Blizzard: AI and Games

Blizzrd, AI and games? Sounds like a match, but exactly how?

As a long-time lover of the game World of Warcraft (WOW) since middle school, I am often amazed by how intelligent some of the Non-Player Characters (NPC) are in response to my actions and how creative and dedicated Blizzard (the company that created the game, named Activision Blizzard and was acquired by Microsoft on Oct. 13th, 2023) is in pushing out a massive amount of game contents through regular updates. These would make me question how they do it and whether AI plays a role in any part. The simple question is: they did not use AI from the beginning, yet they are trying to implement it in all aspects of creating processes now, and we are not just talking about WOW.

A bit intro to Blizzard

As one of the most significant and most impactful video game companies in the world, Blizzard has made many top-selling games throughout its history, including the Warcraft series, World of Warcraft, Diablo series, StarCraft series, Overwatch series, Hearthstone, and a lot more [1]. Blizzard was initially merged with Activision in 2008, the largest merger in the video game industry back then [2]. It had a revenue of 7.53 billion dollars in 2022 and is hiring around 17,000 employees [3]. Simply put, Blizzard is a legend (not biased at all!).

Art creation accelerated with AI

Online multi-player video games require teams of talented artists, designers, and game creators and constant updates to the game content so that the games would attract players to stay and always have or anticipate new things they can explore. This means that the game-creating process is rather highly labor-intensive. In the era of 2022-2023, with the advances of generative AI tools in various forms, such as Stable Diffusion, Midjourney, and ChatGPT, one can easily imagine the utility of this software within game creation. In fact, according to a New York Times article, Blizzard announced an internal AI tool for art generation called Blizzard Diffussion, which is fine-tuned based on Stable Diffussion [4]. Blizzard diffusion could generate art in different scenes without games, including the landscape’s looks, buildings, characters, gears, and even the stories behind them. Although Blizzard did not disclose any details on how exactly they were or were planning on using AI, it was made clear by its vice president of global insights, Andrew Guerrero, that they would use AI to accelerate their creative work, making the process more accessible [4].

As an interesting note, here is a picture generated by Stable Diffusion Web when I entered the prompt: “A new world of Warcraft character who is a male troll, a legendary hunter with his wolf pet alongside, greeting the player” [5]. AI’s performance here is sub-optimal, yet it has high potential.

AI-driven NPC

Though AI cannot replace humans in the real world, it can do so in video games as non-player characters. In games like World of Warcraft, players constantly interact with NPC for actions such as delivering quests, purchasing or selling goods, fighting together or against, delivering real-time story scenes, etc. Some of these actions may be done with straightforward algorithms, but many could only be achieved with AI, given the dynamic nature of challenging the players [4, 7]. Online gaming companies like Blizzard regularly use AI to detect robots, hackers, and players’ toxic behaviors to maintain a healthy community [6]. In my experience in the World of Warcraft, many bosses can employ targeted spells onto individual classes of players, making the game a lot more challenging and fun.

I would also like to give a special shout-out to AI in StarCraft II, a strategic competitive game where one to three players control units to conquer the other team of one to three. Aside from AI usage during artistic content creation, the game utilizes particular types of AI to act as the opposing robotic player against the human player for training purposes. The AI needs to respond and adjust its strategies dynamically to the actions it observes from the human player. There are many difficulties one can change for how expert the AI would be and how hard it is to beat the AI. The use of AI here is very similar to the usage of AlphaGo by DeepMind in Go matches, employing reinforcement learning to optimize every action for the best potential outcomes of winning the game. Interestingly, DeepMind also participated in StarCraft II with the AlphaStar AI, which could outperform 99.8% of all registered human players [8]. There is also a community behind building AI to play the game with a ranking system specifically for AI. This topic of would ultimately contirbute to the debate of human vs. AI.

As we can observe, AI has been used extensively by either Blizzard or the players to optimize player experiences as the purpose.

Value capture with AI

Blizzard’s ultimate goal to employ all these forms of AI is to keep their existing players and attract new players to play their games. But how exactly does Blizzard capture the values created by AI using AI? Although few data points are available to us, I will assume their AI-driven business model practices for pricing optimization, customer behavior analysis, and future trend predictions. As a video game company, Blizzard can record all user data points, including time, clicks, favorite characters or elements, etc. A big-data-driven approach with AI would allow Blizzard to understand their customers and provide targeted marketing and product deliveries to capture values.

Challenges of AI and potential solutions

Although AI tools may sound promising in this case, concerns about copyright abuse, job displacement, and hallucination were raised. They could only be justified as the field experience and mature under regulations. Until then, the usage of AI at Blizzard was very much up to the individual designers. Given the nature of generative AI, I doubt any current copyright laws would prevent such acts from happening. A needed change to Blizzard is a thorough guideline on AI usage within the company. Such guidelines require an organizational change of Blizzard with the incorporation of an AI management team that controls the future ways of using AI within the cooperation. Tremendous opportunities lie within AI usage, and a management team dedicated to this field would allow Blizzard to construct strategic and organized plans to capture values.











Woebot: The Robo-Therapist Supplying Comfort, Consolation, and Care at Scale


Relume – Web Designers Get A Massive Productivity Boost in the No-Code & Generative AI Era

Student comments on Blizzard: AI and Games

  1. Very interesting post, Sam! It is fascinating how Blizzard is enhancing player experiences and keeping their massive gaming community engaged. I’m not a video game person but the art caught my attention. I’m curious how can game companies like Blizzard use AI creatively while making sure they respect intellectual property. Have they taken any proactive measures in this direction so far?

  2. This is a cool post, Sam! The question I have for you is whether the potential for AI in gaming can be unlocked without the use of cloud computing, meaning always online games. With Xbox leaking a hybrid console for 2028, I think it’s interesting to consider whether we’re going to see the real arrival of AI in games when we get next gen consoles.

  3. Nice post! It’s cool how AI is used for NPCs in Blizzard games. I noticed you mentioned AI is not able to replace players in these kind of games. I am curious, what more is needed for AI to actually control player characters, not just NPCs? This could really change how we play games. What do you think?

  4. Thanks for the post! Though I rarely play games, I’m also impressed by some AI-created cartoon characters and synthetic characters in TV series and movies. It is the combination of AI-powered creation and increasing computing power that makes the gaming industry develop so rapidly these years. Besides, I found you mentioned people are now training AI models to win the game. Will it develop quickly in the next few years and negatively affect users’ gaming experience?

  5. I’m reminded a lot of the old single-player Diablo games, which regenerated new dungeons maps, monsters, and layouts every time you restarted the game. Back then, this was extremely revolutionary, since it gave the games near infinite replay value without it getting stale, and surprisingly, it wasn’t particularly buggy (i.e., no monsters in walls, or unplayable areas emerged). I think one super interesting use of AI in games is personal adaptation of single player games. For example, in an exploration game, depending on how well you are doing, games can dynamically adjust the environment to suit your skill level, to ensure you are always adequately challenged. I know many games already do this to a limited degree, but AI can really push this to the frontier, by crafting monsters customized towards your particular weaknesses, or by crafting new worlds on the fly based on your style. As someone who really prefers the value of single player games, I think this could add a component of dynamism that is missing from the modern gaming landscape.

  6. This is a great post Sam, thanks for the detail. I would be curious to understand the implications of Generative AI in gaming as it relates to deepfakes. How real can the AI-generated NPCs get so that they are indistinguishable from humans in the game? This is the ultimate form of the Turing Test.

    Another question is how much this impacts the jobs that game designers do to create the characters in the game? Will this cause job displacement or just up-level the jobs that creatives do at these gaming studios?

Leave a comment