How AI can enhance your wargaming

[@JohnMichaelson]

If you want NotebookLM's silly podcast to be more professional sounding, you can add the prompts like "Avoid filler words like 'uh', 'wow', 'interesting', etc." and "Present in a professional manner at a college reading level" and it will eliminate a lot of the stupid cutesy back and forths. It's something we've struggled with in our test cases trying to sell it to the higher-ups, but they are amazed at the accuracy it digests material with minimal guidance.

[@RvTWargames]

THAT is a great tip. Thank you, John. That's a slide and hat tip in whenever v3 of what seems to becoming a series comes out! I'm sure Google will give us more voices and more tones of voice soon-ish, but for now I'll definitately be using this from now on.

[@alanshackelford6450]

Excellent.

[@alanshackelford6450]

So let me get this straight . . . somebody, somewhere decided having it speak like dolt was added value?

[@RvTWargames]

It's probably quite culturally specific.

[@JohnMichaelson]

@@alanshackelford6450 In a world where everything is "super" and "incredible" its kind of excessive hyperbole about everything seems like par for the course. I wonder if they fed it a bunch of existing podcasts to train its initial style from, or if that's how the dev team actually discuss things amongst themselves? 🤔

[@Dng5660]

Great! The ia for wargames...wat a new world of possibilities. Very interesting video. Thanks!

[@RvTWargames]

Yes, I feel there are a lot more uses, if only I could put my finger on them.

[@Simplified-z1k]

Absolutely taken on Notebook LM for RTW3 with a stack of Naval Books -- game changer in terms of speeding up my learning curve to some degree... Great Tip and Tankyou V much.. Awesome... Got notebooks setup on all things Naval :) WW1 WW2 RTW Naval in general Jutland etc etc..

[@RvTWargames]

@@Simplified-z1k once you start it just runs away with itself. I've converted all of my Kindle books into PDFs so that I can add them to NotebookLM.

[@kiwinesss]

Such a great idea, especially for a forgetful old (63 years old) chap as myself. I will definitely be including this into my gaming protocols. Thanks!

[@RvTWargames]

Exactly. I and my 65yo brain wouldn't be without it!

[@kiwinesss]

@@RvTWargames Nice to meet you, it is not often I find 60+ gamers, we are a rare breed.

[@man_vs_life]

This is really great content. Please keep it going with it (I know you will). 👍

[@RvTWargames]

Enjoying ploughing this exciting groove.

[@johnwlawton8833]

Outstanding! Thank you

[@RvTWargames]

😁 You're very welcome

[@SimpleWineChannel]

Hi, Dickie! I asked NotebookLM a question concerning an opportunity to situational retreat a unit in PD or key terrain in BCS wargame ("If a unit in key terrain and the combat result is situational retreat, may it chose to retreat 3 hexes instead of losing 1 step? I can't find it in the rules") and got this answer: --- ## Choice Between Retreat and Step Loss in Key Terrain You are correct that a unit in Key Terrain facing a Situational Retreat result can choose to **either retreat 3 hexes or take a step loss**. The rules explicitly state this option: > Given that 5.5c does not apply, units with this Combat Table result will **Retreat 3 hexes** and flip to their Move-side, or (if in **Key Terrain** or Prepared Defense) make no Retreat but **each unit in the Defender’s stack must take one Step Loss**. Therefore, the unit is not forced to lose a step and can instead choose to retreat 3 hexes. This rule highlights the tactical decision-making involved in *BCS*, allowing players to weigh the value of holding key terrain against the potential cost of unit attrition. --- Do you agree with it?

[@RvTWargames]

Obviously this is wrong. Unfortunately the AI is being hit with a double difficulty. Firstly, the rules in 5.5d say "or" when they really mean "unless". It not a well expressed rule. Secondly, the word 'or' has two meanings and it is not always obvious which is being meant. The more common is two connected choices: I could reply to you today or tomorrow. The second is two exclusive choices (known as XOR in programming): either I'm dead or alive. The rules are using an exclusive or: it is a situational retreat with a retreat or it is a key terrain with a step loss. But the AI is reading it (as might many humans) as a connected or: you can do this or that, your choice. Interesting example, thanks.

[@MichaelKarlosky]

Can AI be used to serve as a double blind umpire for wargames? For those unfamiliar, double blind is a play format where each player plays their own copy of the game and a third person (the umpire) determines which of the opposing units a player is aware of due to line of sight or some other proximity metric. Nearly all wargames automatically grant extreme omnipotence to both players at all times. Being able to play double blind without burdening some unfortunate chap with the job of umpiring would dramatically change how you’d play and would be more aligned with real world military doctrine. Reconnaissance units would finally find their proper place!

