VERSES AI Inc. announced the filing of a patent application covering a milestone invention for automating the generation of intelligent software agents directly from data sets that can interact with both software and hardware systems, such as robots, drones, sensors, and actuators. Generative AI and Large Language Models (LLM), such as OpenAIs GPT, Googles BARD and Metas LLAMA,cel at creating content based on patterns inferred from their training data. However, their comprehension of underlying data remains rudimentary, mimicking rather than understanding, and they lack the ability to incorporate to new information post- training.

This can produce inaccurate, unbiased, and potentially harmful responses which have resulted in calls for global AI regulation to ensure that AI can be aligned with human values and goals. With the goal of developing human-centered intelligent agents, VERSES has employed a neuromorphic approach to its AI research and software development based on neuroscience research known as Active Inference which simulates the brain's processes for learning and problem-solving. VERSES AI Inc.'s new invention is designed to streamline and automate the creation of 'intelligent agents' or digital taskmasters.

The process begins by creating a structured representation of the world, known as an HSML graph, and transforming it into a blueprint for how the agent should behave. This not only brings a new level of intelligence and adaptability to smart systems but believe also represents a significant step forward in their scalability. Building on this, the next phase of the process tunes the agent to perform within a specific context, such as operating in a drone or vehicle or as a personal assistant or managing a smart home, warehouse, or manufacturing facility.

By tailoring the agent to the particular context, task and hardware, the process seeks to create a seamless and effective operation. These two aspects of VERSES new intervention work together with the aim of creating a more advanced, adaptable, and effective AI system. The result is expected to be a new class of Agents that have the capacity to evaluate and consider their responses prior to making them, and to assess the context of a scenario before determining an appropriate course of action.

In essence, to enable AI to be able to "think" before they speak or act and to constantly learn and update their understanding of users, and the domain they operate in. This differs from ChatGPT and related models that are currently on the market, which cannot weigh decisions or update themselves in light of new information. Until recently, the adaptive benefits of Active Inference agents were tempered by the fact that they relied on labor-intensive, hand-crafted methods to encode the generative 'world' model they use for reasoning and decision-making.

This limitation restricted the widespread adoption of this approach to AI due to the scalability challenge it presented for real-world deployment. With the recent breakthroughs at VERSES believe this limitation will be overcome, paving the way for the automated creation and adoption of a new generation of adaptive intelligent agents. After years of advanced research and development, VERSES, in a recent series of technical breakthroughs, has successfully overcome this limitation enabling the automated generation of Agents from small data sets.

The development of this technology was led by VERSES Chief Scientist, Professor Karl Friston and R&D and engineering teams who have developed a transformative method for automatically generating intelligent software agents directly from domain-specific data that utilize HSML (Hyperspace Modeling Language), an explicit knowledge modeling language currently being developed into the P2874 IEEE standard that enables the translation of any multimodal data set (text, image, audio, sensor data) into a generative "world" model upon which an Agent can reason and act on. The invention not only establishes a foundation for the automated creation of VERSES intelligent agents but also lays the groundwork for "guidance systems" for other AI. These systems are expected to contribute to safer and more efficient operation, better user alignment, predictive accuracy and the potential regulatory compliance of Large Language Models (LLMs) and other foundational models for text, audio, video, IoT devices, cameras, autonomous vehicles and robotics.

This patent application reflects the process of this standards-based AI agent generation and forms the basis for VERSES' ongoing advancements in the AI space. The technology is intended to be demonstrated in the Company's upcoming version of KOSM OS, an operating system for generating and running agents in the cloud or on devices or robotics systems as well as in GIA, their general intelligent agent personal service, expected to be released later this year. The issuance of the patent is subject to receipt of approval from the USPTO.

In all industries including, but not limited to, manufacturing, logistics, healthcare and education, the Company's advancement is intended to create more efficient and effective AI solutions through the advent of increasingly intelligent, adaptable, and autonomous systems. The patent application is the latest in a new series related to VERSES' recent technological advancements in the AI landscape.