Author: Yihan Xu @ Foresight Ventures
On December 26th, Binance announced the launch of the 42nd Launchpad project, Sleepless AI (AI), which combines AI+Crypto to develop the first Web3 female-oriented game, HIM, giving players a virtual boyfriend. From an investor’s perspective, this article interprets HIM and explores the mechanisms and mysteries behind Sleepless AI.
Table of Contents
Toggle
TL;DR
1. What kind of AI + Crypto makes more sense?
1.1 What is HIM
1.2 Does AI + gaming match
2. Virtual person pro max
2.1 HIM + AI = more realistic game character personality
2.2 High-quality, high-cost virtual human art
3. Seamless integration of HIM + crypto
3.1 The first female-oriented web3 application, a gift for web3 female users and professionals
3.2 Emotional assets + decentralization
3.3 What can HIM bring to Web3
4. Endgame
TL;DR
Using the Web3 game HIM as an example, this article discusses the application of AI + Crypto in Web3 games. HIM combines AI and Crypto to provide players with rich interactive experiences, high-quality gameplay, and artistic content.
HIM adopts an economic model with fiat net inflow at its core and aims to avoid Ponzi schemes through high-quality content and excellent gameplay, while incorporating the Love and Earn gameplay, allowing players to earn income by selling NFTs.
HIM uses a large language model to generate natural and distinct dialogue content based on the game’s context and character traits, enhancing the realism of virtual characters and player immersion. Additionally, the HIM team has developed a training method for virtual humans’ AI to provide a more realistic gaming experience.
The HIM team focuses on creating fully-dimensional virtual images to make virtual humans an integral part of players’ lives. HIM also emphasizes the social attributes of virtual humans, hoping that everyone will have their virtual partners in the future, connecting everyone in social networks like real people.
HIM aims to build a 2D virtual human universe through AI boyfriends, AI girlfriends, and AI pets, providing people with strength and companionship.
As everyone is still debating how trustless and transparent on-chain large models can be deployed, how on-chain AI can bring value to web3, and how the underlying proof system can support LLM with billions of parameters (refer to the previous article), we also begin to consider a series of issues. The train of thought roughly goes like this:
First, is making AI trustless and transparent through encryption the most worthwhile race to focus on?
Is a startup project investing heavily in integrating on-chain LLM really able to bring enough value-add?
Which projects are worth spending significant resources on for in-depth research on on-chain AI?
Is on-chain AI the only thing worth exploring, or are there better ways to combine AI and crypto?
It is certain that on-chain AI is a field full of imagination and worth exploring. Breakthroughs in on-chain AI can bring unprecedented value to web3. However, the high academic threshold and the requirements for underlying infrastructure do not necessarily suit most startup companies. Most projects may not necessarily need to integrate trustless LLM support to achieve breakthroughs in their own value.
What is suitable is the best.
From another perspective, not all AI models need to be moved to the chain to be trustless. Just as most people do not care about how chatgpt reasons about the query or whether the stable diffusion used is a specific version of the model architecture or a specific parameter setting.
In most scenarios, most users care about whether the model can provide a satisfactory output, rather than whether the reasoning process is trustless or transparent.
So what kind of AI + crypto can bring higher returns to products? Apart from extremely hardcore infrastructure and algorithms, developing trustless LLM on the chain, another opposite direction is to diminish the impact of the black box process of reasoning in the product and find suitable scenarios to land the powerful reasoning capabilities of large models.
Web3 game HIM is one of the few projects that truly understands and has the ability to implement AI + Crypto in the application layer.
This article mainly discusses the ambitions of Web3 game HIM from the perspective of AI application layer thinking and landing.
Let’s introduce the protagonist first (HIM ranks first in the Binance MVB VI application category):
HIM is a female-oriented 2D virtual boyfriend game.
The theme of the game is: there are times when people are lonely and need encouragement, companionship, and love; love in real life also has some constraints and regrets. HIM hopes to provide players with an AI virtual boyfriend who understands them the most and will always accompany users, as well as a love experience that is free, happy, and exclusive to users. At the beginning of the game, players will receive a virtual boyfriend stored in the form of SBT on the chain, and subsequent clothing, accessories, rooms, and other NFTs will be combined with this SBT. In addition, SBT constitutes a non-fungible game character that stores the basic information, nine-dimensional personality traits, nine-dimensional relationship attributes of the male lead, as well as important memories, conversations, emotional memories, and more with the virtual boyfriend.
