Since the launch of Roblox, the discussion of the Metaverse in China has continued to heat up.
The concept of “Metaverse” is commonly described as a virtual world experience constructed by science fiction works such as Snow Crash and Ready Player One.
At present, the Metaverse is taking shape, and there are already many companies in China cutting into this track from the perspective of social, game and technology.
In the opinion of rct AI CEO Yuheng Chen, the domestic exploration of Metaverse is still in the state of a follower, and the real Metaverse is a virtual world entertainment experience that is more than just a game.
In a virtual world dominated by native virtual creatures, any object and subject can have intelligent decision-making capabilities. In the Metaverse, people can interact with any object to meet not only efficiency-oriented needs, but also entertainment-oriented needs. The way people feel immersed in the Metaverse is not the impact at the image level at first, but more through the interaction at the logical end, where they feel that the objects in the virtual world are the native species in it and participate in the construction and development of the Metaverse together with them.
Therefore, Yuheng Chen believes that the future “virtual human” will become the middle point of the Metaverse, providing users with services equivalent to the APP on the mobile Internet. To this end, rct AI is also building different technical solutions to lay out the “brain” of the avatar, i.e., to enable the avatar to act and make decisions intelligently.
And in the layout process, rct AI chose to cut in as a tool, hoping to iterate AGI technology with the full digital scene of the game. Yuheng Chen believes that users’ needs for content consumption and interaction are constantly increasing, and without the assistance of AI, these needs are difficult to be met more efficiently. AIGC can improve the efficiency of creation, achieve personalized experience for everyone, and greatly enrich the interaction experience.
Pure AIGC is to create virtual world content generation and creation on behalf of the virtual human brain. rct AI currently provides AI-based solutions mainly for the gaming industry, including intelligent NPCs, automated QA, dialogue systems and AI companions, continuously achieving technology accumulation and iteration in the process.
The reason why we chose to serve the game industry is also because on the one hand, the game industry is keen to use the latest technology to create new ways of playing; on the other hand, the large amount of player interaction data from games can be accumulated and used to improve the intelligence of AI. After this, rct AI can launch more standardized models and development tool models based on data and scenarios, allowing small and medium studios, or even individual developers, to create richer interactive content in the game world in a low-code way.
Company fundamentals: entrepreneurial opportunity, basic philosophy, team size
What was the opportunity of starting the company? How is it based on the judgment of the present and the future?
From the two aspects of technology and product experience, technology is actually the main reason we selected this direction of entrepreneurship at that time in 2018, because what we were doing before was to apply AI to smart speakers, voice assistants, these scenes are actually a tool scene, they provide services to consumers while actually meeting the actual needs of users, the needs of real life, we found that at least in 2018, technology is a big bottleneck, it may be difficult to have a disruptive change in the next 10 years 20 years.
Now the so-called AGI (Artificial General Intelligence), in fact, there are two sides, the downside is that now all the AI it really reason and think in a way that is different from the human, we have no way to know how they think in the end, how the human consciousness is in the end, so we can only go up in some mathematical level to copy, to find ways to copy our intelligence, we copy out the intelligence is now a set of deep learning based on neural networks, the reasoning logic behind it and our human reasoning logic is completely different.
It is because contemporary AI shows two sides, in some other scenarios, it shows a very efficient or very accurate task completion in specific scenarios, so AI can be used in various industries in recent years, which means that if you want to go to really achieve AGI, or to achieve a human-like intelligence, in fact, the current technology is a big bottleneck.
So we switched from the field of tools to the field of entertainment, in fact, the game has been serving as a good scene to experiment with new technologies, and for us to be able to take all the accumulation we already have in AI, and then to do the exploration in the C-end of the consumer market, and to allow players to experience the way to iterate forward with the technology.
This switching process has many practical advantages, for example, all the data of the game are structured data, the game itself is a human-defined system, and all the rules of system operation are human-defined.
When we do real-world services, people need to understand the real world first, and they need to mark all the knowledge, information and entities in the real world, but obviously we do not have a set of standard specifications to define the real world, so the final implementation results are uneven, and a lot of manpower is needed to do the marking work. However, in the digital content experience like games, which is naturally digital, all the processes from data generation to the end are carried out in the digital world, and there is no confirmed relationship between it and reality.
