A team of former Google DeepMind researchers has launched an innovative AI behavior engine designed to revolutionize video gaming by enhancing the interactions and behaviors of non-playable characters (NPCs). Based in Canada, Artificial Agency has emerged from stealth mode with $16 million in funding and aims to distinguish itself in the rapidly advancing world of AI-driven gaming.
Traditionally, NPCs are limited by decision trees and pre-written scripts, which restrict player experiences and often lead to repetitive and unengaging dialogues. Artificial Agency's behavior engine breaks this mold, empowering developers to act as stage directors. By assigning each NPC specific motivations, rules, and goals, the engine enables dynamic responses to player actions. This technology can integrate with existing games or form the foundation of entirely new gaming experiences.
Based in Edmonton, Alberta, Artificial Agency enters a competitive arena featuring notable competitors like Inworld, which also focuses on AI-generated NPC behaviors, and Nvidia, which has been developing AI-powered NPCs for quite some time.
Brian Tanner, co-founder and CEO of Artificial Agency, believes that the incorporation of AI-generated NPCs is the future of game design. “The conversations we have with gaming studios are not about if, but when,” he said. “Dynamic interactions and responses enabled by our system will soon become standard in the gaming industry.”
Recently, the startup secured $12 million in a seed funding round co-led by Radical Ventures and Toyota Ventures, increasing its total funding to $16 million after an earlier $4 million pre-seed round from Radical Ventures. Other participants in this seed round included Flying Fish, Kaya, BDC Deep Tech, and TIRTA Ventures.
The Adoption of AI NPCs
A key concern for startups like Artificial Agency is whether gaming studios will embrace their AI technologies. Some fear that major studios may choose to develop their own solutions or exhibit reluctance to integrate generative AI into their flagship games due to the potential risks of hallucinations and the technology's untested nature.
While keeping specifics confidential, Artificial Agency claims to collaborate with “several notable AAA studios” to refine its behavior engine, with plans to bring the technology to a broader audience by 2025.
Daniel Mulet from Radical Ventures noted that many gaming studios initially attempted to create these behaviors independently, diverting focus from actual game development. He asserts, “Seeing numerous companies trying to build this themselves reveals a significant opportunity to provide a universal platform.”
Overall, game developers show increasing interest in generative AI, though some reservations persist. A recent survey of 3,000 game developers by GDC and Game Developer for the 2024 State of the Game Industry report indicated that nearly half currently utilize generative AI for various tasks, especially those deemed repetitive. However, only about 21% believe generative AI will positively impact the industry, with 42% expressing strong concerns regarding its ethical implications.
Mulet credits the founding team’s extensive experience with Google DeepMind as a point of confidence for creating top-notch tools to enhance NPC behavior. DeepMind has a proven track record of pioneering advancements in AI, such as developing AlphaGo, the first computer program to defeat a world champion at Go.
While Google shifted its focus to the Gemini model, Tanner and his team pursued developing game agents capable of replacing traditional NPCs.
Revolutionizing NPC Interactions
In a demonstration, co-founder Alex Kearney showcased an NPC powered by their behavior engine in Minecraft (the startup did not disclose other games in development). The NPC, named Aaron, was designed to be friendly and resourceful. Offered basic functions like moving and interacting with objects, Aaron was asked by Kearney's in-game character to fetch supplies for a mining adventure.
Mirroring a real ally, Aaron sought out armor, tools, and food from various chests—actions it was not explicitly programmed to perform. After Kearney mentioned being gluten-free, Aaron apologized and provided a suitable meal of cooked chicken. This demo illustrated how Artificial Agency's AI NPCs can engage in meaningful dialogues and execute complex initiatives without the need for detailed scripting, thus potentially saving time for developers.
The Financial Implications of AI NPC Costs
Tanner estimated that this five-minute demo incurred only $1 in AI inference costs, noting that a year ago, such expenses could reach up to $100. He anticipates continued cost reduction due to advancements in GPU efficiency and AI model optimizations. The startup currently employs open-source models, including Meta’s Llama 3, and Tanner predicts that one year from now, that same demo could cost a mere penny or less.
The critical question remains: who will bear the costs associated with AI inference? Although Artificial Agency believes that integrating AI NPCs will not elevate end-user game prices, investor Mulet expressed uncertainty. He indicated that while game studios are willing to license Artificial Agency's technology, implementing it could lead to additional monthly fees for gamers.
“The cost of inference for these systems means that it may be positioned as a premium feature,” Mulet elaborated. “Will gamers be open to paying $2.99 a month or $12.99 a month? That’s still uncertain.”
Ranking
