ChatGPT wakes up AI games?

Image source: @VisualChina

Wen | Liu Kuang

"Make new games in seven days", "AI fully completes all games!" ”...... Following all kinds of AI paintings, AI synthetic photos, videos, AI has opened a new carnival in the field of games.

For a long time, games and AI have always had an intimate relationship of "mutual support" - on the one hand, games serve as AI research bases, positively driving the progress of AI technology; On the one hand, AI is positioned as an advanced technical tool, bringing more possibilities to the game experience. Today, ChatGPT has completely exploded the field of AI, industries related to artificial intelligence have been deeply affected, and the relationship between games and AI has become more deeply connected.

AI is the living water of games

With the successive explosions of AI painting, AI synthetic pictures and videos, the importance of AI technology application has become more prominent, and the discussion of "AI+" in all walks of life is increasing, among which the topic of "AI+ game" is frequently mentioned by NVIDIA, Microsoft, Tencent, Mihayou and other domestic and foreign manufacturers, the reason is very direct and simple - in the current stage of artificial intelligence technology entering the practical stage, AI is easier to play in the field of games, and the value highlighted is more intuitive.

On the game development side, AI can run through the entire link of game development, production, and operation, and promote the high-quality development of the game industry. In the field of games, the time cost and money cost behind high-quality game masterpieces are high, and with the productivity improvement brought by AIGC technology, the impossible triangle of "high quality - low cost - short time" in game development may be broken.

For example, in the game development process, developers can first use AI to generate a large number of design drafts, and then screen and modify them on this basis, which effectively reduces the game production time to a certain extent. At the same time, the application of AIGC can liberate some human resources from a large number of repetitive tasks, allowing them to focus on more creative links, thereby making the development of "good games" more time-saving and labor-saving.

On the game user side, AI can make the machine more "soul" and enhance the user's interactive experience in the game. Usually, NPCs in the game are made by the game team in advance to make fixed texts to talk to users, the dialogue content is rigid and lacks "soul", and adding appropriate AIGC functions can make them achieve real "speaking".

Taking the "Against the Water and Cold" mobile game with the "game version of ChatGPT" as an example, the NPC characters in this version of the game can give logical behavioral feedback based on the dialogue content, and players can freely generate dialogues with intelligent NPCs to achieve deep interaction with NPCs.

Coincidentally, Mihayou's "Honkai: Star Dome Orbit" also added AI technology to the NPC's behavior mode to make it more natural to interact with users. It can be seen that to a certain extent, with the blessing of AI, game products can be more exquisite and more in line with user needs.

Of course, from the perspective of game industry participants, the core driving force for many big manufacturers to collectively increase the size of AI games is not only that AI technology can help games complete the construction of realism and immersion, but also that AI can continue to self-correct and improve with the help of massive data generated by games.

Games are a testing ground for AI

As we all know, the four major elements of AI research are algorithms, data, computing power, and scenarios. However, in the face of problems such as difficult algorithm testing, scarcity of scenarios and data, expensive computing power, and complex and changeable real-world scenarios, there are not many environments suitable for AI research. But in the game scenario, AI can complete experiments relatively easily, reducing the cost of trial and error.

On the one hand, game avatars can repeatedly train AI interactive capabilities such as dialogue, listening, and speaking. Looking back at the development of "AI+ games", the most classic is the impact of board games on AI evolution. There have long been board games that inspired AI thinking and trained the DeepMind AlphaGo series to defeat human players; The existing Meta uses the "Diplomacy" game to experiment with the negotiation, persuasion and cooperation capabilities of the AI CICERO.

It is reported that Meta's CICERO can infer the player's beliefs and intentions through dialogue, and generate dialogue according to its plan, which is currently able to play human-level performance in "Diplomacy". The data shows that in 40 anonymous online matches, Cicero averaged more than twice as many points as human players and ranked in the top 10% in more than one match.

In the game, there is no shortage of data resources, players will generate a large amount of data in the game every day for AI for deep learning, choose to explore AI capabilities in complex game games, is a good way to "make AI closer to humans".

On the other hand, most scenes in reality can be simulated in the game, providing a good reference for AI landing in real scenes. In the game, high-simulation restoration of real scenes can be carried out to create suitable conditions and environments for AI learning, and the AI research team can continuously polish AI algorithms and verify technical effects through complex game scenes.

For example, in the study of autonomous driving AI algorithms, researchers extract training data from GTA in different scenarios, and then provide these labels to machine learning algorithms, so that they can identify cars, pedestrians and other objects displayed on real streets to further optimize the effect of autonomous driving.

