• Hi Guest!

    The costs of running this forum are covered by Sea Lion Press. If you'd like to help support the company and the forum, visit patreon.com/sealionpress

ChatGPT and creating alternate history

From my side, ChatGTP has been the best research tool I've ever used. Since I write WarPorn alt history, getting the technical details right is just as important as the historical details. So it's AMAZING to just ask a question and get an answer. I do worry that I'm going to over-rely on ChatGTP, and get some errors or anachronisms, but that's why I have Beta readers. Here's a case that would have taken me a bit of time to run down the air intercept range of an F-106.

-HC

latest_5_star.png
 
That's no surprise as you're basically using it as a search engine for something verifiable. That's the best and least controversial use of it.

So how does Google compare then with ChatGPT, Google more the here is the link, the rest is op to you, while ChatGPT says, this is the answer but you cannot trust me 100 % if it is the correct one, thus you have to Google to see if ChatGPT gave you the correct one.
 
I assure you it isn't.

It's not, but technical research does matter, and it's surprisingly often a "brown M&M" that shows a lack of quality elsewhere. For every work that overloads on technical details (and they do exist), you have a million "machine gun pistols with heat seeking bullets" and "M60 Abrams with mounted flamethrowers", and even worse.
 
From my side, ChatGTP has been the best research tool I've ever used. Since I write WarPorn alt history, getting the technical details right is just as important as the historical details.

latest_5_star.png

How do you know it is accurate, though?

It's not, but technical research does matter, and it's surprisingly often a "brown M&M" that shows a lack of quality elsewhere.

I always underestimate the effectiveness of a weapon's stated capability in my stories, as tech rarely performs that well the first time, or any time.

Most projectiles miss the target they are aimed at. And bombs hit the wrong people, even precision-guided ones. Or don't go off at all, like Black Buck 2.
 
I always underestimate the effectiveness of a weapon's stated capability in my stories, as tech rarely performs that well the first time, or any time.
Michael Farmer's Tin Soldiers (great underrated tank novel by a tank veteran) has a great and accurate line where the protagonist adjusts the claimed kill rate of any airstrike down by one full echelon (ie if they say they hit a company equivalent, it's a platoon at best, etc...) whenever he hears of one.

But I'm not talking detailed stuff like that. I'm talking stuff like this, from Ian Slater's Battle Front:

reactive appliqué armor, that is, armor made famous by the Israeli armored divisions who had stuck the slabs of steel-Kevlar-ceramic sandwiches on their tanks to prevent the Syrians’ Russian-made T-72 TOWS from penetrating.
 
It's not, but technical research does matter, and it's surprisingly often a "brown M&M" that shows a lack of quality elsewhere. For every work that overloads on technical details (and they do exist), you have a million "machine gun pistols with heat seeking bullets" and "M60 Abrams with mounted flamethrowers", and even worse.
Literally none of that matters.
 
I don’t think that level of detail is essential to any work of historical/alternate fiction, although it can be a useful addition if adding to the immersion of a narrative (The depth of Hilary Mantel’s research being the classic example of this). But examples like the above just seem like minutiae for the sake of minutiae rather than adding anything to narrative or characterisation. In any case I don’t think ChatGPT is anywhere near good enough for research and there’s no attribution of what it produces.
 
In any case I don’t think ChatGPT is anywhere near good enough for research and there’s no attribution of what it produces.
I've been asking it technical questions that can't be answered just by citing stats like "When did aircraft gain the ability to reliably destroy tanks directly?"

(Answer: With the invention of guided weapons, but it said WW2).

I've not been the most impressed.
 
I can only assume that this is picking up on socialism = all things (some of the) US right hates?
It has its own definition and it just goes with it. I asked it about Kautsky's Agrarian Question and it just spouted what it knew about socialism and agriculture in the Soviet Union. It is not very good at anything not easily replicable from its model and will just make up stuff when it can't do that.
 
How do you know it is accurate, though?



I always underestimate the effectiveness of a weapon's stated capability in my stories, as tech rarely performs that well the first time, or any time.

