"The Future of History Writing in the Age of AI" Topic

"I use AI — specifically ChatGPT, Claude, or Gemini Notebook (formerly NotebookLM) — daily. As an editor and writer, I realized early on after the release of ChatGPT in November 2022 that I had a choice: adapt or become obsolete.

Generative AI is impacting every industry, not just mine. This includes history writing.

The good news is that AI is easy to use once someone starts experimenting with it, and it can be useful for brainstorming, editing, and research. But for all its benefits, there are some serious issues, notably when it comes to accuracy.

As we navigate the "jagged frontier of AI," history writers will need to adapt their writing styles.

But first, we need to recognize the merits and pitfalls of AI…"

Emerging Civil War

link

Armand

I find the AI adapts to me.

GIGO

John, yes, I agree. The more of yourself you put into an AI the better the results.

Actually, it appears all the good stuff is going in, old rare books, musicians and actors entire careers, so on; and slop, messed up hands, endless loops, horrific writing, so on are coming out.

And, as doc pointed out elsewhere, it can take multiple attempts to get the desired result, wasting resources, where a human is more likely to present a better result.

I've seen some doozies on ink master, tattoo nightmares, but I think AI would take the cake.

There are ways to prevent the GIGO and AI slop but it is easier to complain than find an easy solution.

The ChatGPT image has problems with details on vehicles but I was able to find an easy work around. It's not magic but does save a lot of time and will get better. Using AI for images is much easier than learning a graphics package and cheaper than hiring someone.

I think of LLM as an idiot savant who has access to all of the knowledge available but is incapable of making a final determination of how accurate it is. You need to narrow down the prompt, direct it to areas of known accuracy and to stay away from things like Wikipedia. It's all in the prompt.

You can even tell the AI that is did a terrible job and to try again and look harder and stay away from slop. It actually works!

If you use it expecting for some super intelligent response and 100% accuracy without some knowledge of prompting you'll be VERY disappointed and it is your fault.

Wolfhag

Wolfhag, yes, precisely.

Start stockpiling your pulse rifles in the 40 watt range because you're going to need them. Skynet.

Thanks

Armand

No interest in AI while writing. I even dislike auto-correct apps.
As an ex- college professor and HS teacher, a lot of AI is the same as plagiarism.

"it is easier to complain than find an easy solution."

Oh, I know the solution. Putin already apparently did it. I just think we're gonna need a bigger boat.

"I was able to find an easy work around."

Perhaps you'll pass that onto the WW2 slop makers.

"AI for images is much easier than learning a graphics package and cheaper than hiring someone."

A bit like a particular European country in the 40s. They found its cheaper to feed a lot of people, if they don't need to eat anymore.

AI is cultural genocide.

"You need to narrow down the prompt, direct it to areas of known accuracy and to stay away from things like Wikipedia. It's all in the prompt."

Sounds like a full time job. An artist or writer could create something more (or less, if asked, I'm sure) original in a reasonable time frame, saving an unnecessary education.

"You can even tell the AI that is did a terrible job and to try again and look harder and stay away from slop. It actually works!"

So the polar opposite of Google AI then.

"and it is your fault."

"No."

I certainly never asked for it. It's the fault of the people who have allowed it. Perhaps on the individual for using it, but certainly not for not knowing about it or wishing to.

Perhaps it's any complainant's fault they can't draw/write? Perhaps they should take a course on that instead?

"You need to narrow down the prompt, direct it to areas of known accuracy and to stay away from things like Wikipedia. It's all in the prompt."

This doesn't do much for others AI utopia we're all meant to be living soon.

Maybe if you live in a standardized cube surrounded by noisey data centres.

Though people won't have time for recreation and games if you have to "narrow down the prompt" all the time. Remember it's supposed to be infiltrating all aspects of life.

H, you have to master tools to use them effectively. I have now been actively and DAILY involved with AIs for two years, and am still learning, yes, how to do prompts. But also how to stretch my own imagination in terms of asking for help. The problem is not the AI, the problem is ME.

You are aware, from another thread, that I have been working on casualty markers for my skirmish games. Spent several hours yesterday, and then again today, trying to get images right, then get them formatted to take to FedEx to be printed. And now I can do in minutes some neat things that six months ago I could not do at all, and last week took hours. The AI image function has improved, but also my own mastery of it has developed. And it has been greatly amusing and entertaining and satisfying.

