Axiom: A New AI Approach Inspired by Human Brain's Learning Mechanism

A new AI system called Axiom uses a deep learning alternative inspired by the human brain's modeling of the world to efficiently master simple video games. Axiom is equipped with prior knowledge about physical interactions and uses active inference to update its model based on observations. This approach draws from the free energy principle, a theory that explains intelligence using principles from math, physics, and biology. Axiom outperforms conventional deep reinforcement learning in various simplified video games with fewer examples and less computation power.

A new AI system called Axiom, developed by Verse AI, is making waves in the field of artificial intelligence by offering a novel approach to game-playing. Inspired by the human brain's modeling of the world, Axiom uses a deep learning alternative that outperforms conventional methods in various simplified video games [2].

Axiom is equipped with prior knowledge about physical interactions in the game world. It employs an algorithm that models its expectations of how the game will respond to input, updating its model based on observations—a process known as active inference. This approach draws from the free energy principle, a theory that seeks to explain intelligence using principles from math, physics, and biology [2].

Unlike traditional deep reinforcement learning, which requires extensive experimentation and computational resources, Axiom masters various simplified versions of popular video games like drive, bounce, hunt, and jump using far fewer examples and less computation power [2]. This efficiency is particularly notable in the context of large-scale AI applications, where computational resources are often limited.

The free energy principle, originally influenced by the work of British Canadian computer scientist Geoffrey Hinton, was further developed by Karl Friston, a renowned neuroscientist and chief scientist at Verses. Friston's work emphasizes the importance of considering not just learning but also the ability to act in the world [2].

Some industry experts view Axiom as a promising alternative to current AI systems, potentially leading to more efficient and accurate AI agents. Gabe René, CEO of Verses, believes that Axiom could revolutionize the way AI learns and interacts with the world, potentially even in finance and other industries [2].

While Axiom's potential is promising, it remains to be seen how widely it will be adopted and how it will impact the broader AI landscape. As AI continues to evolve, innovations like Axiom could pave the way for more efficient and intelligent systems.

References:
[1] https://www.usnews.com/news/business/articles/2025-06-10/video-game-performers-on-strike-for-almost-a-year-over-ai-issues-reach-a-tentative-deal
[2] https://www.wired.com/story/a-deep-learning-alternative-can-help-ai-agents-gameplay-the-real-world/
[3] https://www.facebook.com/groups/DeepNetGroup/posts/2498388973887303/