Excitement and Anxiety Collide at GDC2024

Some are happy and some are sad.

On March 23, the five-day Global Game Developers Conference (GDC) 2024 came to an end. Although from a player's perspective, the news about new games announced by major manufacturers at conferences attracts the most attention; for industry insiders, its meaning and value are multiple.

Advertisement

On the one hand, it gives developers from all over the world the opportunity to gather together (in the past few years, such opportunities were very rare), and many participants have shown off their “star chasing achievements” – working with admired teams. People and developers communicated and took photos. On the other hand, many practitioners will expand their network and meet potential partners and investors at the exhibition. Many supporting organizations and groups will also hold their own exchange meetings and small meetings at GDC.

At the same time, major manufacturers showcase their cutting-edge technological achievements at GDC, and many summits focus on a specific technology. The hottest topic this year, and also the topic that attracts the most attention in the GDC preview, is the application of machine learning and generative AI in the industry. As expected, some major manufacturers brought some demos or closed-door demonstrations to show the effects that this technology can currently achieve.

This year's GDC has another special feature. Due to the overall economic downturn and shrinking industry scale, manufacturers around the world have been affected by layoffs. So this time GDC also saw a large number of developers trying to find new job opportunities. GDC officials have therefore set up some special seminars to deal with this situation.

This gives this GDC a special quality of “surprise and frustration”.

Advertisement

More than 30,000 people attended this GDC

Domestic manufacturers participating in GDC: It has become the norm, focusing on technology

According to the practice of reporting on international events, we should introduce the domestic game manufacturers participating in GDC this time. But this is becoming more and more normalized and taken for granted, and the summits and invited speeches given by many manufacturers reflect a strong technical nature.

Rubik's Cube Studio Group and Tianmei Studio Group under Tencent Games shared a lot of development technologies at this GDC. These include the ray tracing technology used in the big world scene of “Dark Zone Breakout” on the mobile terminal, as well as the frame prediction technology required to achieve 144 frames on the mobile terminal; the cross-platform development pipeline in “Operation Delta”; and “Naruto Ninja” “Game” how to commercially apply reinforcement learning technology to a fighting game for the first time.

NetEase Leihuo Studio participated in the China Board Game Summit and showcased the current Chinese board game market and representative works. At the Machine Learning Summit, they shared some game rendering technologies based on machine learning.

In terms of independent games, domestic manufacturers have also made a lot of moves. In fact, for independent game developers, GDC has always been the most direct way to obtain game feedback from industry insiders. This GDC Independent Game Summit also introduced many particularly practical development and project management skills.

Domestic independent game publisher GameraGames will participate in the exhibition with a variety of games it publishes, both domestic and overseas. In addition, they also held a seminar with the theme “The Rise of China's Independent Games and How to Enter the Chinese Market”. This is a hot topic in the overseas game industry at the moment, and it also provides overseas developers interested in entering the Chinese market with an opportunity to directly showcase themselves.

Chinese practitioners related to independent games at GDC include Remai Games, as well as individual developers and individual publishers who purchased tickets out of their own pockets. This opportunity to communicate face-to-face with players and peers from all over the world and get feedback is extremely valuable.

GameraGames held a networking session during GDC

Industry Focus: Generative AI

A topic that has aroused industry expectations since before the official opening of GDC is generative AI.

After nearly two years of development, almost every major manufacturer has set foot in this field to a greater or lesser extent. At present, manufacturers' attempts can generally be divided into two aspects: on the one hand, they try to apply AI technology to workflow to simplify work steps and improve efficiency – this step usually does not include content directly presented to players, and It is applied to the original material or program. On the other hand, they are trying to train generative AI so that the AI ​​can intuitively react differently to the player's actions in preset situations, thereby providing varied content. Ideally, this would make the NPCs in the game more realistic and believable, improving the overall immersion.

The result of these attempts are various new AI tools demonstrated by some manufacturers at this GDC, including AI engines adapted to game development needs and some prototype demonstrations. According to developers who arrived at the scene, these demonstrations were very popular, and related lectures and discussions were often full and even had long queues.