[@RvTWargames]

That is a great question and a really interesting idea. Naturally, I asked AI (Perplexity in this case) Double-blind gameplay in tabletop wargames is a method used to simulate the fog of war and limited intelligence that real military commanders face. Here's an analysis of how it works, the umpire's role, and how AI could potentially serve as an umpire: Double-Blind Gameplay In a double-blind wargame: - Players cannot see each other's units or movements directly on the game board - Each player maintains their own private map of unit positions - An umpire mediates interactions and provides limited information to players This system creates uncertainty and forces players to make decisions based on incomplete information, much like real military operations. Umpire's Role and Information Needs The umpire in a double-blind game requires: - Complete knowledge of both players' unit positions and movements - Understanding of game rules for combat, movement, and reconnaissance - Ability to determine when units come into contact or sight of each other - System for relaying limited information to players based on their actions The umpire must process player orders, update the master game state, and provide appropriate feedback to each player without revealing too much. AI as Umpire A generative AI tool could potentially assume the role of umpire in a double-blind wargame by: - Maintaining the complete game state digitally - Processing natural language orders from players - Applying game rules to resolve movements, combat, and reconnaissance - Generating appropriate, limited responses to each player The AI would need to be trained on: - The specific game rules and mechanics - Natural language processing for interpreting player orders - Principles of fog of war and limited intelligence in military operations Challenges for AI umpires include: - Ensuring consistent and fair application of rules - Generating realistic and appropriately limited responses - Handling edge cases and unexpected player actions While an AI umpire could potentially streamline the double-blind process, it would need to be carefully designed and tested to ensure it maintains the spirit and challenge of limited intelligence gameplay.

[@man_vs_life]

Pretty much the Holy Grail, along with having an AI opponent, especially where you can request a certain style of gameplay (aggressive, cautious or whatever). Surely this will be possible sometime in the future. Hope so. 🤞

[@war_gamer]

I wonder if we could up load maps and have LOS checkers! When thats a thing it would be neat.

[@RvTWargames]

It would, but sadly not yet. Generative AI is built on large language models (LLMs) - so words, quzillions of them. Maps are pictures with very specific encoded messages in them (terrain and counters). Some more visually orientated AI can recognise patterns in the pixels, say a face, but only after a great deal of machine learning training. And even then, it doesn't really 'understand' faces, it only recognises them. Looking at a wargaming map as we do, we absorb a huge amount of visual information because we've been trained on it (via reading the rules) because we can see things (actually we are very very good at seeing). LLMs can't do that. Still, one day!

[@MaitresChezNous21]

Great video ! From you experience in reading an OCS setup, do you think the AI can "understand", "see", the hex grid and the position of the units on it ? And provide in a second step a view of concentration of forces ? Do you think we may ask AI to add an overlay of a real map to the OCS hex grid and say whether Stalingrad is well defended ? I'm pretty sure we should be able to "read" an OCS map, to decide by pattern recognition the terrain of each hex (and hexsides, and paths) and may be should we be able to determine the strength or weakness of some position...

[@RvTWargames]

What a great question. So off the cuff I think not, currently. ChatGPT, Claude, Perplexity, Gemini (which notebooklm uses) and all the rest are large language models - so they deal in words. So they cannot 'see' a map, for example. Words have specific meanings (if sometimes vague or multiple meanings). Pixels from an image are much harder to understand. Some AI can recognise patterns, like a face, but they don't 'understand' the face and what the various bits are for. So equally, it wouldn't understand terrain. Theoretically you could explain the hex grid to the AI, give it so many rows and columns, and explain the he numbering system. I think it would understand that. But then you'd have to explain in words that hex A.01 was open terrain and hex A.02 was woods etc. Which would be beyond tedious! So sadly no. Or not yet!

[@mhorohello]

@@MaitresChezNous21 Some ai models can take images as inputs so it could be worth experimenting.

[@RvTWargames]

I've found that they can certainly take structured text like a process flow and turn it into mermaid code and put that into apps like Draw.io or Mermaid.live to turn into diagrams.

[@codyashworth7470]

Where did you go?

[@fredschwarz2376]

Thank you for this video. I am trying to make a source document for BGG feed using Feedly like you show, however, I can only get it to show threads going back about 20 days and not there entire three years' worth. I select "oldest" as you show. What might I be doing wrong?

[@fredschwarz2376]

I notice when I am on the BGG forum page and click the RSS button, the result I get only shows the last 19 days or so.

[@RvTWargames]

It's a limit many organisations put on their RSS feeds: they only allow feeds from now (or recently) going forward. They don't allow the RSS feed to go far back in time. Which is frustrating for us, as hand copying old messages is a massive bore. Obviously this means this technique works best for new games, or games that have had a recent errata published.

[@fredschwarz2376]

@@RvTWargames OK, thanks.

[@johnclifford9853]

Leave no Stone unturned! 😂

[@RvTWargames]

We'll have a proper chat about it next week