Exactly. Uses keep on popping up. Today we were thinking of a house rule and so I asked my notebook of 23 sources if why the Rule As Written was as it was, and it answered that it wasn't explained, but then went in to offer three plausible reasons while the RAW might be designed that way, which prompted a much richer and more rounded discussion around the proposed house rule (and we came up with a fourth possible reason as a consequence). Uses keep popping up!
@@RvTWargames I missed the first one (work and immediate vacation) and the notification on the second pushed me to what I missed! I may jump into TCS as it’s an area I’ve dabbled in often and Lee‘s videos are helpful but there’s a lot of meat on the bone!
Really enjoyed you discussing the practical application of technology to understanding problems. The wargaming application is just a bonus! I'd love to hear more about how you apply technology like this or other less cutting-edge tools in your gaming. I've enjoyed the RtW episodes when you break out diagramming tools, spreadsheets, and maps.
Thanks Sabre. I'm musing about a follow up video to show some new use cases, and new features that have recently been released, and how to combine it with other AI tools to create helpful results.
Damn, that is impressive. This would be so good for complex role playing games like Shadowrun, for example, where you're trying to search through half a dozen books just to draw up a character.
Excellent stuff, many thanks. I'll definitely give this a try. Like you, I write the occasional book review. Not for much longer, I suspect. Thanks again, Andrew
The golden key here seems to be supplementing rules, charts etc with forum Q&A. I've been experimenting with a couple of games and have parsed all relevant forum posts from "a popular games forum" and the results are proving _very_ promising, tapping in on all the ambiguities, edge cases etc of rules that have been expressed and clarified. Notebook LM is truly amazing.
Interesting. How did you manage to parse "a popular games forum" given its epic inaccessibility and the fact it doesn't even work properly on a phone! I assume that the likes of Consimworld are all chopped up into many separate pages and FB won't let you read private group pages without being logged into FB.
@@RvTWargames Nothing fancy. Copy n pasting into a text file and running a parsing script on it to remove all the irrelevant stuff and to format it in a way that that makes it clear where a thread starts. Only on a couple of fairly new games at this stage so not too much work. Having to face the bigger ones at some stage soon though. That'll be a lot of work but the potential is pretty juicy.
Once you have extracted and loaded the forum discussions into the tool, can you ask a question to only get the game designer clarifications related to a specific topic?
Yes if: 1. You can extract the discussion (love to know how you'd do that) 2. You know the names of who give authoritative opinions on rules questions - then you can give that name(s) in the question (prompt) you ask. 3. You might try creating a prompt around terms like errata, rules question, clarification. 4. If it's a game series you might want to limit it to the name of particular games, so you don't accidentally pick up game specific corrections.
I've been experimenting with this idea for a few weeks and was of course very excited to see your video! I tried using ChatGPT's "my GPTs", but now that I've seen what NotebookLM can do this is clearly superior. A great use case is selecting series rules and game specific rules and then asking for contradictions to find out where the GSRs overrule the series rules. The LLM is very good at identifying these differences and it "understands" why these contradictions occur. Tested with GTS 2.0c and GTS Utah.
Nice to see I'm not the only one tinkering with this stuff. Love the series rules vs GSR contradictions use case! I'll give it a go on my GTS NotebookLM as I'm playing Utah in a few weeks.
Glad that the "podcast" was breaking down how war is always political at 11:00. Lots of young wargamers absolutely refuse to think critically about what they're representing and will tell you with a straight face that wargames arent political.
Been using notebooklm for 1914 twilight in the East. One 50 pages rulebook, 3 playbooks. Want to check something without going through 4 books or a long index? Just ask and you get a comprehensive answer with sources from the 4 books. Absolutely excellent.
@@Nico165 great to hear. I have that on my shelf of shame of want to play but not played yet! Are there any other sources you can add? Quality UA-cam videos? Files on BGG such as player aids and AARs? Or are you just happy with the core documents?
@@RvTWargames I started with the official docs. There are new charts only in image format and my first attempt to load them was not successful. Need to try again. I discovered in your video that UA-cam was a possibility and will try it. Another idea I have is adding other historical sources (starting with Wikipedia pages about the battles involved to make it simple at first) and see if I can have interesting answers to questions like "how can you do this historical strategy with the rules from this game?"