What attracts more attention is the self-developed game AI engine GiiNEX released by Tencent. According to official information, the role of this engine is to “provide rich AI solutions for the entire game life cycle based on generative AI and decision-making AI technology.”

To put it simply, this engine can help developers use AI to generate needed materials at different stages and in different fields. For example, if developers need to write an NPC dialogue, GiiNEX can help them generate the text; if they need to make the characters do some fixed things, GiiNEX can help them generate behavioral logic; if they want to create a city scene, GiiNEX can help them generate it. 3D city models, even levels and music can be generated. These contents can be used as assistance to greatly improve the efficiency of developers in producing high-quality content.

Tencent conducted a live demonstration of urban layout using GiiNEX: developers only need to draw the main features of the city, such as water bodies, main roads and mountains, and can quickly generate a complete vector road network layout. After setting up special building areas, dividing urban block functions, and adjusting building height distribution, a realistic and mature blueprint can be quickly formed. On this basis, GiiNEX can also easily automatically generate the appearance of a single building, or quickly generate a three-dimensional indoor scene using multi-angle photos of the room. This is many times faster than modeling and painting a scene by hand.

Decision-making AI technology can be widely used in game testing, simulated gameplay and other scenarios to help developers quickly iterate games.

GiiNEX can help developers save manual labor in many aspects

At the player level, GiiNEX can also help them generate UCG content more conveniently in some games. Currently, this technology has been implemented in “Yuanmeng Star”. When players are building, they only need to enter preset keywords, and architectural reference drawings, theme color schemes, etc. will be generated for their reference.

As for the generative AI that can directly affect the player's immersive experience, the main representative is “NEONPC” released by Ubisoft at GDC. However, Ubisoft adopted a closed-door demonstration format, so the number of people who can directly experience it is limited.

According to Ubisoft’s official statement, NEONPC aims to allow NPCs and players in the game to interact as naturally and realistically as possible: “This prototype inputs the character personality, background story and other elements created by the game writer into a large language model, and then uses Nvidia's Audio2Face application and Inworld's large language model (LLM) are trained to enable NPCs to make appropriate language responses based on scene context, player input and other factors.”

In other words, after the developer sets basic identity information for the NPC, the generative AI will conduct autonomous conversations with the player based on the acquired language ability. This prototype can be applied to many different types of games in the future, including various adventure puzzles, open worlds, and management games.

Although it was a closed-door test, there were still videos of actual conversations leaked. It is worth noting that AI not only generates dialogue for NPCs, but also simultaneously generates dubbing and real-time expression and lip matching. Although it is still a bit stiff as a prototype, with unnatural pauses and an awkward tone, the presentation effect is relatively complete.

At the same time, players can directly use the microphone to talk to NPCs in natural language instead of selecting options on the screen. After AI performs speech recognition, it will give its own answer based on keywords. For example, in the demo, NPCs tried to convince players to join a resistance group. The player said that he has a wife and daughter and it would be too risky to join. The NPC responded accordingly, “We want to create a better world for the people we love… This is a risk worth taking,” and when asked by the player, he said, “I also had a son.” In these conversations, NPCs not only respond to the player's words, but also gradually reveal the backstories written for them by the developers.

Ubisoft's NEON PC can respond to conversations input by players using their microphones

At this stage, it can be said that NEONPC generates an “advanced chatbot” in the game. But if you want to recognize all of the player's natural reactions and provide feedback, such as randomly evaluating how the environment makes you feel, this prototype is definitely not enough. They can still only advance the storyline in the demo by triggering keywords and events, but they won't keep reciting the same lines after running out of dialogue like in many games.

Nvidia also demonstrated a “SmartNPC” generated in cooperation with InworldAI at GDC. Similar to Ubisoft's demonstration, NPC can respond to what the player says and interact dynamically with the player. Expressions and voiceovers are also generated simultaneously, allowing them to jointly explore the case in a small scene.