From the perspective of product experience, this thing is a historical inevitability, because human beings’ adaptability to content consumption or interaction is constantly improving. For example, in the picture, the pixel style of black and white at the beginning, then to 8 bits, to 16 bits, to 4K now, as well as real-time rendering of more HD and more detailed animation, these are the trend of people’s increasing requirements in the direction of image or visual.
The same in the logic of the content, our players are also evolving, as reflected in the cost of the game manufacturers can be very obvious. In the beginning, this type of plot-experience games only takes 2–3 years or even 1–2 to create, players are happy with it, and sales are also great. But in recent years, we can see the so-called 3A masterpiece, the development cycle of each game is very long, because it requires a very large volume of content, players consume content increasingly fast, because players have adapted to the interaction mode, for example, players in previously may need 10 hours to explore, but now may only need 5 hours to explore everything, so the remaining 5 hours either they move on to some other game or the game needs to have more content to fill, so in game experience, whether from the image or from the logic or interactive content, the player’s demand is significantly increased.
In general, in the direction of product experience, we see a trend of the times. Now players all over the world have come to a time when they need more interactive content, and players are more accustomed to the way of interaction in the digital world and more adapted to explore new content, which requires AI to help the game to generate more content to give players a better and richer dynamic experience.
What is the mission of rct AI? What kind of technology path is being followed?
Our technical route, initially, started with big game companies or co-producing games with IPs, we may not be able to really influence their core gameplay and core business model, but from these big business-side customers, we got enough data and scenario accumulation, which is very important to enhance the intelligence of our AI.
In the second stage, after accumulating enough data and scenarios, we can launch some more standardized and integrated models based on these data, which allow players, small and medium-sized studios, and even individual developers to create richer interactive content in the game in a very low-code way.
The direction we’re going it’s not a game editor, like Roblox does is low code to create the content of the scenes in the middle of the game, like building a scene, level, world. We are creating the content behind the world in a low-code way, how the player interacts with these characters and the world, how these worlds and characters will react to the player’s interaction, these are the things our AI can do.
In terms of the industry worldwide, our mission is to provide players with a richer and more dynamic interactive experience, and to enable people to better understand themselves through interactive experiences, which is a general direction for us.
In terms of specific industry-related missions, we act as a bridge between the Chinese content industry, or the Chinese game industry and the European and American game industries, so that the Chinese game industry can take advantage of the new technology application scenarios based on data and artificial intelligence to the international stage and be accepted by the mainstream European and American industries.
Can you tell us about the background of the rct AI team?
Our current team has a very diverse background, with the founding team mainly coming from Raven Tech, Inc.
At the very beginning of this company, although most of the team did not have experience in Europe and the United States, we went directly to the United States and then took root there. We entered Y Combinator in 2019 and experienced three months of incubation, which helped us get in touch with very many core resources in Europe and the United States, and we kept contacting various investors later, together with the team’s ability to adapt, it took some time for us to take root in the European and American markets.
It is true that we are not a purely technical team, but I have done a lot of R&D work in various directions before, and first of all, I am very passionate about games. One of the criteria for recruiting for the team is the knowledge of games, how many games we have played, and how deep we think about games. The core technical and engineering team has very strong product thinking and can combine theory and practice very well.
Culturally, we have Canadians, Americans, Iranians, Ugandans on our team, with many different cultural backgrounds joining us, it is more inclusive and easier to attract diverse cultural factors. Part of our own mission is to act as a bridge between the Chinese industry and the mainstream industry in Europe and the US, and also because of our multiculturalism and adaptability, it helps us to do a lot of work in the middle.
We now have 60 people on our team, with about 10 interns, and about 50 full-time employees, about 70% of whom are R&D.
Products and Services: Four AI-based solutions
What kind of products or services does rct AI mainly provide? What are the service targets and industries?
We now mainly provide artificial intelligence-based solutions to the game industry, there are four main solutions, the first is intelligent NPC, the second is automated QA, the third is the dialogue system, the fourth is AI accompanying the game.
The core technology behind these four solutions is very similar, in fact, they all give the virtual character brain, so that the virtual character can act intelligently to make decisions, only that we send it to different scenarios to use.
For example, automated QA is to send our AI to the game to play the player, after which it can play the game 24 hours a day, which can significantly reduce the cost of human testing.
