• He explains how AI is a gamechanger for developers, who are using ChatGPT to produce code along with image generation tools.
• However, he also warns that the current generation of AI systems still don’t have the capacity to manage other complex tasks linked to gameplay and coding.
How are independent video game studios using AI in their day-to-day workflows?
AI has always been present in the world of video games, but it’s important to point out that the term “AI” is now being used in a different way. These days, when we speak of AI, we are usually thinking of models like ChatGPT, but other more primitive forms of AI, like bots that perform random actions, have long been used in the field of video games.
Machine learning has now enabled independent developers working on new games to extensively explore their technical potential, notably by using ChatGPT to generate code for game mechanics, and also to translate games more efficiently.
It is essential to point out that artists are not being fully replaced by AI, but that it helps to simplify their work.
How has AI simplified the creation of content for studios?
In the world of mobile games, we’ve found that AI can simplify the generation of visual content. For example, an artist who draws a safe can then use an AI model like Stable Diffusion to to open the safe and add gold coins inside. This has enabled artists to save time that used to be devoted repetitive tasks, because they no longer need to manually complete every stage of a process. However, it is also essential to point out that artists are not being fully replaced by AI, but that it helps to simplify their work.
Can you tell us about new tools that are available and AI’s potential to boost the creation of artistic assets for games?
New tools like Scenario.com, which launched earlier this year, can easily generate characters and other game assets. However, some artists may be reluctant to use them. Even if they are not, they still need time to learn how they work, so it will take at least a year or even 18 months for these tools to revolutionize their professions. Artists need to learn how to use tools like Stable Diffusion and Midjourney effectively, which takes time. Just as an understanding of code is essential for working with GPT, so too is an artistic vision for designing game assets.
What other limits have you encountered?
AI still doesn’t have the capacity to manage complex tasks linked to gameplay and coding. For example, OpenAI recently tried to integrate functionality for the loading of entire GitHub repositories onto ChatGPT, but this project was halted, and the reasons for that decision remain unclear. The question of personal data has also become a major issue. So there’s still a lot of work to be done before we can take full advantage of AI in our field. Having said that, things can change very quickly: Meta has filed patent applications for the generation of 3D animal assets from real elements. Combined with technologies such as Lidar, we believe that there will soon be very powerful tools for generating realistic content.
Arthur Guérin