Excellent. You might try something a bit more heavy weight. Norman Stone wrote one of the first depth coverages of the Eastern Front in WW1. It's been superceded by another chap coincidentally also called Stone (what are the chances) - unrelated. If you can find some things about them or interviews that could be extra heft to your wiki articles. You might also give Google Scholar and ResearchGate a try and see if there are academic articles that would fit. Good luck with that
@@RvTWargames thanks for the recommendation ! I went all in, bought the ebook version and uploaded the pdf on my notebooklm twilight in the east space. Below is my first attempt. Things got a bit mixed up between different sub answers but this looks promising. Similarities Between the Twilight in the East Supply Rules and the Historical Situation Described by Norman Stone Here's a comparison of the supply rules in Twilight in the East with the historical situation described in Norman Stone's book, The Eastern Front 1914-1917: ● The importance of railroads: Both the Twilight in the East rules and Stone's book emphasize the critical role of railroads in supplying armies during World War I. Stone describes how pre-war arrangements failed to anticipate a protracted conflict, leading to a severe shortage of munitions in the Russian army. The reliance on foreign suppliers proved unreliable due to their own wartime commitments and production issues, highlighting the inadequacy of Russia's existing infrastructure for sustained warfare. In the game, this dependence on railroads is reflected in the rule that all supply, except for fortress supply, originates at railroad hexes. ● Limited range of supply: Stone points out that the Russian army's supply range from a railway was estimated to be around 120 kilometers, largely due to their modest requirements and disregard for their horses. Similarly, in the game, a unit's supply range is restricted by the capabilities of depot and Corps Train units. Stone emphasizes that exceeding this range could cripple an army's offensive capabilities. This is consistent with the game's mechanics, where units operating beyond their supply range suffer penalties. For instance, units in low supply or out of supply face negative consequences like reduced combat effectiveness and potential surrender. ● Challenges of logistics and transport: Stone highlights the many logistical and transport challenges faced by the Russian army, particularly during the early years of the war. The inadequate rail infrastructure, the chaotic state of foreign trade, and the conscription of skilled labor all contributed to delays and shortages. These challenges are mirrored in Twilight in the East through rules like the limited capacity of depot units and the vulnerability of Corps Train units to enemy attacks. For example, if a Corps Train unit is forced to retreat, it must retreat three hexes and is flipped to its Extended Mode side, potentially disrupting the flow of supplies to front-line units. ● Impact of supply shortages on military operations: Stone explains how the lack of ammunition and supplies severely hampered Russian offensives and contributed to their defeats in 1915. Commanders often attributed their failures to shell shortages, even when other factors like disorganization and poor leadership played a significant role. Twilight in the East replicates this by imposing negative consequences on units in low supply or out of supply, representing the detrimental effects of inadequate logistics on combat effectiveness. For instance, units out of supply have their combat effectiveness reduced and risk surrender if both "Demoralized" and "Isolated". ● Fortress supply: The game's rule about fortress supply reflects the historical reality that fortified positions often served as logistical hubs and provided a more reliable source of supplies for nearby units compared to those relying on extended supply lines. By incorporating these historical aspects of supply into the game, Twilight in the East provides players with a deeper understanding of the logistical challenges and constraints that shaped military operations on the Eastern Front. The game's mechanics accurately portray how critical a steady flow of ammunition and supplies was to the success of any military operation during World War I.
@@Nico165 WOWZERS! You've got the bones of a video script right there! Stick in a few historical photos and map screen shots from Vassal with the odd arrow drawn on it to show pertinent things and BOOM! A cool 20 minute or so video discussing how the game apes history in key features. If you reverse the question - how does game supply and history differ, you get a bonus critique of the game too. I do believe a new use case is born!!!
Many thanks for investigating this complex and not well understood from a (more critical) view than others. I have started experimenting with this type of LM for answering rules board game questions; starting with games that I know well. I still do not trust these tools to not make mistakes!
Well, I tried NotebookLM with my fave game of Unhappy King Charles by GMT. It also has a set of rules that have held up to repeated scrutiny over the years. The rules are very complete but are often hard to located. I uploaded the rules and the playbook PDFs and started to ask real world game rules type questions. "How many movement points does an army with 4 brigades receive"; that worked well. However, when I asked the question can the King disperse before battle, in a variety of different ways, it repeatedly said the King could not disperse. That is very incorrect. the citations listed did not point to the area of the documents that caused such a ruling. I hazard it may be due to the examples of play in the rules and these being taken as rules rather than an example. So, points for effort but not a usable tool yet.
I'm unclear why the examples of play would contradict the rules? Is there a confusion of language here? We still need to know enough to know when the AI quotes sources that are being misleading. So it did its job as well as it could given the clarity of the sources and gave you the tools to spot that. We should never assume that AI will always be right. Knowing stuff is a tricky contextual interpretive thing. You should never trust them not to make mistakes. We have to keep responsibility for using our judgement. At least until our Robot Overlords take over.