Intelligent NPC is also a very important piece of our solution. Intelligent NPC includes dialogue, and on the basis of dialogue can make more complex interaction and behavior with the player. For example, the branch characters, some passers-by, some villagers, the feedback of these passers-by and villagers will be driven by our AI, according to the different decisions of the player, different interactions to make different responses.
Dialogue system is a little more simplified than the intelligent NPC, we will remove some of the complex behavioral interactions, equivalent to some of the behavior it does not involve physical, there will be no physical damage, in fact, is a relatively simple single layer of behavioral interaction, this piece we now also do a very large application direction, the highest priority of a solution, is to do AI pets.
Although this direction is said to be AI pets, in fact, all non-human creatures can access our dialogue solution, our first online customer case, Oasis, has been launched in the Brazilian market last month, and has gotten relatively good data in the past two weeks. The typical Metaverse experience, in such a social game, will come to interact with his pet, including some very interesting interactions, the player’s different interactions and conversations will affect the character of the pet, that is, the character of the pet will change with the player’s upbringing.
The AI companion is equivalent to adding our brains to all the players’ teammates or opponents, many traditional AI companies are doing matchmaking, matchmaking can of course also do AI companionship, and also has a huge imagination and market space, in addition to the technical advantages of easier Showcase, but our AI is more as a teammate, with the collaboration between players, there is some emotional interaction. For example, the type of games we choose to do AI companion is more SLG, strategy games, rather than combat games. Strategy games are not so fast-paced, more emphasis is placed on the collaboration of players , such as I will give you a little something, I will move my city next to your city, I will do the task with you, is to have this kind of personalized performance.
In the next phase, we are doing there is a standardized platform called Meta Being, which allows players to allow individual developers to create the brains behind the virtual characters or virtual creatures of the game in a low-code way, for Meta Being we have a big piece of imagination and market space, that is, we are now also doing something related to NFT. Now all the NFT is static, it is only a collector’s item concept, homogenization is very bad, whether painting, video, or model, in fact, all remain visual, but our technology can actually give these NFT a richer brain in a very fast and easy way, can evolve with the player’s interaction, through the process of nurturing can produce more value.
AIGC: not pure AI production content, the core is to solve the interaction problem
rct AI has mentioned the AIGC concept on many occasions, can you explain what AIGC is?
First of all, AIGC is not only the brain behind the interactive content we are doing, for example, sandbox games, like Minecraft or No Man’s Sky, the so-called process of generating the game world is also called AIGC. The real AIGC is the brain behind the logic, which simply determines the order of all the information composition. The AIGC is the brain behind the logic, which simply determines the order of all the information, such as what it does first, then what it does, what it says first, then what it says again, so that the player’s experience can really feel dynamic, the AIGC is actually to enrich the interaction.
The logic we are talking about is actually not important for matchmaking and competitive games, because the logic behind competitive games does not require AI to assist people to create, as long as you set the rules out, the rest is more visually generated content, but in fact, in the direction of logic, is a completely different direction of the game experience.
Another direction is through exploration, to meet the user’s desire to explore, and then the user in this game it will continue to explore and interact, and then through the exploration and interaction to dig out different possibilities, through the exploration and discovery of pleasure, such as Rougelike games, games, simulation games, they are to explore as the core of the fun, and then go to the user, so in this direction, the logic behind these games is very important.
AIGC is actually not pure AI-generated content, pure AI-generated content is possible in the future, but not possible now, because people still need to set a basic world view or character setting attributes and parameters for AI, so for now, AIGC still plays more of an auxiliary creation, solving the role of production efficiency. Our emphasis on AIGC is actually about logical content, not anything related to visuals.
In addition to generating content, another key piece is the recommendation algorithm, the recommendation algorithm is actually to help our algorithm to iterate according to the player’s interaction, and then finally be able to adapt to the different needs of each person, and finally achieve personalization for everyone, it is far more efficient than human. Simply put, our AI generates 1000 different kinds of feedback, for 1000 players with different interaction roles will make some different feedback, in the previous planning may seem only 500 kinds of feedback is reasonable, is what they can think of.
This means that if it was done by human, human could only produce 500 possibilities, and the production speed may take a month.But in the 500 kinds of possibilities our AI produced, it may have 200 kinds of possibilities that the planner did not think of, and the planner might not think and might not believe that players will find these 200 kinds of possibilities interesting, but some players might find these 200 kinds of possibilities particularly interesting, so it actually improved efficiency.