Not everything is on the internet. Nothing in, nothing out.

(smile)

Armand

I know nothing about graphics design software other than MS Paint from the last millennium. All I could do is draw stick figures with it.

I wanted to design the graphics for a box cover and tried ChatGPT without taking any lessons. Here is the result:

In my prompt I told it to look over other WWII cover art and what the best practices are to have an attractive cover. It was a battle scene in Ukraine between a Panther D and T-34/85 which is on fire. I did have to edit some small changes using MS Paint.

Whether you like it or not does not matter. There is no way in my lifetime I could do this without AI and a designer would want $50 USD/hour or more.

Wolfhag

Cheaper and easier doesn't mean better.

After all, what pictures are AI going to steal from if people aren't painting or taking photos?

If everyone uses ai instead of actually making images, and AI only has slop as new material to copy, it can only create degraded images.

If it's so hard to spot ai images online now, and AI used images on line, and AI makes mistakes in making images from what it finds online…

So, have fun while it lasts.

PS

There's a worthy experiment for the above,or the doc, take a WW2 tank, feed it in, then keep feeding it's slop back in for, 10, 50, 100 generations.

I'm keen to see the results.

PPS

And so it begins.

The above is just first generation slop, right?

Already the tank on the left has some very odd looking tracks. The spare tracks look worse than what they may need to replace. Okay, so maybe they already replaced some and just hung the old for armours sake. Sure, no harm in a story to go with the image. The front track has some weird circle, perhaps gleemed from a 40k tank eye??? And the edge of the tred from Bart Simpson.

Someone in the Jerry can factory needs to be shot. You know the meme, "What the blank is even that!" *Not a can, but still screwed up.

I'll keep looking.

"best practices are to have an attractive cover."

Shameless plagiarism?

Hmm…

PPS

So, let me get this straight.

AI rips off other's work, then, to make it all legal, it changes it just enough. Ditching historical accuracy for a Wes Craven routine. Of course, camoflauge is completely all the same and not unique at all, so no need to bother there. Straps? What are those?

I love these types of pictures. Hours of fun.

I do love picking pictures like this apart. Its not personal.

I remember when a local town had an event. They put out a program book with a photo of the crowded jetty…

You can imagine.

Half poles, people chopped up, the more you looked the more you saw. I apparently caused a stir when I contacted the publisher to complain.

It does have its rewards, like when I was shown a recalled coke poster with some ice containing a woman's lips and a man's…

Then there's chip art. That's safe to look up. Not sarcastically.

There are vids on YouTube where people call out WW2 AI slop, not dissimilar to the picture above. They have a good time picking them apart.

Ok.

So, the tracks behind them don't look right. The angles and perspective looks wrong.

That dead guy? If that's his neck?? Then his left hanging arm has an extra section in it, or his shoulders are too long.

I'm not too familiar with what those green things are?? Eyes? Anyone know?

Where is it's right headlight?

Look, if your doing things for experts, even just armchair ones, don't give them this. Unless it is the game, spot the historical inaccuracies.

It may be just as cheap??? To look into those sites selling photo licences. Getty and what not. I think you can just pay a small fee to use an image. I'm sure they would have lots of tanks. Just double check if they can be edited, if you want to put a scene together in Photoshop or paint.

Thanks

Armand

No problemo. Picking apart pictures is my type of fun.

The area where the other headlight should be, next to the cross, directly under the turret on the front panel.

You can see a patch of yellow that doesn't seem to match the rest and it is flat without detail, the same with the patches of brown and green directly below it.

I suspect this is one of those paint edits?

I'm guessing it really screwed up the light??

Ah, wait. The other light, that is there, is an obvious paint job. All different, looks to be floating above the underlying image. It does have some 3d shading. At a guess it may have been created in and pulled from word??

Any chance of posting the original image? For comparison?

On the other tank, the tool box, hatch and grey circle between the green things look like possible paint edits also. Flat floaty solid shapes of odd colour.

I'm not sure how much was AI compared to stuff added later, but there are some possible issues with the pictures down the bottom.

