VERSES AI Inc. (CBOE: VERS) (OTCQB: VRSSF) is adding the final touches to a groundbreaking new research paper with major implications for the AI sector. It introduces an efficient new alternative to current machine learning methods being widely used by technology developers.
Chief Scientist Dr. Karl Friston and a team of researchers will be publishing the paper by the end of this week on arxiv.org. The document is titled “From pixels to planning: scale-free active inference.”
It details the advantages of using Renormalizing Generative Models (RGMs) and their method of learning instead of conventional mechanisms like deep learning, reinforcement learning and generative AI.
“RGMs are more than an evolution,” VERSES chief executive Gabriel René said, “they’re a fundamental shift in how we think about building intelligent systems from first principles that can model space and time dimensions like we do.”
One method to rule them all, René says
The “active inference” in the paper’s title refers to a neuroscience and physics inspired AI framework used for identifying the ways in which biological systems, like the human brain, consistently refine their predictions based on sensory input. They do this with the intention of becoming more and more accurate.
René and the VERSES team’s flagship Genius AI platform was inspired by these scientific fields and principles.
“Within Genius, developers will be able to create a variety of composable RGM agents with diverse skills that can be fitted to any sized problem space,” Chief Product Officer Hari Thiruvengada said, “from a single room to an entire supply network, all from a single architecture.”
The science behind active inference has been solidified and is considered to be a valuable potential alternative to some of the most advanced AI systems. But unfortunately, there has been no viable means or pathway to scalable commercial solutions. However, with the introduction of RGMs and their adjustable “scale-free” technique, the company says those days will soon be a thing of the past.
“This could be the one method to rule them all!” René said. “With it, we can design multimodal agents that can not only recognize objects, sounds and activities but can also plan and make complex decisions based on that real world understanding from the same underlying model.”
Read more: VERSES AI levels up with global standards for intelligent system interoperability
Read more: Illustrious market research firm Gartner recognizes VERSES AI for its innovation
RGMs will reduce AI development costs, expand capabilities
This soon to be published paper will detail how RGMs using active inference are capable of performing crucial tasks like other AI platforms. These include image classification, object identification, natural language processing, file compression and content generation. Except, RGMs do it even better.
These models are a versatile and universal architecture, VERSES says. They can be customized for performing the same tasks that AI programs today perform, but with far superior efficiency.
“The inference process itself can be cast as selecting [the right] actions that minimize the energy cost for an optimal outcome,” Friston described.
Moreover, it explains how RGMs achieve a 99.8 per cent rate of accuracy while using 90 per cent less data. The Modified National Institute of Standards and Technology (MNIST) digit recognition task subset measured these statistics. It is a well-known benchmark used in the machine learning sector. VERSES expects other upcoming research papers to further verify these findings on RGMs.
“We’re optimistic that RGMs are a strong contender for replacing deep learning, reinforcement learning, and generative AI,” Thiruvengada said.
Read more: 3/10 generative AI projects will be scrapped before 2026, Gartner says
Read more: Zurich’s Lakera raises millions to protect LLMs from malicious users
Prominent neuroscientist leads the way
The company’s chief scientist explained that instead of relying on conventional datasets and “brute force training” used on different models by others, RGMs expand their knowledge by learning about the underlying structure and concealed causes of their observations.
Lead scientist Friston is one of the most renowned and widely cited neuroscientists on the planet. Notably, he is one of the world’s foremost authorities on brain connectivity research. And now, he is applying his expertise on the brain to help revolutionize the emerging AI sector.
Friston recently discussed the future of the AI industry with New York University professor Gary Marcus in Davos, Switzerland.
VERSES AI is a sponsor of Mugglehead news coverage
rowan@mugglehead.com