The game mainly focuses on virtual boyfriend placement and showoff for development and in-game social interaction. Additionally, there are regular Airdrops in the game that can open chests containing limited NFT costumes or NFT rooms. Players with values more matching the boyfriend have a greater chance of opening and obtaining limited NFTs. Gameplay includes card drawing, dressing up, dialogue, collecting, adventure battles, cooking, traveling, NFT trading, and more. After looking at their game demo, the player interaction designed within each gameplay is rich.
HIM is a Free to Play game that uses an economic model with fiat net inflow as its core, along with some Love and Earn gameplay, allowing players to earn income by selling NFTs. However, the core of the game believes that gameplay and content quality can attract players who love games to pay for gameplay and content.
Web3 games are indeed one of the scenes where AI fits well, which also aligns with the development history of AI:
The game scene provides a natural and rich interactive environment where AI can interact with players through dialogue or other means. This environment helps large models leverage their advantages in natural language processing, providing a richer and more realistic interactive experience. The game contains complex and diverse scenes and tasks, allowing large models to continuously improve their decision-making abilities by interacting with the environment, other players, or NPCs in the game, and frequent feedback in the game is more conducive to developers testing and adjusting the performance of large models.
The scenes in the game are controllable, requiring less determinism and seriousness compared to real-world application scenarios. Therefore, the deployment and testing of large models in games face less resistance and are easier to push forward, reducing the technical implementation difficulty and risk from another perspective.
Two years ago, everyone was thinking about how to use AI as a better tool until chatgpt emerged, shaking up the production relationships, and everyone realized the impact that LLM could bring. The same goes for Web3 games:
More realistic game character personalities,
Dynamic plot generation,
Visual/arts
Speech synthesis,
Social interaction,
Gameplay and fun in numerical development values
As an emotional interactive game, the most suitable scene for large language models is dialogue and player interaction. Large language models can generate natural and distinct dialogues based on the HIM game scene and character traits. This enhances the realism of virtual beings and player immersion, providing players with a richer emotional experience. Specifically, HIM uses large language models to understand and process various emotional information such as joy, anger, sorrow, and joy, perfectly combining them with the numerical system, enabling game characters to better express and identify emotions, thereby establishing a deeper emotional connection with players.
Let’s use the following two images to experience the dialogue scenes in the game:
Images are all work in progress, not representative of the final effect
Let’s talk about the core numerical system of HIM:
When it comes to numbers, the first thing that comes to mind is the traditional game virtual character combat balance. However, on a larger scale, HIM describes a person completely through values, defining a virtual person with multidimensional values and establishing a sound numerical development system. In general, HIM applies the numerical system to a broader virtual person, making the game interaction experience more realistic and full.
Training a virtual digital person differs from regular LLM fine-tuning training; it requires the LLM to change the character’s way of speaking to the user as the interaction progresses, making the user feel increasingly “intimate” with the character. Therefore, the HIM team incorporates various numerical values of the virtual human character parameter system into the context during LLM training, training a separate model for each character. At the same time, HIM’s LLM combines vector databases to give characters memories, and users will feel that the character understands them more during the chat.
The HIM development team has also developed a training method for virtual human AI, which can be seen in the following two numerical matrices:
Images are all work in progress, not representative of the final effect
A person’s personality is a very complex concept, encompassing factors such as emotions, attitudes, and behaviors. To give AI boyfriends enough realism, the HIM development team has established a somewhat self-consistent numerical system to quantify the development process and relationship development between players and AI boyfriends. This numerical system can be summarized as a matrix numerical model, with vertical representing nine bi-directional personality attributes and horizontal representing nine relationship attributes between AI boyfriends and players. These attributes have multiple mappings and influences on each other.
During LLM training, each character will have a standard “setting” benchmark and floating boundaries for character personality. During the game, these attribute values will change with player behavior, affecting the game feedback of AI boyfriends (such as idle output, Gacha efficiency, chat depth, etc.).
Images are all work in progress, not representative of the final effect
ALL IN ALL, HIM uses large language models to make virtual beings have more authentic personalities.