In addition, Nvidia also demonstrated Audio2Face technology in the early access of the MMO game “Jade Dynasty”. This technology can automatically generate corresponding mouth shapes based on different dubbing languages, eliminating the need for staff to manually adjust the character's animation presentation. This may bring more convenience to the development of multilingual games and localization of games.

King, a major casual game company, shared its experience in using AI for level design at GDC. Candy Crush Saga is a classic matching game that focuses on endless levels. According to official disclosures in the speech, there are currently more than 16,000 levels. On the one hand, King is working on using generative AI to help generate levels; on the other hand, the AI ​​will predict and imitate player behavior and test how playable and interesting these levels are. These can greatly reduce duplication of work for developers, while also giving them the ability to run hundreds of tests simultaneously.

Getting through tough times together

Although AI technology has attracted a lot of attention and domestic manufacturers have also shown their image, this GDC is still not as exciting and enjoyable as it seems. Every year, GDC conducts an industry survey before it is officially launched, issuing questionnaires to developers in various fields and types of work to describe the overall status of the industry. In January this year, a questionnaire was distributed to 3,000 practitioners, and practitioners' concerns about layoffs and the ethics of generative AI technology dominated the majority.

In 2023, more than 9,000 people will be laid off in the gaming industry, and 1/3 of those who participated in the questionnaire have been laid off, or have colleagues who have been laid off. Some developers who came to GDC said, “We can't pretend everything is fine during this week in San Francisco.” In order to express and vent these dissatisfaction, a former Epic Games producer and a “Fortnite” designer organized a “Collective Primal Scream” in the park opposite the Moscone Center where GDC was held – this is A flash mob event that supports game practitioners to vent their anger and powerlessness towards the environment in the form of screaming and shouting.

GDC organizers apparently noticed this, too. Before the exhibition started, they expected that there would be many unemployed developers who wanted to come looking for jobs. Therefore, the organizers provided low-price tickets for the unemployed and set up a bulletin board for potential employers and job seekers. In terms of venue layout, GDC has also added many areas for discussion, allowing people to get to know each other while resting, or to talk while walking.

The hidden dangers brought about by the development of AI have also attracted the attention of some industry insiders. A GDC survey in January found that 84% of game practitioners surveyed were “somewhat worried” or “very worried” about the ethical issues of generative AI. When the rapid development of generative AI is paralleled by large-scale layoffs in the industry in 2023, many practitioners feel the pressure, especially screenwriters, artists, and voice actors.

To this end, many industry union organizers also appeared at this GDC, and there was a discussion group called “Now it's time” to encourage practitioners to establish or join unions and unite to maintain own legitimate rights and interests. The International Theatrical and Stage Employees Alliance, which mainly focuses on the rights of voice actors, is one of the more active organizations, and Microsoft has recognized its union of 600 Activision members.

Practitioners try to form trade unions to safeguard their rights and interests

Studios of different sizes have different applications of generative AI. According to the GDC survey, 51% of respondents said that their company has some policy on generative AI, of which 2% mandate its use and 12% completely prohibit its use. Triple-A vendors are quicker than independent studios to establish rules regarding the use of generative AI, with 21% of triple-A vendors and 9% of independent studios banning its use entirely. Most of the participants who came to the scene were more interested in AI tools than the content actually generated by AI.

Conclusion

According to statistics, there are 30,000 game developers from all over the world attending GDC 2024. This is indeed a rare gathering.

The different faces among the bustling crowd represent the special period in which the industry is living: the dawn of excitement brought about by cutting-edge technology and the depressing overall shrinkage coexist. If the generative AI tools demonstrated by manufacturers can be smoothly applied to the development process, it will indeed greatly improve efficiency and free some people from boring repetitive work; at the same time, whether it is because of AI or the overall industry Developers who have been affected by the environment and manufacturer decisions are looking for solutions to their own problems.

But one thing is for sure: people want change. The development of technology and communication are ultimately for the well-being of practitioners and the possibility of creating better works. In fact, this is what people who attend GDC and those who follow GDC every year hope for: to meet an opportunity, a colleague or a friend, and to have a change.

Advertisement