@@RvTWargames What confuses/worries me if how did it get it so wrong? I have been throwing simple, yet real world, questions at the LM and it has been (overall) rather impressive in summarizing concepts and listing exceptions. But for dispersal it just gets it wrong in a game-breaking way!
This is so useful. I started looking at using ChatGPT and Perplexity a few weeks ago for rules outlines, summaries of rules etc but on my controls (rules I already knew) there were errors that I put down to use of multiple/wrong sources. This solves that - and with a lot of output types. Full steam ahead with NotebookLM it is :) Thanks
It is. You can't use AI if you don't know what it's talking about, but citing it's sources means you can bring your knowledge and expertise to assess what it's saying and have confidence it's correct, or identity when the rules are fuzzy or contradictory. Great stuff.
Great idea and great tools for this great warfare series. In planning a multiplayer game of ocs where player would be CinC, Army groups and Armies commanders and the referee would be all the Corps Commander, moving the pieces on the board. Players write orders each turn and have access to a Vassal situation and reports. I plan to use IA to translate in some military form the written orders and the reports…
Ambitious! Love to hear how that goes. I helped with a four aside Enemy at the Gates, with an Army Group commander controlling the direction, supply and reinforcements and the air force, and three army commanders. Worked really well.
The real magic will happen when these tools are powerful enough to digest the whole forum (or fora....) devoted to a game. Every serious game has an online forum or three with a vast compendium of rule interpretations, AARs and tactic/strategy musing ... I never have time to read the whole forum for any game so bring on the AI summarizer and miner... I think I'd pay real money to use a tool like that!
I'm very interested in this topic. In fact, I'd like to do some Comp Sci. research on how rules could be rewritten for clarity, or even implemented in a VASSAL like setting using AI and Rules->DSLs->Game.
If the game has more than maybe 30 pages of rules that are complex, make it a computer game. The board wargames I enjoy the most have 20 pages of rules or less. For me its about enjoying the play, not combing rulebooks.
Well then you are an ideal candidate for a spot of AI help to cut down the cognitive load of absorbing the bigger games. While there are plenty of great games with shorter rule books, war, being inherently complex, usually needs bigger rules to do justice to combined arms, air and perhaps naval integration, supply, politics and anything else the designer thinks it's important.
Very interesting. I would hope that if you ask a Question, that cannot be answered by within Source documents, that it clearly states it has no Qualified answer 9from the source ? And so giving confidence that it does not attempt any hallucinations outside the source material.
If it is completely outside what's in the source documents it is pretty good at saying so. If the language of the documents is vague, then it will usually qualify its answer. And if it does find something, but thinks its for a higher reason it will usually try and explain that difference by referring to that reason. Of course, in places where a human might get it wrong, it might get it wrong too. But it will usually will have used qualifying words when it does so.
Hmmmm, interesting ideas. So, if you took a long video of yourself playing, with a mic to record your commentary. That would include talking through the dice rolls (it only understands text, and your voice would be transcribed into text). Then it could certainly summarise that into the skeleton of an AAR. You'd need to edit it and correct it, but it would distill hours of play into something manageable. (And if the first draft is too long, ask it to summarise it further). Probably best to chop up the video into hour long pieces. For the battle plan. If there are good design notes plus any AARs or playthroughs from other people (either articles or videos), then asking it to suggest a battle plan would be an interesting exercise. Alternatively, you could add PDFs of a history book or two, and ask it for guidance on how to plan your play based on history and the rules of the game. These are advanced uses, so the results might be interesting or rubbish. Probably best if you use a logic tree prompt to help guide it's thinking (see the next video).
@@RvTWargames Thank you very much for your answer and all that work. I did a first experiment with Chat GPT yesterday. It was able to create feasibles ideas for the engagement for both sides of a cold war gone hot scenario, based on the tactical doctrines of the time. Nevertheless it lied to me when I asked for a specific scenario in the game - the answers where just too general. In the end it is as you say in the next video: the quality of the answers depends on the formulation of the prompts. As soon as I help myself to a tablet, I will go on and try out these ideas.
Great question. I think the short answer is none. War in the East is too prone to hyper micro management. Hearts of Iron is too sand boxy, allowing massive diversions from historical reality. And both are at the wrong scale. Matrix Games has some candidates: Panzer Corps 2, Order of Battle, Decisive Campaigns. But for me, they all miss the mark. Their integration of different arms doing different things (infantry, armour, artillery, air) is poor. The battle play is very attritional. Supply, surprise, and reinforcements are all quite mechanistic. Don't get me wrong. They are fun. But they are just not in the same class of sophistication as OCS or BCS or GTS systems.