I'm no stop watch expert but that one looks wrong.

And the far away 2 dot on top of the die looks larger than the closer dot??

Not sure what that terrain thing is?

Back to the main pic and there seems to be pictures hidden in the smoke. Can you find captain hooks hook?

A look at the original, pre paint, image may be clearer and easier to see.

Hint. Hint.

The H Man,
Let me clarify why I made the original post.

I wanted to show what a beginner with very little knowledge of artificial intelligence (AI) could create using an older, free version of ChatGPT and a two-sentence instruction. An instruction given to AI is called a "prompt." I was sharing an experiment, not claiming expertise in graphic design or presenting the image as the best that AI can produce.

It is fine not to like the image. I am not offended, and I agree with several of your criticisms. ChatGPT generated the image from my instructions; I did not draw or paint it myself. Anyone can have an opinion about art. You do not need professional training to say what you like or dislike. However, judging one image and judging the capabilities of AI as a whole are different questions. What direct experience have you had with AI and image generation?

After creating the image, I asked ChatGPT to research game packaging and marketing practices, then suggest changes based on those practices. I was experimenting with a possible game box cover, not preparing it for publication. I also had concerns about possible copyright or intellectual property issues. Intellectual property refers to creations people may have legal rights to, such as artwork. Those concerns would need to be addressed before any commercial use.
I sent the revised image to two friends with decades of experience in game design and production, including numerous projects funded through public contributions. One is in the Wargaming Hall of Fame. Both thought the image worked as game packaging because it communicated the subject and gave potential buyers an idea of what was inside.

That does not make every detail historically correct. Many game covers use stylized artwork rather than exact historical renderings. In my view, a cover needs to attract attention and communicate the game's subject, as well as provide an appropriate level of visual and historical accuracy.
I agree that the image has flaws, and my own graphic design experience is limited. My disagreement is with using this beginner's experiment as the basis for broad conclusions about AI. Neither of us has enough experience with the newest tools to make a complete assessment of their capabilities.
More detailed instructions and good reference material can help AI produce a more historically accurate vehicle image. They do not guarantee a perfect result. Someone who knows the vehicle still needs to check the details.

Copyright and accuracy are legitimate concerns. AI can also produce "hallucinations," meaning information that sounds convincing but is wrong or invented. Restricting an AI tool to selected source material and checking its answers can reduce these problems, but I would not claim that every copyright issue or error can be eliminated.

Here is the TMP post I mentioned, which fooled some of our members:
TMP link

The older, free version I used had limitations, especially for images requiring precise details. My experience has been more useful in other areas. These tools change quickly, so an assessment should identify which tool and version was used and what it was asked to do.

Docmcb, whose "doc" refers to his PhD, lives in the next town over. We are practically neighbors. He knows more about this than I do and uses several current paid AI tools. I am watching his experiments before deciding whether to spend more money myself.
Here are a few examples of how I have used AI or seen it used.

For my local fire department, I used ChatGPT to help create a fundraising letter with graphics and QR codes. A QR code is a square pattern that someone can scan with a phone to open a donation page. I do not have computer graphics skills, so that assistance was useful.

For a nonprofit organization, I asked ChatGPT to research its locations across the state and organize information about local activities, fundraising, and community outreach into a database. I also asked it to compare those activities with similar organizations elsewhere in the country. Writing the instructions took about five minutes, and the work saved me many hours of research and organization. The information still needs to be checked before it is relied on.

ChatGPT then, acting as an assistant, offered to develop a statewide implementation plan with goals for the first 30, 60, and 90 days. I agreed, and it produced a draft quickly. The organization's president liked the result. I could have done the work manually, but AI helped us get a useful starting point much faster and saved hours of work.

My son develops hardware for a company that makes military communications and electronic warfare equipment. Hardware means the physical equipment; software means the programs that run it. Electronic warfare involves systems that detect, disrupt, or protect electronic signals.

A software problem was delaying one of his new products release date. One evening, he used Claude, another AI assistant, to work through the problem. After identifying the issue, he asked Claude to write code to address it, then tested the code himself. The company installed and ran it the next day, and it worked. He received a $20,000 USD bonus, equal to 10 percent of his salary, and was promoted to director of hardware and software engineering. He has no four-year degree, and his experience with AI is self-taught.

