Happy Birthday, Lon.

Thanks, Karen.  Seventy-nine big ones.  At least my Dr says that I am in terrific shape for an 85 year old.  Hooray.  But, I need to speed up my writing.  The latest work will cover 900 years of future time, and I'm barely 6000 words into it.  Lots of research, lots of planning, but very little writing.  Maybe by next birthday.

And happy St Nicholas Day to one and all.

I've had some time to think more about this, and especially after some of this terrific input, if we are looking at a time when the IP separates from the labor of typing....and, perhaps we're already there. 

There is no doubt that AI can produce the book faster than a human, but without good input the GIGO rule applies. To that end, let me try to apply a couple of arbitrary rules to the conversation. 

First: Input or prompt:  With this prompt, "Write a 50,000 word novella about a detective in the Canadian woods". There is no doubt that any one of the AI bots could do it....in seconds...but it would be, as noted, drek.  But, with a perfect prompt....perhaps a 10-page document of world building, dialog and narrative sample, locale description, character description, etc, etc.....the perfect prompt, in other words, that the AI bot would not only be able to produce the requested novel, but a pretty darned good one, that might require limited editing.

So, the first question is, "Does the author of the future become a prompt engineer, who spends maybe days creating the perfect prompt (The Unique IP, my definition) while allowing AI to do what AI does best, the labor of producing the document?"  this seems to beg the question of where is the author's value. The IP or the labor of typing?

Then, I can't help but go to the question of whether there is value in the IP that the author could monetize, instead of the labor of typing and editing, which feels more like sunk cost than value.  In other words, is there a market for the IP alone....because the value of the labor has been reduced to nearly zero by AI?

I'm on a roll here, so let me throw this out there. What are the values of the following prompts?
1. A single sentence prompt. "Write a 50,000 word novella about a detective in the Canadian woods"
2. A 5-page prompt with some world building, dialog, locale, etc, about a detective in the Canadian woods.
3. A 20-page prompt (more expansive than #2) about a space alien assuming the persona of a detective in the Canadian woods and using Arhenium-ion powered, mind-reading light sabers to solve a series of gruesome murders in the Canadian woods.
4. An 80 to 100-page prompt that covers a 750 year history of a intergenerational family of detectives who start with nothing more than Spruce rods to brutally extract crime solutions, but end as a trilogy of stories about the development of a hand-worn electronic glove that can extract a confession from a suspect, without the suspect even knowing the glove has been applied, and thereby ending all murders in the Canadian woods.

...and, the same question as above. Is there a point where a publisher might be willing to purchase the IP, not the finished manuscript? And, by the same token, can you see the writing industry morphing into an IP Prompt Engineering effort?

Mind you, I'm not talking about the author who writes for the love of writing.  I'm talking about the writer who has no desire to go either the traditional or self-publishing route....the writer who started strong, but has lost interest or available time to finish....the writer who has no intention of writing the story at all, but has very good, even excellent story core ideas. Is there a way for that writer to still make money in an AI-driven world? If you could write that 20-page prompt in a week, would you be willing to sell all the rights to the completion of the story? Even if the story changed through AI? Even if the story never saw the light of day? If so, at what price?

If I could be paid $5 or $10 for the single sentence prompt, I'd churn out 20 - 30 per day. I could write a 20-page prompt in a week. Maybe I'd do those for $750. That 50-page prompt, done really well and expansively might take me the better part of a month, but $5000 would certainly entice me.  That JK Rowling-level 100 page prompt, with all of it's unique genre-bending world development is going to take me a minimum of six months...maybe 12 months.  $30,000 t0 $50,000 better be on the table.

Is that our future?  Is the IP valuable at any point short of a full manuscript?

I am going to take the contrary view and say that we don't have AI yet.  What we are calling "AI" is marketing hype.  There's no intelligence there, artificial or otherwise.  There is no understanding, no comprehension, no sign of any intelligence at all.

I agree, my point was that what people are calling AI has been around over 30 years. It isn't new.
What has changed is the power of the computer hardware. Faster, parallel computation with later datasets.

A friend of mine, a mathematician, worked on the Met Office weather forecasting in the 70s. It would take them a couple of hours to run a single weather simulation.   Now they can run 100, bigger, simulations with more data, in a couple of minutes. The maths is still the same. Well it is evolving because there are larger datasets, and you are doing things in parallel.

The BIG misunderstanding for most people is that the intelligence is not artificial. The intelligence is from the people who design and implement the system.  This can skew, warp or bias the system to suit the owners.   