@@RommelsAsparagus not really. The RR rail cap question isn't clear in the rules. The fault is creating a clarification on Consimworld, which many don't it can't use, and not reflecting that back in the consolidated errata, which is also missing the last couple of games.
Good question. Certainly yes if: 1. The chart PDF is in text, rather than a picture of a chart 2. The chart content is in English, rather than in code, unless the code has a clear key. If the chart PDF is an image, it can use OCR to read the chart. The more English like it is, the more chance of success. If the chart is coded, say like a combat results table, the more the AI will struggle as will not be trained on the code. Likewise, it will struggle with scenario listing as it won't necessarily understand the map hex references nor the abbreviated unit references.
Just checked GTS rules. And it fails on pretty simple questions: Can independent unit during Formation activation not In Command make Rally action? His answer is YES (it's wrong). And: can artillery in Artillery Park make Opportunity Fire? His answer is YES (it's wrong, only Emplaced Art can do this). He even suggests quotes and they are contradictory.
notebooklm answer: In Conclusion: Even though an independent unit is not In Command during Formation Activation, it can still attempt to Rally. However, it must pass a Troop Quality Check, with its Troop Quality Rating reduced by one, for the Rally to succeed.
notebooklm answer: Artillery units with orange, brown, or black Fire Ratings located within an Artillery Park can perform Opportunity Fire. Example: If an enemy unit enters a hex within three hexes of a friendly Artillery Park, triggering an Opportunity Fire situation, the artillery unit within the park can attempt to perform Opportunity Fire using its Direct Fire capabilities. In summary, while primarily deployed for Indirect Fire support, artillery units in Artillery Parks maintain their ability to respond to immediate threats through Opportunity Fire when enemy units approach within their Direct Fire range.
I've created a GTS notebook with the official rules, the unofficial rules (rewritten for clarity - available on BGG), the game specific rules (exclusive in GTS-speak), and the transcripts of tutorial videos from honcho Lee Forester and the usually solid Nathan (Wise Guy History). Can an independent unit during Formation activation that is not In Command conduct a Rally action? An Independent Unit that is not In Command cannot make a Rally action during a Formation activation. Here's why: ● Independent Unit Activation: Independent Units (those with white or black stripes) can only be activated during a Division Activation or the Formation Activation of the Formation to which they are attached. ● Command Requirement for Rally: Units that are Out of Command must pass a Troop Quality Check to Rally successfully. ● Out of Command Status: Units are considered Out of Command when they are outside their Leader's Command Range. This applies to Independent Units as well. ● Command During Formation Activation: During a Formation Activation, Independent Units are considered In Command only if they are within the Command Range of the Formation's Leader. Therefore, if an Independent Unit is not within the Command Range of the Formation's Leader during a Formation Activation, it is considered Out of Command and cannot perform a Rally action.
@@RvTWargames it is inconsistency that makes it not very usable. If it makes mistakes (using oficcial rules, and they are clear, what the answer should be) - I do not think we can rely on it. Yet.
Aha. So my quip wasn't completely off target 🙄 Not to mention that it is the only PDF you are expected to purchase. As if ASLers haven't paid enough! And, I suspect they are copy protected so that would fail in NotebookLM too. I stick to ASLSK myself. Really should sell my ASLs. My remaining life on this planet is too short to learn them now!
Great video! Lots of food for thought. You’ve prompted me to go explore a number of things… Thanks!
@@jjflash30 you're very welcome. I'd be interested to hear of any success you have.
Thank you so much for this! A wormhole of all wormholes for gaming study! Seriously, this looks incredibly intriguing!
Exactly. Uses keep on popping up.
Today we were thinking of a house rule and so I asked my notebook of 23 sources if why the Rule As Written was as it was, and it answered that it wasn't explained, but then went in to offer three plausible reasons while the RAW might be designed that way, which prompted a much richer and more rounded discussion around the proposed house rule (and we came up with a fourth possible reason as a consequence).
Uses keep popping up!
And if you haven't already, see the second in this series.
@@RvTWargames I missed the first one (work and immediate vacation) and the notification on the second pushed me to what I missed! I may jump into TCS as it’s an area I’ve dabbled in often and Lee‘s videos are helpful but there’s a lot of meat on the bone!
That sounds an combo of rules and stuff plus high quality videos.
Really enjoyed you discussing the practical application of technology to understanding problems. The wargaming application is just a bonus!