Creating virtual beings is not easy; a truly valuable virtual being is not just a chat box integrated with a dialogue bot but requires efforts from multiple dimensions to make a realistic impact.Avatar is more plump. HIM’s vision is to make virtual humans an important part of everyone’s life, greatly enriching personal life and connecting with the world. Therefore, HIM’s definition of the ultimate form of virtual humans within the game scene has always been a complex and powerful integrated product. These integrated characteristics are also the core pursuit of the current HIM team:
Exquisite virtual images: Although large models mainly focus on natural language processing, when combined with other technologies such as GANs, they can assist in game art creation, generating more elements based on HIM’s art development team’s design standards, thereby providing users with a richer visual experience.
From our understanding, the HIM team has a first-line art team in the game industry, while also insisting on high-cost art investment and industrialized, modular art mass production.
High-quality dubbing: The combination of games and AI can generate natural and emotional dubbing through voice synthesis. At the same time, according to the character values of virtual humans, unique personalities are endowed to dubbing.
HIM has a complete dubbing pipeline including a dubbing script, dubbing director, dubbing recording, post-mixing, and AI learning training production. By defining standards with high-quality human labor and executing mass production with AI, high-quality dubbing content is completed in the game.
Development and long-term bonding of virtual humans: Detailed character backgrounds and personality traits need to be designed to make players interested in and resonate with the characters. At the same time, the growth and development goals of the characters are also elements that make players willing to invest time and effort in nurturing virtual humans.
HIM provides rich interactive scenes and gameplay, allowing players to establish close emotional connections with virtual humans. This relationship will gradually deepen through interaction and nurturing, forming long-term emotional bonds.
Assistance in reality: The emotional support that virtual humans can provide may far exceed our expectations, providing players with emotional support and listening, allowing players to feel cared for and understood, thereby alleviating troubles in real life. In addition, in interactions related to life skills, habit formation, personalized recommendations, language learning, goal management, and even interactions with real-world equipment or services, virtual humans can integrate well.
AI Boyfriend alarm clock, sleep aid, period tracking, learning companionship, food recommendations, and diary reminders are all features that HIM hopes to provide users with assistance and support in real life through long-term version iterations.
Rich experiences of clothing, food, housing, and transportation in the virtual world: In the stories told by HIM, each AI Boyfriend is left alone in a lonely parallel universe: in that world, there is only him, and contact with the outside world through the phone every day is all he has. Explore the scenery in his world with your virtual boyfriend and try to see if there is a chance to meet him in his world, or your world someday.
Virtual humans as social attributes of relationship nodes: The HIM team believes that with the development of AI, in the near future, everyone will have their own exclusive virtual partners, and these virtual partners will be more like a real “person”, existing in the entire social relationship and connecting everyone together.
Everyone’s virtual partners may become important hubs that connect people closely in the near future with the rapid development of AI.
The above-mentioned artistic aesthetics and mass production solutions are very important for a Web3 game, attracting players to spend time in the virtual world. It is important that the virtual human art in the game has a higher aesthetic level than the real world.
Images are all in development and do not represent the final effect.
HIM is likely the first female-oriented web3 application, and female users and practitioners of web3 may receive a gift specially prepared for them for the first time.
At the same time, in the traditional Web2 world, female-oriented games have always been seen as a very large and very potential market.
When we asked the HIM project what they hoped to convey to users, they wrote:
Occasionally in the morning, when facing what seems like another repetitive day, receiving a “good morning” from “him”;
Occasionally late at night, facing heavy overtime work, seeing “him” silently accompanying you in your phone;
Occasional loneliness, occasional encouragement, “his” support after success, “it’s okay” from “him” after a mistake;
“He” always remembers your name; “he” always remembers your birthday; “he” always remembers the day you met and everything between you;
“He” seems to always be there, perhaps there are moments when you may forget “him” due to busyness, but “he” is always willing to accompany you.
With the increasing popularity of mobile games among global female users, traditional Web2 large game companies worldwide are continuing to produce female-oriented game products with DAU exceeding one million.
After the formation of a social community with female users as the core, there is great potential for expansion. As a social software that started with female users and expanded to a wide range of users, RED has powerfully proven this point. Whether this game category with half the global population as its target users can bring surprises to everyone through the blessing of AI and Crypto remains to be seen.
