Imagine you're playing a video game where NPCs (non-player characters) not only converse with you but also understand your needs and even accomplish complex tasks without explicit instructions. This isn't science fiction; it's what Canadian startup Artificial Agency is currently working on.
A team of former Google DeepMind researchers has created an AI behavior engine aimed at transforming traditional video games into more dynamic experiences by enhancing NPC behaviors and interactions with players.
Image source note: The image was generated by AI, authorized service provider Midjourney
Artificial Agency has just completed a $16 million funding round and emerged from stealth mode. They believe their behavior engine will set them apart from numerous companies using AI to generate more realistic NPCs.
Traditionally, NPCs are guided by decision trees and pre-written scripts, which often limit the number of outcomes players can experience. For example, most NPCs in games respond to player actions with only a few repetitive dialogues, often appearing unrealistic and dull.
Artificial Agency's behavior engine discards this framework, transforming the role of game developers into more of a stage manager. The engine requires developers to assign each NPC a set of motivations, rules, and goals that determine how the NPC responds to the player. This technology can be integrated into existing video games or serve as the foundation for new ones.
This startup, based in Edmonton, Alberta, is entering an increasingly crowded field. Competitors include Inworld and Nvidia, which have been working in the AI-driven NPC space for some time.
Artificial Agency believes integrating AI-generated NPCs into video game design is the future. "Our conversations with these studios are often not about whether to adopt, but when to adopt," co-founder and CEO Brian Tanner told TechCrunch. "The dynamic interactions and responses our system allows will become the standard in the gaming industry in the coming years."
Artificial Agency recently raised $12 million in a seed round led by Radical Ventures and Toyota Ventures. Prior to this, it received $4 million in an undisclosed pre-seed round from Radical Ventures, bringing the total funding to $16 million.
Tanner estimates the AI inference cost for a five-minute demo to be about $1, but notes that this would have cost $100 a year ago. Artificial Agency expects costs to continue to decline due to improvements in GPU efficiency and AI model optimization. Currently, the startup uses open-source models, including Meta's Llama3.
Mulet says his venture capital firm believes game studios are willing to pay licensing fees for Artificial Agency's technology, but once deployed, it could lead to monthly fees for players.
"The inference costs associated with running these systems mean it must be some sort of premium feature," Mulet says. "As a player, would you pay $2.99 or $12.99 a month? It's too early to say now."