a reader be able to connect with them the same way you connect with your favorite author? If "Spare" was written by the equivalent of the Queens iPhone, would you care?

I think asking if AI will replace flesh and blood authors is a bit like asking if cars and motorbikes will replace flesh and blood 100-meter runners, or why we still have Chess tournaments (and I'm sure Go tournaments) when there are machines that can outsmart us all.

It's not just about speed or accomplishment. It's also about the human behind it. We would lose something if we had AI-generated novels, or robotic 100-meter runners, or chess tournaments between computers. It wouldn't be about us any longer and we'd stop watching, participating and, I'm sure, generating money for someone.

However, I fear the advent of AI might put even higher pressure on us to be visible, conspicuous, and online personalities and whatnot.
Click to expand...
Should publishing fame find me one day , I will be hiring out my customised Hatsune Miku to fend off the virtual mobs.
sabukaru.online

The World's Most Iconic Virtual Idol: Hatsune Miku<br/> — sabukaru

Hatsune Miku is a vocaloid [vocal android] software virtually personified as a 16 year old idol from the future with bright blue twintails and eccentric dance moves.
sabukaru.online sabukaru.online
 
I think eventually AI will be able to replace us as entertainers, but only if there is a huge shift in the culture towards it. Think endless streams of your favourite show on tap, whenever you want it, custom made for you personally. That's too tempting for someone who isn't too squeamish about where that entertainment comes from. Human made material might not go away but it might get marginalized.
 
Wayne Mack said:
When creating a textbook, a writer is likely relying on his or her accumulated knowledge and cross-referencing supporting texts. If, instead, one wanted to use a Machine Learning program to create something of similar quality, it would require carefully curating the texts provided to the program to include the most current interpretations and avoiding questionable or invalided information. I suspect the curation of data would require more effort than having a knowledgeable individual write it.
Click to expand...
And I think it goes further than just having the most current understanding of the topic. When I was an undergrad, the standard physical chemistry text was written by a guy called PW Atkins*, and so far as I can tell that was the case for students of my generation at other universities**. There was a clarity and elegance to the way in which Atkins presented the topic that simply made his text the go-to reference.

Textbooks get updated, and newer information gets incorporated, so there has to be something to the quality of the presentation that makes it worth producing new editions. A quick bit of Googling shows Atkins is now on it's twelfth edition.

Atkins also wrote a text on quantum mechanics. It wasn't a standard or recommended text at the time, but one of my professors recommended it directly to me. I bought a copy one Saturday morning and read it over the weekend like a novel. Again it presented quantum mechanics so clearly and elegantly that it just flowed like a story***, and so far as I am aware, Atkins's quantum mechanics book is now very much a standard or recommended text.

I am not convinced that a chatbot cobbling together the current knowledge on a topic will be able to bring that to a new textbook.


* And in one of those odd twists of fate, I ended up working for a guy who had been one of Atkins's PhD students.
** I'm pretty sure we have two copies in the house, mine and the Biskitetta's
*** It probably helped that at the time I was seriously immersed in the mathematics of quantum mechanics and operator mechanics
 
I would say that the formulaic writing practices by many authors and taught in creative writing classes is already a form of AI.
It is just a soft machine, generating product. Formula vs algorithm, you can barely get a cigarette paper between them.

I suppose it comes down to the notion that if you think original writing can be taught, then you will believe that AI can do it.
 
Biskit said:
I am not convinced that a chatbot cobbling together the current knowledge on a topic will be able to bring that to a new textbook.
Click to expand...
What is also true is that just because one is an expert about a subject doesn't mean that they are able, or even qualified to write about the subject such that the maximum amount of information is transferred from the page to the brain.

Part of writing is subtly connecting the dots in the background so that the reader gains an automatic understanding of the entire picture that is being presented by being able to continually build a picture using everything that is presented by using common ideas that they already know. Just getting flashes and snapshots yields partial understanding which requires multiple passes through the same material to get an idea of what is going.

Its like a very simple but elegant sketch of a cat or dog that brings up memories that fill out the complete picture versus a crude sketch using the same amount of ink but falls far short of evoking anything except the common name of what is. The ink is the data, the arrangement of the ink is a combination of art and science.

One of the points mentioned in the NY Times article about students using chatbots was that material written by the chatbot was considerably better arranged than that of a typical student.
 
Rick Beato, although basically talking about music, covers the broad span of AI taking over the arts (and homework assignments) here.
Although he doesn't mention it here, I think we are seeing an amplification of the zeitgeist that brought us the "Unplugged" movement a couple of decades back where musicians had to "prove" that they could play without massive technical assistance.

 
You must log in or register to reply here.
Thread starter Similar threads Forum Replies Date
Culhwch Query Letter Critique Critiques 12
Serendipity Writing Style Query Writing Discussion 5
J.L. Borstlap Query/Synopsis writing Plato Writing Discussion 4
Tyburn Writing a query to an agent. Writing Discussion 2
Warren_Paul Writing Query Letters Writing Discussion 3

Similar threads


Back
Top