Entering the next era of AI

  • * Register (or log in) to the Neural Network to add this session to your agenda or watch the replay

  • Date
    28 September 2026
    Timeframe
    15:30 - 16:30 CEST, Geneva
    Duration
    1 hour

      Artificial Intelligence (AI) is now entering its third phase. The first phase was driven by basic research; the second by scaling where ever-larger foundation models were trained on massive datasets with massive compute. The third phase is now beginning: the industrialization of AI, bringing intelligence into machines and robots that operate in the physical world. Industrial AI demands architectures that are computationally efficient, have a small memory footprint, maintain a persistent state, and support state tracking together with advanced memory mechanisms.  

      This AI for Good session will introduce xLSTM, an architecture designed for this transition. With its novel Memory mechanisms and efficient recurrent computation, xLSTM combines long-range state tracking with linear-time processing and constant memory at inference. TiRex, an xLSTM-based time-series foundation model, demonstrates how this approach delivers efficient predictive intelligence for complex, evolving systems.  

      For robotics, however, prediction alone is not enough. An intelligent robot must maintain a working memory of its current situation, task, and environment, while drawing on an episodic memory of past experiences. Together, these capabilities allow robots to recognize familiar situations, learn from interactions, retrieve relevant experiences, and continuously adapt their behavior. This lays the foundation of truly autonomous industrial intelligence.

      Are you sure you want to remove this speaker?