The Creators of Baldur's Gate 3 Details Its Application of Machine Learning for Next Divinity Game

The studio behind acclaimed RPGs like Baldur's Gate 3 and Divinity: Original Sin recently shown its new project, generating immense anticipation within the player base. However, recent remarks from the company's lead designer have brought nuance to the narrative, focusing on the developer's philosophy toward AI tools.

A Tool for Ideation, Not Replacement

In a new message, Larian's director explained that the developer is using machine learning for specific supporting functions. These include fleshing out presentation materials, producing initial artistic references, and writing placeholder text.

Importantly, Vincke stressed that the end material in the game will be crafted exclusively by human writers. "Larian is creating all the content manually," he stated.

Our studio is continuously expanding our team of writers and are actively assembling writing teams.

Given that this area is being specifically called out — we currently have over twenty artistic staff and have job openings for further artists.

Each initiative we do is incremental and aimed at letting our team spend additional energy on actual creation.

Any machine learning application used well is supplementary to a creative team workflow, never a stand-in for their skill.

Addressing Concerns and Clarifying the Vision

The admission of employing this technology initially generated concern among portions of the fanbase. In reaction, Vincke provided further elaboration on social media.

"At Larian, we employ machine learning to research ideas, in the same way we use the internet and reference books," he stated. "In the conceptual ideation stages we use it as a basic framework for composition which we then substitute with original concept art."

He added, "Our studio recruits creatives for their unique talent, not for their capacity to execute what a AI generates."

Three Pillars of Practical Application

Vincke had in the past broken down the company's practical method to this technology, defining its use into key areas:

• Handling Monotonous Jobs: Areas like polishing mocap data, audio processing, and pipeline-specific tasks like adapting animations for different models.
• Fast-Tracked Experimentation: Using systems to speedily create rough versions of gameplay ideas to test concepts before complete implementation.
• Long-Term Aspirations: Exploring how machine learning could in the future facilitate emergent gameplay, especially in managing unforeseen permutations in a detailed game universe.

He explicitly noted that core creative disciplines — like music composition — are not areas where the company is cutting human talent. Conversely, Larian is actively hiring in these very positions.

"Our studio is not shipping a game with any AI components, and we are certainly not planning on reducing creatives to replace them with artificial intelligence," Vincke summarized.