Bio-Inspired AI Architecture 'Neuraxon' Unveiled
The firm QUBIC has introduced Neuraxon, a new bio-inspired AI architecture. The system uses trinary logic, as opposed to binary, to create evolving neural systems. This approach is being explored for applications in adaptive design, relevant to fields like architecture and urban systems.
- The founder of QUBIC, Sergey Ivancheglo (also known as Come-from-Beyond), has a history of pioneering blockchain technologies, including creating the first functional Proof-of-Stake concept with NXT and co-founding IOTA, which utilized a Directed Acyclic Graph (DAG) structure. His work on Qubic extends to a "Useful Proof-of-Work" model, where mining power is repurposed for AI training. - Trinary logic, which uses three states (-1, 0, +1) instead of binary's two, is not a new concept; the first modern ternary computer, "Setun," was developed in the Soviet Union in 1958. The core advantage of a ternary digit (trit) is its higher data density, carrying approximately 1.585 bits of information compared to a single bit. - Recent research into ternary computing has focused on using materials like carbon nanotubes to create reliable three-state logic gates and even ternary neural networks (TNNs), which can reduce computational complexity and memory requirements in AI applications. - The concept of "adaptive design" in architecture involves creating buildings that can dynamically respond to environmental conditions and user needs. AI platforms like Autodesk's Forma and Sidewalk Labs' Delve are used in urban planning to run thousands of simulations for factors like daylight, noise, and wind, optimizing designs before construction begins. - The discussion around human-AI collaboration is shifting from viewing AI as a tool to a "co-creator," raising questions of authorship and agency. Art collectives like teamLab use AI to create dynamic digital artworks that respond to and interact with the audience in real-time, blurring the lines between artist, viewer, and the work itself. - For developers building creative AI tools, a multi-tool workflow is common, where practitioners chain different AI models together—for example, using a large language model for concept generation, an image model like Midjourney for visualization, and then a video model like Runway to add motion. - To build these complex workflows, developers are increasingly turning to AI-native IDEs and CLI tools that integrate AI directly into the coding process. Popular tools in this space include Cursor, an AI-first code editor, and command-line agents like Aider and Claude Code, which can modify and commit code based on natural language prompts.