Can it be understood that AIGC is still at a relatively early stage?
It can be understood in this way. There are different ways to understand AIGC, if we understand AIGC in a broad sense, there are actually a lot of people doing it now, for example, Ubisoft, for example, some game manufacturers with self-research capabilities, doing this thing is the merchandise that players see in the game recommendation, which is actually AIGC, we emphasize the AIGC is a deeper layer, is to solve the most core interaction behind it.
So we are actually strictly speaking AIGC is doing the virtual human thing, there are actually several companies in the United States working on this direction, their core technology is mainly with OpenAI cooperation, they are mainly based on OpenAI technology to do engineering landing, so now the whole industry in general is more cutting-edge. Of course, it’s normal, first of all, this direction of the Metaverse and gamification experience has not been around for a long time, and Roblox only started to receive more attention when it was listed in February and March.
In fact, most people are still doing content at the beginning of the trend, that is, to be able to see and feel the story. The companies that focused on the infrastructure are not favored in the beginning, because there is no implementation, so the infrastructure will be weaker at first, we are determined to create the brain behind virtual people, this is the purest AIGC, but also needs a process of development.
Metaverse: avatars will be an important portion
What is rct AI currently doing, where does the connection to the Metaverse manifest itself, and what specifically is being explored?
The virtual human brain that we are working on is actually a very important piece of the Metaverse. We believe the real Metaverse is much more than the entertainment experience of a game in a virtual world. One of the most important roles it plays is that it can map almost all social activities performed by humans in the real world to the virtual world, except for some activities that require physical interaction, all of which can occur in the virtual world.
Previously, people would think that the interaction or activity in the real world is much more efficient than the interaction in the virtual world, but the success of Gather Town proves one thing, in fact, the interaction in the digital world can be more fluid, many social activities can be mapped online. In fact, there are many carriers of social activities in the real world, and these carriers are actually interactions between people or between people and services, and when your social activities are mapped to the virtual world, virtual people can be the same as real players in the virtual world, they exist in equal, and they provide services to players.
For example, if I want to socialize and go to a party, I may have to go to a community, find a certain community of interest and post in this community, and then find a friend who is like-minded, and then I’ll have to add his WeChat, through WeChat make an appointment with him at a certain place at a certain time, and I have to interact with him before I know whether my interests are the same as his, and if they are the same, I will go. In order to get there, I’ll have to take a taxi, and there will be a series of service jumps between applications in between. But in the virtual world these things are seamless, because the avatar can serve as an intermediate bridge between people and people, it can naturally interact with each player, it knows each player’s preferences, what kind of services they need, and then to provide services to each player.
For example, in the social scene my virtual character is Santa Claus, but then Santa Claus is also the virtual character of many other players, but through interacting with me it learned that I may have just recently lost my job, I’m rather depressed, and when it chatted with another person, it learned that the person is recently recruiting, it feels that you two seem to be suitable, then it can act as an introduction to introduce me to that person, because both me and that person are friends of the virtual character, there will be a natural trust between the two of us, so it is easy to connect together.
The reason why we think virtual people will be very important to the Metaverse and why what we do is reflected as virtual people in the Metaverse is because we think virtual people will serve as a middle point for people to get services and information in the future Metaverse, and it will serve as an APP in the future Metaverse, which will let everyone have both real world life and virtual world life, and these two lives exist at the same time, and then the one who provides services to everyone in the virtual world is virtual people.
Future Directions and Planning
What is the next overall development plan of rct AI? In what areas will we continue to explore in depth?
There will be two levels, the first level is the product implementation, one of the direction is the game experience, another direction is the pan-entertainment, that is, with NFT and virtual world more related experience, these two directions are our next phase of focus, there will be some landing cases and user-oriented products launched.
The second level is to promote the direction of scientific research, as I mentioned earlier, all the AI, whether based on neural network machine learning, deep learning, or reinforcement learning, behind them are not the way people really think, and the human reasoning process is still different, so it still needs a lot of scientific research. In this direction, we will invest a lot of human and material resources to do the underlying research ourselves, and also cooperate with many research institutions or research-oriented companies in the industry to do some underlying innovation.
