New machine Learning from how the brain models and understands the world is an approach capable of mastering simple videogames with remarkable efficiency.
Axiom is a new AI system that offers an alternative artificial neural network, which dominates modern AI. Axiom was developed by Verse AI and is armed with knowledge on how objects interact in the virtual world. It then uses an algorithm to model how it expects the game to act in response to input, which is updated based on what it observes—a process dubbed active inference.
This approach is inspired by the Free Energy Principle, which is a theory that aims to explain intelligence through principles derived from mathematics, physics and information theory, as well as from biology. Karl Friston developed the free energy theory, which is now used by many scientists. “cognitive computing” Verses for company Verses
Friston, who lives in London, told me via video that this approach could be particularly important when building AI agents. “They have to support the kind of cognition that we see in real brains,” “He said” “That requires a consideration, not just of the ability to learn stuff but actually to learn how you act in the world.”
Deep reinforcement learning is the conventional way to train neural networks to learn to play video games. This involves tweaking parameters in response either to positive or negative feedback. It is possible to create superhuman-like game algorithms, but the process requires lots of testing. Axiom can master various simplified video games like drive, jump, hunt, or bounce using fewer samples and lower computation power.
“The general goals of the approach and some of its key features track with what I see as the most important problems to focus on to get to AGI,” says François Chollet, an AI researcher who developed ARC 3, a benchmark designed to test the capabilities of modern AI algorithms. Chollet also explores novel approaches to machine-learning, using his benchmark as a way to see if models can learn to solve new problems instead of simply copying previous examples.
“The work strikes me as very original, which is great,” He says. “We need more people trying out new ideas away from the beaten path of large language models and reasoning language models.”
Artificial neural networks are the basis of modern AI. They’re based on the same wiring as the human brain, but they work fundamentally differently. In the past decade or so, computers have been able to do a wide range of things using neural networks. This includes transcribe voice, identify faces, and create images. Deep learning is responsible for the latest generation of large language models, which are used to power increasingly complex and garrulous computers. chatbots.
Theoretically, Axiom promises to be a faster way of building AI. It might be especially effective for creating agents that need to learn efficiently from experience, says Gabe René, the CEO of Verses. René says one finance company has begun experimenting with the company’s technology as a way of modeling the market. “It is a new architecture for AI agents that can learn in real time and is more accurate, more efficient, and much smaller,” René says. “They are literally designed like a digital brain.”
Ironically, since Axiom is an alternative for modern AI, deep learning and machine learning, it was the British Canadian computer scientist who first influenced the Free Energy Principle. Geoffrey HintonThe award was given to. Turing award Nobel Prize and his pioneering research on deep learning. Friston was Hinton’s colleague at University College London.
If you want to learn more about Friston, and his free energy principle I recommend that you visit the Friston website. this 2018 WIRED feature article. Friston’s work has also had an influence on the development of an exciting new theory of consciousnessIn a review of a 2021 book by WIRED, the author describes.