Most projectiles miss the target they are aimed at. And bombs hit the wrong people, even precision-guided ones. Or don't go off at all, like Black Buck 2.


I agree, the accuracy is always going to be questionable, and that's what I mean about becoming overly reliant on it. But like I mentioned, When I run it through the Beta readers, if they don't pick it up, I'll call it a win.

-HC
 
I still don't get why. If you're handing over a degree of creativity it defeats the point. Even if you're editing the results yourself I'd rather just write it all myself. I'd rather do the research myself given how flawed it can be. It seems to be favouring productivity for productivity sake over fun. That's what this should be, fun.
 
I still don't get why. If you're handing over a degree of creativity it defeats the point. Even if you're editing the results yourself I'd rather just write it all myself. I'd rather do the research myself given how flawed it can be. It seems to be favouring productivity for productivity sake over fun. That's what this should be, fun.
Its All so wild.

Research and writing are challenging, they're not impossible. No. Matter how many times people who lack any creativity try to make it out to be.
 
Last edited:
I find the prospect of disengaging with writing to the extent of just posting the results of a GPT to be a profoundly miserable prospect, and I get especially sad seeing this being suggested as a mode for the future on a writing forum

Media creation hasn't been a bottleneck for a long time. Editing and curation is where the focus is now.

If there are some people who get some sort of enjoyment out of reading AI generated stories, then that is all the reason I really need to justify it being there. Though I anticipate that there will always be a market for non-AI generated stories, simply because some people will feel that it isn't real if there isn't some human behind it all.

Though I am amused by the question of how such people would engage with fiction if we reached such levels of computational sophistication that it was impossible for even the most astute of literary critics to tell whether something was written by a human or a computer. Let's say you gave them a piece of fiction and wouldn't tell them whether it was written by an AI. Would they then be unable to enjoy it or engage with it until they had proof that it was actually written by a human? If you told them it was by an AI would they then instinctively find it to be empty and lacking in meaning even if it was written by a human? And conversely, would they find it meaningful, even engaging, if you told them it had been written by a human even if it was written by an AI?

But in answer to your final question, I'd say that doing TLs and creative writing is something we do pretty much purely for own entertainment and amusement.

The art world has had that issue ever since the creation of the camera. There are even digital cameras that allow almost limitless opportunities to take photographs. The key is know what to take pictures of and that is what distinguishes someone taking photographs from an artistic photographer.

From my side, ChatGTP has been the best research tool I've ever used. Since I write WarPorn alt history, getting the technical details right is just as important as the historical details. So it's AMAZING to just ask a question and get an answer. I do worry that I'm going to over-rely on ChatGTP, and get some errors or anachronisms, but that's why I have Beta readers. Here's a case that would have taken me a bit of time to run down the air intercept range of an F-106.

-HC

latest_5_star.png

The maximum radar range isn't going to matter that much for the F-106.

What that isn't telling you is that the F-106 was designed to operate under ground control intercept so the radar range usually wouldn't be a factor. Aircraft of that era also often operated under rules of engagement requiring visual target confirmation because of the difficulty of identifying targets. Lastly, the missiles used by the F-106 were rather short ranged and not particularly effective, especially against anything smaller or more maneuverable than a subsonic bomber.

I don’t think that level of detail is essential to any work of historical/alternate fiction, although it can be a useful addition if adding to the immersion of a narrative (The depth of Hilary Mantel’s research being the classic example of this). But examples like the above just seem like minutiae for the sake of minutiae rather than adding anything to narrative or characterisation. In any case I don’t think ChatGPT is anywhere near good enough for research and there’s no attribution of what it produces.

That's not really minutiae, that's easily researched information that is off to such a comical degree. It would be like having cowboys riding around on horses with semi-automatic Colt .45 pistols. Cowboys used the Colt .45, but it was a revolver and the most popular one was single action and had to be cocked every shot.
 
cowboys riding around on horses with semi-automatic Colt .45 pistols

The slightly underwhelming sequel to Guns of the South.

"I've travelled back and given cowboys M1911s"

"thats only about 40 years more advanced, why did you do it?"

"They can hold them sideways and look dope as fuck"
 
Back
Top