YGOrganization | [Master Duel] Upcoming Products and Latest on the AI project

I am pleased to introduce our new AI rulers.
-New Design House: Blue-Eyes Max
“A deck using a powerful “Blue-Eyes” monster as a secret weapon. Many “Blue-Eyes” support cards like “White Stone of Ancients” and “Blue-Eyes Jet Dragon” have related effects that directly affect your Graveyard. Keep an eye on your Graveyard status. Once you set up and summon formidable monsters like “Blue-Eyes Chaos MAX Dragon,” your victory is secured! These cards are included:
Blue-Eyes White Dragon (UR)
Blue-Eyes White Dragon (UR)
Blue-Eyes Chaos MAX Dragon (UR) Available on February 7th (times apply) -New Optional Pack: Rise to the Future
Cards included:
Diabellstar the Black Witch (UR)
Snake-Eyes Flamberge Dragon (UR) Cosmic Quasar Dragon (UR)
TG Glaive Blaster (UR) Emperor Charles the Great (UR) Available on February 7th (times apply) -New Option Pack: Name of Champions
Cards included:
Lady Labrynth of the Silver Castle (UR) Destiny HERO – Destroyer Phoenix Enforcer (UR) Baronne de Fleur (UR) Dingirsu, Orcust of the Evening Star (UR) Accesscode Talker (UR) Available on February 7th (times last) – ———
In addition, during the event, there was a brief presentation about the AI system (starts at about 47 minutes). Ken’ichi Kataoka (Master Duel/Duel Links Senior Producer)
Mr. Hattori (Programmer) Q: Can you give us a straightforward explanation of what this AI project is all about?
Mr. Hattori: This project is a system that utilizes Python for Master Duel. In short, with AI experience and Yu-Gi-Oh! OCG, anyone can create a Duel AI that can play Master Duel.
Q: Does “everyone” mean, even someone outside of Konami?
Mr. Hattori: Yes, that is correct. Initially, we want to focus on a Duel between two AIs generated by individuals or a group outside of the Master Duel design team.

The graphics on the screen illustrate that the special features are designed for the AI to interact directly with the game (top image).
The text in the middle of both these images explains that AI-based services and capability evaluation etc. create a responsive system.
The image below explains that it is also possible to use the console simulator to run Duels at high speed with only a text view. The two circles on the right explain that information from Duel and the board is processed and distributed to the AI, which responds with its inputs and requests. Assistant AI (written in Python) is an AI framework/learning model. Q: This looks very different from the usual screen we see in Master Duel, can you give us a brief overview of what we are looking at?
Mr. Hattori: Yes. The first thing I want to show on the bottom left, we see the amount. This percentage is the winning odds from the AI moves. In the box on the bottom left, we can see all the estimated wins, but on the bottom right, we can see the top 5 moves. So we’ll be focusing on the AI when choosing move #1. Right now, the chance of winning is low, but soon, there will be a move that increases the chance of winning, so let’s pay attention to it. (The AI uses The Melody of Awakening Dragon to seek Blue-Eyes Chaos MAX Dragon and is now able to Ritual Summon, increasing the chance of success to about 80%). some, but that’s not it, is it?
Mr. Hattori: That’s right, these aren’t CPUs, but external software running Master Duel. In other words, machine learning is being used to train AI. User data is not used at all, and the AI is learning from scratch by doing Duel and Dueling over and over again, without relearning the rules.
Question: How many times can you say “repeatedly”?
Mr. Hattori: They have trained Dueling and Solo Mode CPU for 100000 times.
Q: So even if you don’t teach the rules, you just make it Duel 100000 times, and the result is this current AI?
Mr. Hattori: Yes.
“He probably played more Duels than humans!”, “It’s another course!” Q: There is a lot of information on the screen, which looks great!
Mr. Hattori: Yes, we thought it would be interesting for anyone watching to show what the AI was thinking on the screen. Ken’ichi: We are game developers, however, we are not AI experts. That’s why we thought it would be interesting to see what kind of AI could be created by a company that works well in this field, and what kind of Duels would be between these AIs, and that’s why we created this project. “Who can build the strongest AI?” Of course, it is also possible for AI players to Duel people, so we also want to see what kind of Duel will happen between the “world’s #1 AI” and the “World Championship winner”.
Q: So an outsider, someone not from Konami, can participate in this?
Ken’ichi: Of course, this isn’t a system that we’re just going to give to everyone to create their own AI. [protocols/rules/terms of service] is being developed, and we are currently making progress so that later this year, AI Duel developer titles will be posted on the official Master Duel website.
Mr. Hattori: First of all, we can expect the participation of groups / organizations that have the ability to create AI. If possible, competition between universities or the like.
“It would be great, every university training their AI and then competing against other universities!”
Ken’ichi: Yes, AIs developed by universities are very interesting.
“Yes, they have their own unique quirks depending on their creator” Ken’ichi: So far, this discussion has been mainly for AI enthusiasts, but there will be benefits for all Duelists as well.
Mr. Hattori: We have already shown this in the previous video, but please watch this:

Mr. Hattori: Using the successful AI function, we can create the equivalent of a well-known meter from go/shogi.
Q: I see, as seen in the picture, the player is in a good position, and that is what is shown, right?
Mr. Hattori: That’s true.
“That’s right, with just one look, you can easily tell that the player with the Mystical Space Typhoon is the favorite to win!”
“For us sports commentators, you can just look at the screen and see who’s winning”
“[…]and you can also have a case where someone has a 5% chance of winning, then they draw a new card and grow! Suddenly he can change the course of events!” Q: Is there anything else?
Mr. Hattori: Yes, we are also happy with the function that allows you to predict what the face-down cards will be. If we can look at the next slide…

Mr. Hattori: Here we can see the AI trying to predict what the face down card might be. [Vaguely explains that the prediction could be off, the A.I. could hold back, move aggressively, etc.]
“So the AI thinks: The face-down card can be a Ring of Destruction, right”
“Probably got hit by RoD too many times”
“Who did he fight 100000 times?”
“This is how AI creates its own challenges, depending on who cuts it”
“It feels like the world of VRAINS, where you can search for Dueling data”

Ken’ichi: The idea behind Master Duel is that both participants and spectators can enjoy Duels. This is why we are also trying to improve “what is shown”. We believe that explaining in a very simple way the best parts of Master Duel to the audience is an important part of growing the group of competitive players. This is why we continue to update Master Duel for Duelists, and those who enjoy Duelists. Stay tuned for the future of digital Yu-Gi-Oh games! Then the live show begins. Both Ken’ichi and Mr. Hattori quickly mention that even though AI has jumped 100000 times, this is still considered “young” and “undeveloped” in terms of AI, so it can move unexpectedly. Mr. Hattori emphasizes that when the AI is close to attacking and the opponent controls a face-down card, they are making a decision using cards that may be face-down, and when they choose that they are not a threat (Eternal Life instead of Mirror Force). ), continues with the attack. He also points out that the AI understands the opponent’s Deck in order to make predictions. Mr. Hattori is asked if it is possible to teach the AI about the existence of other Decks so that it can learn how to deal with them, for example, taking a currently powerful Deck, such as “Horus”, and teaching it to the AI, and Mr. Hattori answers that it is possible to teach the AI that way. The AI uses the Ring of Destruction in the Final Stage, and when Mr. Hattori is asked if the use of this card was learned by the AI, he explains that through thousands of Duels, the AI has a high value of the moves it leads to victory. Although the AI does not know the rules of Yu-Gi-Oh, it understands that these moves lead to good results. (Duel unfortunately ends early due to time constraints)

Author: OpenAI