Freelance Traveller

Title#78: AI Redux (without the help of any AI)

This article originally appeared in the November/December 2025 issue.

Following my last look at using AI, things like dictionary, grammar, thesaurus replacement being ok and brainstorming probably ok but taking the fun out of doing it yourself, I’ve been thinking about other possibilities.

I tested ChatGPT’s proofreading ability with an article I’d written. The results were fairly impressive. Firstly, I was surprised I could feed it 1200 words as a prompt, I thought the limit would be smaller. I probably should have tried comparing the results with Notebook LM which can ingest PDFs wholesale and comment on them. Secondly, ChatGPT picked up a typo I’m pretty sure I’d have caught on my final ‘print out read through’ (I still maintain you can’t proofread as effectively on screen). It offered some grammar corrections which were a mixed bag. Then it went on to suggest that I added a few more subheadings than I had. It offered ideas, some of which I liked, some I didn’t but I didn’t want more subheadings than I already had, so I ignored them. What I really liked was its encouragement about the entire piece and that it didn’t have any major reservations or corrections. That helped me through that “it’s all rubbish” phase of any writing project.

One reader of Confession #76 wrote to tell me of a use he’d made of generative AI which I thought was interesting. He’d fed Notebook LM all the PDFs of Deepnight Revelation and asked for summaries. Given how extensive that work is, I thought it a great use of gen AI. Note that Notebook LM keeps all your content within your account so like my correspondent, I wouldn’t have a problem with this as regards copyright worries. Being able to unpick the complexity of several volumes and question it and find content more easily seems a positive support tool. It’s just a pity the amount of material you can feed it is limited, or I’d be tempted to upload all of Traveller! Perhaps I could get it to generate my bibliography entries for me – though I’d like to think I add quite a lot of value in knowing what to include in the summaries and in making the connections you find in the comments.

Which brings me to ‘background research’ on Traveller’s general approach to topics. ChatGPT at least does seem to be familiar with large chunks of various Traveller rule sets and the Third Imperium setting. I’ve not tested it further yet but it can help locate material.  I think this usage is OK and not very different to using the Traveller Wiki.

Summarizing others’ work (TML emails, social media posts, published work) to feed into your own article? Good question. I’d advise students doing similar to check with their tutor if they’re happy with this (they mark the work, not the Library). In the Traveller context I would suggest this is OK but should be mentioned if not cited. I believe I’ve always done this (and thanked) TMLers etc if they’ve fed into anything I’ve produced.

I’ll freely admit the whole thing is a pandora’s box I’d happily firmly shut and bury very deeply – certainly in the academic world. It will be interesting to see how it plays out in the hobby world and at what point I first buy an expensive RPG book that I become certain has been written entirely by generative AI.

POSTSCRIPT: I’m aware of Mongoose’s recent announcement prohibiting all AI use in any Travellers’ Aid Society material. At least that makes the decision easy. Though I wonder if it’s tantamount to Ned Ludd’s followers throwing clogs into the machinery.Yes, I’m aware I’m conflating “Luddism” and the origins of “sabotage”