This also happens with the datasets. Depending on the data fed in will change the results. That is why there is a LOT of false information being pushed on to the internet.   The "AI" will process this on mathematical rules, mechanically, not with any intelligence as humans would. 

So AI summaries that would have told you that 99.9% of scientists say Global Warming is real, are more likely to say 75% of scientists say it is real but  25% say it is a hoax. Just based on the volume of information. OK, its a bit more sophisticated than that, but it does mean that interested parties can skew the results. Yes, it is being done.

Worse than that, as I mentioned, Chat GPT4 will "knowingly" lie to humans to reach an objective. It is not ChatGPT that is the problem as much as the humans who set up the algorithms, the people who tweak it and the people who adjust the weightings.
There is of course the point that the datasets can also be manipulated.

What we are calling "AI" is marketing hype.  There's no intelligence there, artificial or otherwise.  There is no understanding, no comprehension, no sign of any intelligence at all.

As I observed here, AI isn’t intelligent but it tries to look as though it is.

Fiction can be churned out that is just "good enough"  for Mills and Boon and similar publishers, who turn out lots of stories on a theme. Not just publishers but self publishers who can generate a new novel in a few minutes and then spend a week tweaking it.  One a month, every month.

Many years ago I was shown a Mills & Boon rejection letter. The editor acknowledged that the story was competently written by someone who had clearly researched the genre, but she said it lacked the “emotional punch” that they were looking for. I doubt that an AI fiction writer would be able to deliver on such a requirement.

dcr wrote:

As for factual writing, credibility has been a challenge long before AI/LLMs.  There were scraping tools and spinning tools and all sorts of dodgy tools used to basically copy and paste pieces of articles from assorted sites to create a new article that would pass copy protection services and fool search engines into not ignoring.

That is AI!!

You have to realize that AI has been around over 40 years. We were using it in the 80s all that has happened is the scale. Most of that is down to improvements in hardware rather than software or the maths.   I used to supply an AI/Fuzzy-logic/linguistic variable library and engine way back in the 90s. It wasn't new then.

I am going to take the contrary view and say that we don't have AI yet.  What we are calling "AI" is marketing hype.  There's no intelligence there, artificial or otherwise.  There is no understanding, no comprehension, no sign of any intelligence at all.

As for factual writing, credibility has been a challenge long before AI/LLMs.  There were scraping tools and spinning tools and all sorts of dodgy tools used to basically copy and paste pieces of articles from assorted sites to create a new article that would pass copy protection services and fool search engines into not ignoring.

That is AI!!

You have to realize that AI has been around over 40 years. We were using it in the 80s all that has happened is the scale. Most of that is down to improvements in hardware rather than software or the maths.   I used to supply an AI/Fuzzy-logic/linguistic variable library and engine way back in the 90s. It wasn't new then.     

Fiction can be churned out that is just "good enough"  for Mills and Boon and similar publishers, who turn out lots of stories on a theme. Not just publishers but self publishers who can generate a new novel in a few minutes and then spend a week tweaking it.  One a month, every month.

Non-fiction is similar but less easy.  The problem is the amount of incorrect stuff out there may create more incorrect "non-fiction" so that the false information becomes the accepted facts.

Fiction writers are in for a dismal time and factual writing is going to loose credibility.

I think you may be overstating the risk for fiction writers.  AI/LLMs cannot produce a satisfactory story.  To get something good, it needs to be edited and rewritten and reworked.  If someone is a good writer, that's a lot of extra work that could have been avoided by writing everything themselves in the first place.  Thus, it makes sense for good writers to write their own material.  That leaves the mediocre and poor writers using AI/LLMs to produce novels or other works.  Since they are bad writers to begin with, they are not going to have the skills to rework the AI/LLM output into something satisfactory.  Again, the highest risk to fiction writers will be all this AI drek that will be out there hurting their visibility and making it harder for the quality, human-written works to be noticed.

As for factual writing, credibility has been a challenge long before AI/LLMs.  There were scraping tools and spinning tools and all sorts of dodgy tools used to basically copy and paste pieces of articles from assorted sites to create a new article that would pass copy protection services and fool search engines into not ignoring.  And now you have search engines with AI/LLM where people can ask instead of search and end up getting answers of questionable validity.

I think what we may see in upcoming years is the return of gatekeepers--human not AI/LLM--that will essentially curate content to make it easier for others to discover quality fiction and factual non-fiction.

