How does the 'national game' use AI?

Candy Crush Saga, a game with hundreds of millions of users. Photo: Microsoft .

Players who are working their way through more than 18,700 levels in Candy Crush Saga might be surprised to learn that they are solving puzzles designed with the help of artificial intelligence. The game's developer is using AI to improve the experience, serving a large number of conquering people.

Todd Green, general manager of the Candy Crush brand, said that using AI in this way saves time so they can focus on creating new levels. It would be difficult for designers to manually update and adjust the more than 18,000 levels without initial support from AI, he said.

The use of AI in the video game industry has also received mixed reviews. In contrast to Green, some voices, as well as those directly involved in the development of games, have argued that the technology could threaten their livelihoods. This led to the SAG-AFTRA strike in late July 2024.

“We are not putting chatbots in the game, nor integrating AI-powered design experiences for players to use directly,” Green said, while affirming that this technology is not intended to replace human resources in the gaming industry. Instead, he shared, AI will solve existing problems, speed up efficiency and accuracy in work.

In the US, spending on video game content is set to grow to $51.3 billion by 2024, according to data from the Entertainment Software Association (ESA). Of that, consumers will spend half on mobile games, which are also the most popular form of gaming among users aged 8 and older.

A typical chess game. Photo: Candy Crush/YouTube.

Since its debut on Facebook in 2012, Candy Crush has been constantly updated, recently releasing its 300th client version. Gaming giant Activision Blizzard, now owned by Microsoft, acquired game publisher King in 2016 for $5.9 billion .

Candy Crush is in a unique position, says Joost Van Dreunen, author of a book on the economics of the video game industry. The free app is more than a decade old, has millions of players, and serves a “content-hungry audience.” Demand is so high that it makes sense to use AI to offload the workload, he says.

What’s special about Candy Crush is that each level is technically just a single board. “With AI and a database of previously created levels, it makes sense to speed up and expand the creation process to create more content,” Dreunen said. This will create more levels for users.

King uses AI to focus on two distinct areas. The company wants to make sure that new boards are fun to play from the first time. Meanwhile, old boards, including those from years ago, are still worth playing. The AI plays a technical role behind the scenes, not creating new content.

Green shared that because there weren’t many people to test the game and give feedback, they had to get it right from the start. Additionally, the company focused on a group of customers who had played the game before, maybe took a break, then came back out of curiosity for something new.

The CEO said the company also regularly talks to players. By observing how they react to each level, the rate of clearing the level or the frequency of shuffling the board, the design team will have a reasonable optimization process.

Using AI means that instead of improving just a few hundred, the team can improve thousands of levels per week. In part, this is because they can automate the process of drafting improved versions.

The levels are not arranged in order of difficulty. An easy level might appear after several hard levels (or vice versa) to create a sense of variety. “To some extent, this is obviously subjective. It will feel different to everyone,” Green says.

Source: https://znews.vn/tua-game-quoc-dan-dung-ai-nhu-the-nao-post1552966.html