I'd love to hear more about how you apply technology like this or other less cutting-edge tools in your gaming. I've enjoyed the RtW episodes when you break out diagramming tools, spreadsheets, and maps.
Thanks Sabre. I'm musing about a follow up video to show some new use cases, and new features that have recently been released, and how to combine it with other AI tools to create helpful results.
Sounds great, count me in!
Damn, that is impressive. This would be so good for complex role playing games like Shadowrun, for example, where you're trying to search through half a dozen books just to draw up a character.
YOU are NotebookLM's target market!
Excellent stuff, many thanks. I'll definitely give this a try. Like you, I write the occasional book review. Not for much longer, I suspect.
Thanks again,
Andrew
Really glad you found it interesting. The tools keep on improving.
The golden key here seems to be supplementing rules, charts etc with forum Q&A. I've been experimenting with a couple of games and have parsed all relevant forum posts from "a popular games forum" and the results are proving _very_ promising, tapping in on all the ambiguities, edge cases etc of rules that have been expressed and clarified. Notebook LM is truly amazing.
Interesting. How did you manage to parse "a popular games forum" given its epic inaccessibility and the fact it doesn't even work properly on a phone!
I assume that the likes of Consimworld are all chopped up into many separate pages and FB won't let you read private group pages without being logged into FB.
@@RvTWargames Nothing fancy. Copy n pasting into a text file and running a parsing script on it to remove all the irrelevant stuff and to format it in a way that that makes it clear where a thread starts. Only on a couple of fairly new games at this stage so not too much work. Having to face the bigger ones at some stage soon though. That'll be a lot of work but the potential is pretty juicy.
@@man_vs_life nice work
Once you have extracted and loaded the forum discussions into the tool, can you ask a question to only get the game designer clarifications related to a specific topic?
Yes if:
1. You can extract the discussion (love to know how you'd do that)
2. You know the names of who give authoritative opinions on rules questions - then you can give that name(s) in the question (prompt) you ask.
3. You might try creating a prompt around terms like errata, rules question, clarification.
4. If it's a game series you might want to limit it to the name of particular games, so you don't accidentally pick up game specific corrections.
Glad you shared your experiment w/ us. Both entertaining and informative.
😁
We all like teasing ASL
I've been experimenting with this idea for a few weeks and was of course very excited to see your video! I tried using ChatGPT's "my GPTs", but now that I've seen what NotebookLM can do this is clearly superior. A great use case is selecting series rules and game specific rules and then asking for contradictions to find out where the GSRs overrule the series rules. The LLM is very good at identifying these differences and it "understands" why these contradictions occur. Tested with GTS 2.0c and GTS Utah.
Nice to see I'm not the only one tinkering with this stuff.
Love the series rules vs GSR contradictions use case! I'll give it a go on my GTS NotebookLM as I'm playing Utah in a few weeks.
Glad that the "podcast" was breaking down how war is always political at 11:00. Lots of young wargamers absolutely refuse to think critically about what they're representing and will tell you with a straight face that wargames arent political.
Oh dear. War Studies 101.
Been using notebooklm for 1914 twilight in the East. One 50 pages rulebook, 3 playbooks. Want to check something without going through 4 books or a long index? Just ask and you get a comprehensive answer with sources from the 4 books. Absolutely excellent.
@@Nico165 great to hear. I have that on my shelf of shame of want to play but not played yet!
Are there any other sources you can add? Quality UA-cam videos? Files on BGG such as player aids and AARs? Or are you just happy with the core documents?
@@RvTWargames I started with the official docs. There are new charts only in image format and my first attempt to load them was not successful. Need to try again. I discovered in your video that UA-cam was a possibility and will try it. Another idea I have is adding other historical sources (starting with Wikipedia pages about the battles involved to make it simple at first) and see if I can have interesting answers to questions like "how can you do this historical strategy with the rules from this game?"
Excellent. You might try something a bit more heavy weight. Norman Stone wrote one of the first depth coverages of the Eastern Front in WW1. It's been superceded by another chap coincidentally also called Stone (what are the chances) - unrelated. If you can find some things about them or interviews that could be extra heft to your wiki articles. You might also give Google Scholar and ResearchGate a try and see if there are academic articles that would fit.
Good luck with that
@@RvTWargames thanks for the recommendation ! I went all in, bought the ebook version and uploaded the pdf on my notebooklm twilight in the east space.
Below is my first attempt. Things got a bit mixed up between different sub answers but this looks promising.