Including the now widely known ChatGPT, its birth is also due to the disguised promotion of the Dota 2 game project. It is undeniable that games have always been an important training place in the field of artificial intelligence, and may also be the key driver of the butterfly change of the AI industry in the future.

At least from the success of ChatGPT, game technology can indeed drive the innovation and development of AI, so guess that the next milestone product of AI is also likely to be born with the continuous upgrading of game technology.

To sum up, now that ChatGPT is on fire, AI and the game industry are becoming more and more closely integrated. On the one hand, relevant AI technologies have been deeply involved in the game field, bringing new changes to the industry; On the other hand, the simulation environment with good game scenes feeds back the upgrade of related AI technology. And in this interactive relationship, the game track will be able to fully absorb new "nutrients" and achieve real growth.

ChatGPT gives AI games more possibilities

ChatGPT detonated the AI field and ignited the enthusiasm of players on all tracks. It is not difficult to find that at present, AI games have become the core direction of Tencent, Netease, NVIDIA, Microsoft and other domestic and foreign leading enterprises, and the reason is that AI games have shown great development potential in the context of the explosion of ChatGPT.

The first is a renewed increase in profitability. On the one hand, ChatGPT resonates with the game industry and promotes the development of chips, terminal hardware and other related businesses. Take NVIDIA, its chip business chassis has always been a game, the growth of artificial intelligence business can bring it additional profit direction, such as AI game chips This big selling point.

On the other hand, from the perspective of the game product itself, ChatGPT-related technologies to improve the game experience of players will drive the number of users, as well as the payment rate and ARPPU (average revenue per paying user) to a certain extent.

The second is a new extension of application scenarios. On the B side, large manufacturers can use AI technology to provide game developers with more diversified solutions. For example, Tencent launched a 3D virtual scene automatic generation solution, which uses AIGC technology to help developers create highly realistic and diversified virtual city scenes in a very short time.

In fact, not only Tencent, but now deep commercialization of the B-side has become a major keyword common to large manufacturers, and they are all looking forward to the B-side breaking out in the trend of "AI+ games". At present, Baidu and Ali are actively embedding generative AI into their own products, and in the future, like Tencent, it may be sooner or later to try to make the seeds of "AI+ games" blossom on the B side.

Moreover, the development of the two major industries of AI and games has been able to take a step forward. In the field of AI, games facilitate AI training and accelerate the progress of AI research and development; For games, AI drives game innovation and enhances the gaming experience. To a certain extent, AI games can simultaneously raise the development height of AI and the game field, and for players such as NVIDIA and Tencent who intend to grasp both AI and games, it is imperative to promote the in-depth development of AI games.

AIGC continues to go deeper, and the game may become the best gripper for AI

Through the launch of ChatGPT series products, AI technology has received unprecedented attention, but in reality, the influence of foreign AI technology is better, for domestic manufacturers that want to achieve long-term development of AI, it is urgent to find a good grip for AI, and games that integrate text, images, sound effects and other elements have become the best channel for AI technology landing.

At present, major manufacturers are vigorously promoting the combination of AI and games, and "AI+ games" have become a general trend - Ali's Lingxi Interactive Entertainment "Three Kingdoms Strategy Edition" adopts Qiyuan World Game AI solutions, and Mupu Technology's "MLBB" adopts hyperparametric AI Bot.

AI and games have been deeply bound, and with the further implementation of AIGC-related technologies in the future, the value of the game industry will also be re-examined. After all, in essence, games, as a collection of major content elements, can naturally fully enjoy the benefits of content productivity changes.

According to Guotai Junan Securities, the actual sales revenue of the domestic game industry in 2022 will be 265.9 billion yuan, and the expected R&D expenses will account for 15%, assuming that the proportion of AIGC can be AIGC in the future is 50%, then the market size of AIGC in the current game industry can reach about 20 billion yuan. Therefore, the integration with AIGC is expected to increase the market value of the game industry.

However, it should be noted that for the entire game industry, "AI + games" have become popular, bringing new opportunities and many unknown challenges.

For example, the slow progress of AI research and development, and the inability of related product applications to be implemented in the game industry, will affect the commercial value and return on investment of game products. Data security issues involved in the training process of AI models based on deep learning; The development of AI technology has brought about a new round of regulation of game products.

Like the development of anything, AI technology will also have two sides, and the combination of games may not be completely praised, but it is certain that the addition of AI technology is expected to improve the quality ceiling of the game industry, and in the future, with the continuous deepening of AIGC-related technologies, "AI+ games" may open a different "new world".