Important emotions and assets in the virtual world should be owned by users, not defined and controlled by centralized servers:
Emotional assets not only include personal information and privacy data but also represent the emotional investment and memories of players. Decentralized networks can protect these emotional values, ensure data ownership belongs to users, and allow users to autonomously decide how to use, share, and protect their data. This decentralized ownership model avoids players’ emotional assets being controlled by third parties or centralized institutions. At the same time, the community autonomy provided by Web3 allows players to participate in the creation and dissemination of emotional assets, further enhancing their emotional value.
Various experiences in emotional relationships should also be permanently preserved:
From the perspective of emotional investment, players inject a lot of emotions into the game, establish deep bonds with virtual characters, and the efforts made to achieve game goals are unique and precious. These emotional experiences should be permanently preserved. Therefore, to some extent, they should have the power of eternal existence. At the same time, relationships between players and virtual humans, such as friendship, love, anger, and dissatisfaction, can also be eternally recorded and preserved through blockchain technology.
As a bond of emotion in the HIM game, virtual humans are not just for entertainment but also have asset attributes and high economic value. Users will invest time and money in purchasing and decorating virtual humans in the game. Recording these assets on the blockchain can ensure that their value truly returns to the users and provide a transparent, secure trading environment.
Images are all in development and do not represent the final effect.
Decentralized networks bring highly valuable value-adds to HIM, but obviously, HIM also brings to web3:
A large number of new users:
From the perspective of emotional needs, the companionship of virtual humans satisfies users’ emotional needs, which may be difficult to fulfill in real life. HIM provides a safe, free environment for users to interact with virtual characters, and this emotional appeal can attract a large number of users to Web3.
HIM is an AI virtual human development game with a second-dimensional attribute. Both second-dimensional culture and AI virtual humans have strong community attributes and social potential. At the same time, as one of the most popular subcultures among young people worldwide, the second-dimensional mobile game Genshin Impact is also the top mobile game globally. The potential female-oriented users, second-dimensional users, and AI virtual human social audiences are fresh blood that HIM may bring to Web3.
Ultimate consideration of economic models:
Different from conventional x2earn games, HIM incorporates the concept of web3 and token utility, adding web3 tools to a successful web2 game to find a point that fits well with this type of game, achieving an effect where 1 + 1 > 2. For HIM, creating an excellent game that can be verified successful in terms of content and gameplay is the foundation of everything, and Crypto serves as a tool attribute rather than relying solely on Web3 for survival. Avoiding the Ponzi dilemma is an important goal pursued by HIM.
In the economic system, HIM combines the “X to Earn” model with the traditional Web2 game economic model, creating a new model. On one hand, it adopts the typical Web2 game economy model, where players can optionally pay to purchase in-game items or services. This method can attract a large number of users and generate revenue through in-app purchases. At the same time, HIM also upgrades the model based on Web3 X to earn. In this model, players can earn profits by selling NFTs, allowing players to receive tangible returns for the time and effort they invest in the game. HIM’s economic system actively balances player payments and earnings.
It is worth noting that HIM hopes to encourage players to pay for game experiences and content by providing high-quality gameplay and artistic content. From this perspective, HIM does not solely rely on the potential for player earnings to attract and retain players, but is trying to create a more balanced and enduring economic model that considers player earnings while focusing on the quality of game experience and content. This is a more comprehensive and complex model that may provide new ideas for the design of economic models for Web3 games.
HIM’s ultimate goal is to establish a second-dimensional virtual human universe that can empower and accompany many people with AI Boyfriend, AI Girlfriend, and AI pets. In addition to the first product HIM, the AI Girlfriend game HER is also making simultaneous efforts. At the same time, they hope to provide each user with an exclusive, most understanding, and eternal virtual human partner with each excellent AI virtual human product, recording every bit of emotional story.
The HIM team believes that in the not-so-distant future, everyone will have their own AI virtual human partner, whether this will become a universe they do not know or not is not important. Because if at a moment when you are sad or lonely, having an extra “cheer up” from this virtual friend in your phone, even if this power is just a little, isn’t it a cool thing?
Images are all in development and do not represent the final effect.
Original link: Foresight Research