Similarities Between the Twilight in the East Supply Rules and the Historical Situation Described by Norman Stone
Here's a comparison of the supply rules in Twilight in the East with the historical situation described in Norman Stone's book, The Eastern Front 1914-1917:
●
The importance of railroads: Both the Twilight in the East rules and Stone's book emphasize the critical role of railroads in supplying armies during World War I. Stone describes how pre-war arrangements failed to anticipate a protracted conflict, leading to a severe shortage of munitions in the Russian army. The reliance on foreign suppliers proved unreliable due to their own wartime commitments and production issues, highlighting the inadequacy of Russia's existing infrastructure for sustained warfare. In the game, this dependence on railroads is reflected in the rule that all supply, except for fortress supply, originates at railroad hexes.
●
Limited range of supply: Stone points out that the Russian army's supply range from a railway was estimated to be around 120 kilometers, largely due to their modest requirements and disregard for their horses. Similarly, in the game, a unit's supply range is restricted by the capabilities of depot and Corps Train units. Stone emphasizes that exceeding this range could cripple an army's offensive capabilities. This is consistent with the game's mechanics, where units operating beyond their supply range suffer penalties. For instance, units in low supply or out of supply face negative consequences like reduced combat effectiveness and potential surrender.
●
Challenges of logistics and transport: Stone highlights the many logistical and transport challenges faced by the Russian army, particularly during the early years of the war. The inadequate rail infrastructure, the chaotic state of foreign trade, and the conscription of skilled labor all contributed to delays and shortages. These challenges are mirrored in Twilight in the East through rules like the limited capacity of depot units and the vulnerability of Corps Train units to enemy attacks. For example, if a Corps Train unit is forced to retreat, it must retreat three hexes and is flipped to its Extended Mode side, potentially disrupting the flow of supplies to front-line units.
●
Impact of supply shortages on military operations: Stone explains how the lack of ammunition and supplies severely hampered Russian offensives and contributed to their defeats in 1915. Commanders often attributed their failures to shell shortages, even when other factors like disorganization and poor leadership played a significant role. Twilight in the East replicates this by imposing negative consequences on units in low supply or out of supply, representing the detrimental effects of inadequate logistics on combat effectiveness. For instance, units out of supply have their combat effectiveness reduced and risk surrender if both "Demoralized" and "Isolated".
●
Fortress supply: The game's rule about fortress supply reflects the historical reality that fortified positions often served as logistical hubs and provided a more reliable source of supplies for nearby units compared to those relying on extended supply lines.
By incorporating these historical aspects of supply into the game, Twilight in the East provides players with a deeper understanding of the logistical challenges and constraints that shaped military operations on the Eastern Front. The game's mechanics accurately portray how critical a steady flow of ammunition and supplies was to the success of any military operation during World War I.
@@Nico165 WOWZERS! You've got the bones of a video script right there! Stick in a few historical photos and map screen shots from Vassal with the odd arrow drawn on it to show pertinent things and BOOM! A cool 20 minute or so video discussing how the game apes history in key features.
If you reverse the question - how does game supply and history differ, you get a bonus critique of the game too.
I do believe a new use case is born!!!
Many thanks for investigating this complex and not well understood from a (more critical) view than others. I have started experimenting with this type of LM for answering rules board game questions; starting with games that I know well. I still do not trust these tools to not make mistakes!
Well, I tried NotebookLM with my fave game of Unhappy King Charles by GMT. It also has a set of rules that have held up to repeated scrutiny over the years. The rules are very complete but are often hard to located. I uploaded the rules and the playbook PDFs and started to ask real world game rules type questions. "How many movement points does an army with 4 brigades receive"; that worked well. However, when I asked the question can the King disperse before battle, in a variety of different ways, it repeatedly said the King could not disperse. That is very incorrect. the citations listed did not point to the area of the documents that caused such a ruling. I hazard it may be due to the examples of play in the rules and these being taken as rules rather than an example. So, points for effort but not a usable tool yet.
I'm unclear why the examples of play would contradict the rules? Is there a confusion of language here?
We still need to know enough to know when the AI quotes sources that are being misleading. So it did its job as well as it could given the clarity of the sources and gave you the tools to spot that.
We should never assume that AI will always be right. Knowing stuff is a tricky contextual interpretive thing. You should never trust them not to make mistakes. We have to keep responsibility for using our judgement.
At least until our Robot Overlords take over.
@@RvTWargames What confuses/worries me if how did it get it so wrong? I have been throwing simple, yet real world, questions at the LM and it has been (overall) rather impressive in summarizing concepts and listing exceptions. But for dispersal it just gets it wrong in a game-breaking way!
