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

The developer behind acclaimed titles like Baldur's Gate 3 and Divinity: Original Sin recently unveiled its next major project, creating significant anticipation within the player base. However, follow-up remarks from the studio's co-founder have introduced clarity to the discussion, addressing the developer's stance toward machine learning.

A Tool for Ideation, Not Replacement

In a latest clarification, Larian's director detailed that the company is employing generative AI for particular supporting functions. These include fleshing out PowerPoint slides, producing early-stage concept art, and drafting draft text.

Crucially, Vincke made clear that the final assets in the game will be created entirely by real creatives. "Larian is creating all the content manually," he said.

Our studio is actively expanding our roster of concept artists and are actively forming narrative groups.

As concept art is being specifically referenced — we right now have over twenty concept artists and have job openings for more creatives.

Each initiative we do is additive and focused on having people spend greater focus on actual creation.

Any machine learning application applied correctly is supplementary to a developer's process, not a replacement for their talent.

Responding to Feedback and Defining the Path

The news of employing this technology at first generated backlash among a segment of the fanbase. In reply, Vincke provided further detail on online platforms.

"We use these tools to explore references, similar to we use search engines and physical media," he explained. "In the initial planning process we use it as a basic framework for structure which we then swap out with hand-crafted concept art."

He continued, "Larian brings on creatives for their unique talent, not for their willingness to execute what a AI generates."

Focused Uses for Machine Learning

Vincke had in the past broken down the team's practical method to AI and ML, defining its use into key areas:

• Handling Monotonous Jobs: This encompasses motion capture cleaning, dialogue cleanup, and pipeline-specific tasks like retargeting animations.
• Accelerated Iteration: Using technology to speedily create rough versions of scenarios to test concepts before expensive development.
• Future Potential for Gameplay: Exploring how machine learning could in the future facilitate emergent gameplay, particularly in creating player-driven narratives in a detailed game universe.

He explicitly stated that central narrative areas — like music composition — are not areas where the team is cutting creative talent. On the contrary, Larian is recruiting more in these exact fields.

"We are neither releasing a game with machine-made assets, nor planning on cutting teams to swap them out with artificial intelligence," Vincke concluded.