Every year at ChinaJoy, we will encounter some cutting-edge technology experience projects, in the comprehensive exhibition hall full of IP, these booths that seem to be full of novelty are often more popular.
This year, Century Huatong's GTest extreme test information booth lined up in a long line, the scene provided a trial of the brainwave detector, just stick a headband-shaped sensor behind the ear, play any game on any device, you can test the various emotional characteristics that players are excited in the face of different scenes in the game, and the background algorithm then accurately quantifies the attention mobilization of each element in the game to the human brain.
The reporter of the game think tank was fortunate to meet Mr. Wang Jun, CEO of Jice Information Technology Co., Ltd., and took us to the behind-the-scenes team that has been deeply engaged in game QA, content review, and information support for more than ten years.
The Next Generation of QA: From "+AI" to "AI+"
The main business of Jidei Information is to provide developers with comprehensive quality management, product testing and evaluation, including game content function testing, compatibility and adaptation testing, client-server performance testing, network environment testing, AI automation testing, and special political review testing of game content.
In previous test projects, the industry generally used a manual + script solution. For example, if there is a server rollback problem caused by prop interaction in the game, due to the uncertainty of the operation behavior of possible BUGS, it needs to be troubled one by one in tens of thousands of prop combination interactions, which is time-consuming, laborious and boring, and prone to new mistakes. QA people also write scripts to do some repetitive work, but the versatility of script intelligence is limited and relies on the subjective experience of the programmer.
At present, the technology of Jice in the field of AI has been implemented, and the intelligent algorithm can complete the batch investigation of "incurable diseases" quickly and efficiently by virtue of the advantage of computing power. As an AI company specializing in video games, the challenges of self-developed algorithms are also different. Aside from the mature graphics, audio and video content recognition in the market, video games are a carrier with interactive elements, and players' degrees of freedom, input methods, and gameplay logic are very different in different games.
The solution provided by the company is to intervene in the testing of AI tools in the R&D stage, and troubleshoot the possible bugs caused by new functions in real time, so as to avoid most of the early problems.
Taking a medium-sized mobile game as an example, the operation department needs to launch, develop, and prepare each version, and if the update frequency is limited to one month, the previous "manual + AI" method will take a week to conduct preliminary tests on each version, not including feedback and rework after launch, and the number of projects that the QA team can cover at the same time is very tight.
Now, after switching to an AI-based pipeline solution, the test cycle can be shortened to 2-3 days, and the number of problems can be reduced, which can gain more time redundancy for game manufacturers and save a lot of marketing and communication costs.
Compliance review: Everything can be quantified, just look at whether the data is big enough
In the business list of extreme testing, we also saw an intriguing service, AI game compliance review and anti-addiction review for minors. Although the current version number approval has entered a stable period, there is still the possibility of changing back to macro-control methods at any time, and the use of AI algorithms to help pass the review is somewhat "leaking the heavenly opportunity".
Wang Jun said that the purpose of using algorithms is certainly not to test the so-called review red line, and the requirements of the higher-level authorities for the game industry are clear.
When faced with specific problems, AI is fundamentally different from the logic of human thinking, it will derive the probability of an event based on the existing knowledge base, and finally the algorithm and AI operator will analyze the solution. This knowledge base includes:
1. Guidance requirements of the competent department
2. Feedback accumulated from the running project
Through the political and legal issues accumulated in the market process, relevant cases are formed, and with the continuous training and improvement of AI models, the previous manual and AI testing experience can be digitally and standardized. As a sample of big data, whether it is the accumulation of hundreds of projects such as Century Huatong and Shengqu Games, or providing services for more game manufacturers in the whole industry, this algorithm system will continue to be updated and iterated. Small and medium-sized development teams struggle to allocate time and energy to deal with related issues, and traditional manual testing solutions mean inappropriate commission costs.
In the reliability test of anti-addiction protection for minors, AI can also play a role in simplifying the complex, and automatically maintain the anti-addiction function after the game version is updated. When asked whether the age and behavioral characteristics of players will be able to be identified in the future, Wang Jun believes that this will seriously involve the privacy of users, and the anti-addiction system must be connected to the identity authentication system provided by the public security department.
Brain Science and Artificial Intelligence: Cyberpunk in both directions
Then we went back to the booth and there were two test devices for the GTest for players, which simply means that it knows what the testers are "thinking".
According to the readings of the brainwave sensing patch, the monitoring tablet shows the player's "flow level", which indicates the degree of perception of the game content, and then distinguishes a number of emotional characteristics such as "pleasure", "confusion", "frustration", and "immersion". The author first used the MMO mobile game "Celebrating More Than Years" provided on the spot for testing, and the "immersion" can appear in the cutscene stage, and then in the copy puzzle solving link, in the face of the door opening mechanism that cannot be found, you can also get the reading of "confused", and the second time, the author used a music beat game test on the portable device, and when the music enters the climax and the combo is interrupted, it can be recorded on the timeline.
Those who are familiar with game development will surely realize that the future of this technology has gone far beyond the scope of game QA inspection, and it will even initiate a change in the entire development thinking. When designing a scene, should the textures be red or green? Should the length of the game tutorial be 40 seconds or a minute? These design details, which used to be metaphysics, may become a kind of manifestation in the future.
The developer of the machine, Associate Professor Chen Liang of the Business School of East China University of Science and Technology, is looking forward to the positive effect of this invention to explore the arousal mechanism of the human brain, such as the problem of children's lack of concentration, perhaps through the use of game behavior to mobilize specific emotional areas and enhance the brain's interest. Further, it is even possible to develop game mechanics that are both fun and not easily addictive.
Through the research of brain science and the analysis of the computing power of artificial intelligence, the two-way rush between the human brain and the computer may be able to open a breakthrough in game behavior and finally uncover the mystery of consciousness.
End
In the context of "reducing costs and increasing efficiency" in the industry, every progress in AI is affecting the nervous nerves of practitioners, substitution, or mastery, or division and integration. Wang Jun took the development of domestic games as an example, saying that there were only imitation "Legend" type games in China at the earliest, and he also entered Shanda Games 20 years ago by chance, and later various forms of domestic games were born, such as 3D, shooting, MOBA, mobile games, cards, chicken, and even "Black Myth: Wukong", which is about to be released...... But the old games haven't completely disappeared, they continue to come to life.
The game industry in the future is diversified, all kinds of products will have their own space for existence, and there may be explosive models every year, but there will be no more giants in the incremental era, so the development technology of games is the same, and various development methods are not mutually eliminating.
"If you take it as a joke, I believe that AI will also be addicted to games in the future, but the games they like must be games developed by us (humans)"