That example shows why I see AI as useful beyond image generation. It helped someone work through a practical problem, but his knowledge and testing were essential to the result.
I would encourage you to experiment with something that interests you rather than judge the entire technology by this one image. You may find an application that is useful to you, even if AI-generated artwork does not appeal to you.

My opinion, based on the seminars and workshops I have attended and my efforts to keep learning, is that AI can amplify both helpful and harmful human activity, much as the internet has. It can help with research, writing, and technical work. It can also make scams, cheating, and hacking more effective. I am interested in its potential in medicine and computer security, but I do not know enough to predict how far those applications will go.

I have also read reports about concerns over safety and security at AI companies, including OpenAI. I do not have enough detail to judge the specific cases. "Guardrails" are rules and technical controls intended to limit harmful or inappropriate use. The question is how well those protections work in practice.

The internet has spent years responding to security problems as they arise. I worry that AI development could follow a similar pattern, although AI may also help find and fix some of those problems. I do not know enough to make firm predictions. For now, I am interested in learning what these tools can do, checking their results, and using them where they help.

Wolfhag

FYI: Before posting this, I used ChatGPT to correct spelling and grammar and make the wording easier to understand for readers with little knowledge of AI to understand.

Wolfhag, thanks.

There will always be nay-sayers, and I suppose they serve a useful purpose, up to a point, though beyond that point it may be necessary to ignore them.

I hope others will experiment with AI and share their experiences with us.

H Man, have you done much writing of history yourself? or graphic design? Can you share some things you are proud of?

H:

And, as doc pointed out elsewhere, it can take multiple attempts to get the desired result, wasting resources, where a human is more likely to present a better result.

That is not really what I said, except the very first part: yes, it takes multiple attempts. The next two parts are nothing I said, nor would have said, but the H Man's opinions. Which are wrong. It is not a waste of resources to learn how to do things better. And it is by no means clear that an unaided human can present a better result. Better tools produce better work.

So we have the H Man casually mixing what I said with his own unsupported opinions which he seems to attribute to me. Or more likely, just sloppy writing on his part.

Tell me again how much better humans do it?

AI allows me to:

-Create specific types of "things" of much higher quality than I could ever possibly create myself.

-Create things within my time constraints. Without AI, I would never have had the time to create them. I work full time at my W-2 job.

-Improve the things I can create on my own. Error checking, gap checking, quality control, etc… I can complete both work tasks and personal projects to a much higher standard and in much less time that i did for these exact same things 2 years ago.

If I were willing to spend unlimited cash (or at least more cash than I'm willing to spend on things like this), yes, I could pay someone to do most if not all of the things AI helps me with. However AI costs me $20 USD a month which is actually within my budget.

Are we all going to die when AI goes rogue, or be oppressed by power hungry government officials? Most certainly to some degree… at least the later.

Thunder, yes. We will always be oppressed by the power hungry, but sometimes a sling shot beats Goliath.

M experience is the same: mostly I COULD do on my own what GROK does for me, but it would take many hours instead of mere minutes. And since my time is valuable I would NOT do it, so it would not get done. AI expands what is possible.

I spend $50 USD/month, which is also within my budget, and so have ChatGPT and GROK to play off against each other. But mostly it is entertainment: I spend probably ten hours a week directly working with one or both, and 40+ hours of fun a week is pretty good for $50. USD

"Anyone can have an opinion about art"

Relevance?

"However, judging one image and judging the capabilities of AI as a whole are different questions."

Who chose the example to present?

" I also had concerns about possible copyright or intellectual property issues. Intellectual property refers to creations people may have legal rights to, such as artwork. Those concerns would need to be addressed before any commercial use."

Yet chose to publish it on the open internet.

" sent the revised image to two friends with decades of experience in game design and production, including numerous projects funded through public contributions. One is in the Wargaming Hall of Fame. Both thought the image worked as game packaging because it communicated the subject and gave potential buyers an idea of what was inside."

Thank you. That is appreciated. I do have games I'm supposed to be working on. That does help.

"In my view, a cover needs to attract attention"

Done.

"My disagreement is with using this beginner's experiment as the basis for broad conclusions about AI."