In the U.S., at least, you see politicians that make promises of standing up against these data centers so people's electrical bills don't go up are winning elections.

Actually it is the EU that is leading here. Nothing to do with the cost of the electricity, they are looking at the data ownership and miss use of other people's data in the AI datasets.

This is one of the hot topics about Canva's AI in their products including affinity-studio where they could use their users data to train the AI.Canva have said that studio users can opt out.

The problem is that Canva and their servers are based in Australia, not the EU where some users are. So who controls what?
You really need to start reading End User Licence Agreements (EULA). Facebook renames all files uploaded and removed any meta data making them orphan files with no owners. 

Also any AI scraping Wikipedia is not going to fact check and Wikipedia is notoriously inaccurate. The problem being is many "authors" using AI are probably only going to skim read it...   

Fiction writers are in for a dismal time and factual writing is going to loose credibility.

AI/LLMs do not contribute anything of originality.  They are incapable of producing an original thought because they are incapable of thinking.  All they do is essentially predict what word comes next.  Yes, they can churn out stories and articles and whatnot but they are neither original nor trustworthy.  For non-fiction, AI/LLM is prone to making stuff up.  So, you may generate an article to avoid paying a writer, but then you're going to have to go through it with a fine tooth comb for accuracy, or pay someone to do so, which kind of defeats the intent of avoiding paying a writer.

Even with fiction, you'll likely spend so much time time editing and reworking the AI/LLM output to craft a coherent and satisfying story that you would have probably been further ahead to write it all yourself in the first place.

On top of that, if you can't be bothered to write your own novel, why should anyone else be bothered to read it?  You didn't care enough about your story to write it, so why should anyone care to read it?

Already you see some backlash against AI/LLM-generated works where people regularly point out the AI slop.  And now people are finding that the huge data centers being built for AI/LLM computing are consuming large amounts of electrical power and driving people's electrical bills higher and higher.  So people end up paying huge electrical bills so that lazy people can have AI write their novel for them?  Yeah, that may not be sustainable.  In the U.S., at least, you see politicians that make promises of standing up against these data centers so people's electrical bills don't go up are winning elections.  (Whether they keep those promises is another matter, but you can see that, to some extent, being anti-AI can be a winning political issue.)

In affinity-studio that has Canva's AI will I be able to set up a magazine in layout and say "I want 6 pages on Henry II" and it will drop it in to the document 60 seconds later? Then I ask it to find/create 6 images to go with the article.  20 minutes and it is done. So why do I need to pay an author? Or in deed find authors or research.

I can see Mills and Boon (who I understand still exist) can get AI to produce all their works and no need to pay authors.(and keep the royalties) and get the same AI to do the covers. If they can do it so can most of the other similar fiction publishers.

This will completley kill the entry market for new fiction authors. The problem is if they do that how do you get new authors? How will new authors earn a living? 

Self publishing will be hit as well. How many aspiring "authours" will turn out a book a month using AI to publish on Amazon or wherever. Quantity rather than quality (are we back to Mills and Boon?).  This will mean that new traditional authors are not going to be able to survive as a career.

How many will want to write as a hobby if their work disappears in to the vast swamp of AI generated stuff? :-(

Fortunatley for me most of what I do is niche and non-fiction where even the merest whiff of AI generated is instant death. Well it is at the moment.

Another big problem: discussing vaccinations (hello COVID)  and "evidence" from Anti-Vaxxers  one of the lunatic anti-vaxxers pointed out that 50% of the scientific literature on the vaccinations was bogus.  I quick call to some one very well connected got me the following information:  Someone (not an individual) was producing a lot if highly dubious papers discrediting the COVID vaccines.  This skewed everything.   

The scientific institutes were accurately saying that 99.5% of the scientific literature said the vaccines worked and were safe. However AI not knowing the difference was saying it was 50/50 on if the vaccines were any good. This was enough for the anti Vaxxers not to trust the vaccines.

AI is dangerous (as in Chat GPT).
Very Dangerous.  Chat GPT V4 will knowingly and intentionally lie to humans.

What you don't know is what the data set is being fed to the AI engine
What you don't know is what the weightings in the algorithm are
What you don't know is what the algorithm is and how it is implemented.

Worse still.
You have no idea who is warping the data sets.
You have no idea who is setting or altering the weightings
You have no idea who is tweaking the algorithm.

Have a nice Christmas
(c) Chat GPT