End of ‘black box’ AI? Scientists develop blueprint for transparent systems that learn and remember

New Approach to Transparent AI: Loughborough University researchers have developed a new framework for creating AI systems that can transparently explain their learning process, memory, and decision-making, aiming to replace “black box” AI.
Prototype Development: The team has created a prototype system with both a “brain” and memory, which uniquely allows continuous learning without losing previous knowledge and avoids forming false memories, mimicking human-like thinking.
Mathematical Concept: The foundation of the new approach relies on a “plastic vector field,” a mathematical concept that models changes in information over time, enabling clear tracking of the AI’s cognitive processes.
Overcoming Design Limitations: The researchers believe current artificial neural networks’ limitations arise from their design, which hinders control over information learning and storage, making explainability challenging.
Scalability and Future Applications: While the prototype is still basic, developers intend to scale it up and explore applications in various fields, striving to create trustworthy AI for practical, everyday use in areas like healthcare and decision-making systems.
Research Publication: The findings have been published in Physica D: Nonlinear Phenomena, titled “Designing explainable cognitive systems and explaining neural networks with plastic dynamical systems.” A detailed version of the research paper is available online.

End of ‘black box’ AI? Scientists develop blueprint for transparent systems that learn and remember | Media Centre