Go look up the WW2 slop on YouTube. People, right now, with entire channels of such nonsense. Its not just an amateur thing.

"Neither of us has enough experience with the newest tools to make a complete assessment of their capabilities."

Speak for yourself. I simply looked at the image provided and pointed out its flaws. No experience required.

Look up the YouTube vid of the train museum where everything was just AI nonsense posters.

"FYI: Before posting this, I used ChatGPT to correct spelling and grammar and make the wording easier to understand for readers with little knowledge of AI to understand."

FYI: To many words in to big chunks. Less is more, so on. (Well, at least I'm spacing this lot out!)

"There will always be nay-sayers,"

It was the nay-sayers who freed the Jews, bombed Japan, and threw themselves up beaches into machine gun fire. Give us a little respect.

"I suppose they serve a useful purpose, up to a point,"

I think they definitely deserve more respect than that.

"I hope others will experiment with AI and share their experiences with us."

Or, just watch the news. Its been on a lot recently.

"H Man, have you done much writing of history yourself? or graphic design? Can you share some things you are proud of?

Not rewriting, no.

Yes, lots.

Your reading them, dude.

"And it is by no means clear that an unaided human can present a better result."

Most humans can count to five.

And as above, most humans wouldn't have copied a tank from someone else's image without realising it.

"So we have the H Man casually mixing what I said with his own unsupported opinions which he seems to attribute to me. Or more likely, just sloppy writing on his part."

The other 2 points naturally follow the first. Unused results are a waste (example, film design, the most wasteful part, unfortunately, switching a lot to digital to save paper and such. For AI it's mostly water and power) and as clearly displayed in the tank image above,a human wouldn't have let things go that far, thus presenting a better result.

"Tell me again how much better humans do it?"

Tell you again?

Well, it took a human to pick those flaws in the tanks.

"-Create specific types of "things" of much higher quality than I could ever possibly create myself."

Yeah, I think that's it.

AI is definitely the new 3d printing.

It gives people with no skill a unearned sence of superiority. Instead of actually learning a new skill for themselves.

Just pick up a pencil and draw. Or write. We all started somewhere. Its not that hard.

I can recommend how to draw comics the Marvel way as a fun book to start you off drawing.

"Create things within my time constraints. Without AI, I would never have had the time to create them. I work full time at my W-2 job."

Unfortunately, quality takes time. There is no getting around that. Take those 2 tanks above as an example.

"-Improve the things I can create on my own. Error checking, gap checking, quality control, etc… I can complete both work tasks and personal projects to a much higher standard and in much less time that i did for these exact same things 2 years ago."

You need to be very careful there. There was a novel pulled because a proof reader had put it through AI that actually altered it and it wasn't noticed until publication. I think it was a main big publisher. Should be easy to look up.

"However AI costs me $20.00 USD USD a month which is actually within my budget."

Unfortunately, like time, the less value put in, the less value comes out. However skill can help offset this. So you really need one of them. Most comic stores should stock the book, or its available online or to borrow from the library (they should be able to order it in, if needed).

"Are we all going to die when AI goes rogue, or be oppressed by power hungry government officials? Most certainly to some degree… at least the later."

Just target any pipe or cable going to a datacenter and you'll be fine.

H, you are exhausting, and not very helpful. Thanks for playing.

Typical.

No actual response, just childish remarks.

And, of course, suddenly, apparently, it's just a game we've been playing all along.

"In that specific battle scene, the shapes hidden inside the smoke and explosion effects are actually human skulls.
When Wolfhag manipulated the image to depict the Soviet T-34 tank bursting into flames, he digitally layered images of human skulls directly into the billowing black and white smoke plumes.

This hidden visual trick—often referred to as an "Easter egg" or subliminal imagery—is a classic dramatic technique used in military artwork to subtly symbolize death,
destruction, and the sheer grim reality of the heavy combat on the Eastern Front. When the AI smoothed, blended, and integrated his manual edits together, it partially masked the skulls into the contours of the smoke, making them look like a natural but eerie part of the roiling fire cloud."

That's from Google AI.

There does look to be a skull, more a necron head, just left of captain hooks hook.

So is this how it was done, or is AI just as rubbish as is often suggested?