Can you reverse engineer why but looking at the sources it is quoting?
This is so useful. I started looking at using ChatGPT and Perplexity a few weeks ago for rules outlines, summaries of rules etc but on my controls (rules I already knew) there were errors that I put down to use of multiple/wrong sources. This solves that - and with a lot of output types. Full steam ahead with NotebookLM it is :) Thanks
You're very welcome. Good to see I'm not the only one in the vanguard of using these tools.
@@PirateJock The fact that NobookLM can site chapters in its sources is truly a game changer.
It is. You can't use AI if you don't know what it's talking about, but citing it's sources means you can bring your knowledge and expertise to assess what it's saying and have confidence it's correct, or identity when the rules are fuzzy or contradictory. Great stuff.
Great idea and great tools for this great warfare series. In planning a multiplayer game of ocs where player would be CinC, Army groups and Armies commanders and the referee would be all the Corps Commander, moving the pieces on the board. Players write orders each turn and have access to a Vassal situation and reports.
I plan to use IA to translate in some military form the written orders and the reports…
Ambitious! Love to hear how that goes.
I helped with a four aside Enemy at the Gates, with an Army Group commander controlling the direction, supply and reinforcements and the air force, and three army commanders. Worked really well.
Awesome tool! Greatly appreciated.
For sure! Thanks for the appreciation 😊
Very Helpful Video. Thank you very much :)
You're welcome! It's a bit of a game changer for me, so I hope it is for others too.
The real magic will happen when these tools are powerful enough to digest the whole forum (or fora....) devoted to a game. Every serious game has an online forum or three with a vast compendium of rule interpretations, AARs and tactic/strategy musing ... I never have time to read the whole forum for any game so bring on the AI summarizer and miner... I think I'd pay real money to use a tool like that!
Very informative! Thank you for sharing!
Glad you found it interesting.
I'm very interested in this topic. In fact, I'd like to do some Comp Sci. research on how rules could be rewritten for clarity, or even implemented in a VASSAL like setting using AI and Rules->DSLs->Game.
I think that's the direction that things are going, especially if no-code tools like NotebookLM keep on developing so well.
Bravo this is really helpful thanks.
Happy to help. I knew it would click with certain people 😊
If the game has more than maybe 30 pages of rules that are complex, make it a computer game. The board wargames I enjoy the most have 20 pages of rules or less. For me its about enjoying the play, not combing rulebooks.
Well then you are an ideal candidate for a spot of AI help to cut down the cognitive load of absorbing the bigger games. While there are plenty of great games with shorter rule books, war, being inherently complex, usually needs bigger rules to do justice to combined arms, air and perhaps naval integration, supply, politics and anything else the designer thinks it's important.
Very interesting. I would hope that if you ask a Question, that cannot be answered by within Source documents, that it clearly states it has no Qualified answer 9from the source ? And so giving confidence that it does not attempt any hallucinations outside the source material.
If it is completely outside what's in the source documents it is pretty good at saying so. If the language of the documents is vague, then it will usually qualify its answer. And if it does find something, but thinks its for a higher reason it will usually try and explain that difference by referring to that reason.
Of course, in places where a human might get it wrong, it might get it wrong too. But it will usually will have used qualifying words when it does so.
Can NotebookLM use the rules and scenario descriptions to write a battle plan? Or even, if given crude description and dice roll results, AARs?
Hmmmm, interesting ideas.
So, if you took a long video of yourself playing, with a mic to record your commentary. That would include talking through the dice rolls (it only understands text, and your voice would be transcribed into text). Then it could certainly summarise that into the skeleton of an AAR. You'd need to edit it and correct it, but it would distill hours of play into something manageable. (And if the first draft is too long, ask it to summarise it further). Probably best to chop up the video into hour long pieces.
For the battle plan. If there are good design notes plus any AARs or playthroughs from other people (either articles or videos), then asking it to suggest a battle plan would be an interesting exercise.
Alternatively, you could add PDFs of a history book or two, and ask it for guidance on how to plan your play based on history and the rules of the game.
These are advanced uses, so the results might be interesting or rubbish. Probably best if you use a logic tree prompt to help guide it's thinking (see the next video).
@@RvTWargames Thank you very much for your answer and all that work.
I did a first experiment with Chat GPT yesterday. It was able to create feasibles ideas for the engagement for both sides of a cold war gone hot scenario, based on the tactical doctrines of the time.
Nevertheless it lied to me when I asked for a specific scenario in the game - the answers where just too general.
In the end it is as you say in the next video: the quality of the answers depends on the formulation of the prompts.
As soon as I help myself to a tablet, I will go on and try out these ideas.
I hear Xmas is coming, if a chap needs a swanky new Tablet.
What PC game do you think is most similar to OCS ?
or do you recommend to play who could not play OCS.
Great question. I think the short answer is none. War in the East is too prone to hyper micro management. Hearts of Iron is too sand boxy, allowing massive diversions from historical reality. And both are at the wrong scale. Matrix Games has some candidates: Panzer Corps 2, Order of Battle, Decisive Campaigns. But for me, they all miss the mark.
Their integration of different arms doing different things (infantry, armour, artillery, air) is poor. The battle play is very attritional. Supply, surprise, and reinforcements are all quite mechanistic.
Don't get me wrong. They are fun. But they are just not in the same class of sophistication as OCS or BCS or GTS systems.
I noticed you didn't load the errata for that one, and if Chip had put in in there, then I bet it would have answered correctly..
I put in the consolidated errata covering (nearly) everything.
@@RvTWargames Ah, so the AI goofed. Interesting.
@@RommelsAsparagus not really. The RR rail cap question isn't clear in the rules.
The fault is creating a clarification on Consimworld, which many don't it can't use, and not reflecting that back in the consolidated errata, which is also missing the last couple of games.
Can this AI handle charts at all?
Good question. Certainly yes if:
1. The chart PDF is in text, rather than a picture of a chart
2. The chart content is in English, rather than in code, unless the code has a clear key.
If the chart PDF is an image, it can use OCR to read the chart. The more English like it is, the more chance of success.
If the chart is coded, say like a combat results table, the more the AI will struggle as will not be trained on the code.
Likewise, it will struggle with scenario listing as it won't necessarily understand the map hex references nor the abbreviated unit references.
Just checked GTS rules. And it fails on pretty simple questions: Can independent unit during Formation activation not In Command make Rally action? His answer is YES (it's wrong). And: can artillery in Artillery Park make Opportunity Fire? His answer is YES (it's wrong, only Emplaced Art can do this). He even suggests quotes and they are contradictory.
notebooklm answer:
In Conclusion:
Even though an independent unit is not In Command during Formation Activation, it can still attempt to Rally. However, it must pass a Troop Quality Check, with its Troop Quality Rating reduced by one, for the Rally to succeed.
notebooklm answer:
Artillery units with orange, brown, or black Fire Ratings located within an Artillery Park can perform Opportunity Fire.
Example: If an enemy unit enters a hex within three hexes of a friendly Artillery Park, triggering an Opportunity Fire situation, the artillery unit within the park can attempt to perform Opportunity Fire using its Direct Fire capabilities.
In summary, while primarily deployed for Indirect Fire support, artillery units in Artillery Parks maintain their ability to respond to immediate threats through Opportunity Fire when enemy units approach within their Direct Fire range.
I've created a GTS notebook with the official rules, the unofficial rules (rewritten for clarity - available on BGG), the game specific rules (exclusive in GTS-speak), and the transcripts of tutorial videos from honcho Lee Forester and the usually solid Nathan (Wise Guy History).
Can an independent unit during Formation activation that is not In Command conduct a Rally action?
An Independent Unit that is not In Command cannot make a Rally action during a Formation activation. Here's why:
● Independent Unit Activation: Independent Units (those with white or black stripes) can only be activated during a Division Activation or the Formation Activation of the Formation to which they are attached.
● Command Requirement for Rally: Units that are Out of Command must pass a Troop Quality Check to Rally successfully.
● Out of Command Status: Units are considered Out of Command when they are outside their Leader's Command Range. This applies to Independent Units as well.
● Command During Formation Activation: During a Formation Activation, Independent Units are considered In Command only if they are within the Command Range of the Formation's Leader.
Therefore, if an Independent Unit is not within the Command Range of the Formation's Leader during a Formation Activation, it is considered Out of Command and cannot perform a Rally action.
@@RvTWargames I made the same without youtube.
@@RvTWargames it is inconsistency that makes it not very usable. If it makes mistakes (using oficcial rules, and they are clear, what the answer should be) - I do not think we can rely on it. Yet.
TIL the eASLRB is more than 500,000 words! :(
Aha. So my quip wasn't completely off target 🙄
Not to mention that it is the only PDF you are expected to purchase. As if ASLers haven't paid enough!
And, I suspect they are copy protected so that would fail in NotebookLM too.
I stick to ASLSK myself. Really should sell my ASLs. My remaining life on this planet is too short to learn them now!