NXAI has successfully demonstrated that their groundbreaking xLSTM (Long Short Term Memory) architecture achieves exceptional performance on AMD Instinct™ GPUs, marking a significant advancement in recurrent neural network technology for edge computing applications.
NXAI unveils its first time series model, TiRex, based on the novel xLSTM architecture – and immediately claims the top spot in well-known international benchmark leaderboards. Despite having just 35 million parameters, TiRex is significantly smaller and more memory-efficient than its competitors. It not only excels in prediction accuracy but is also considerably faster. A decisive advantage lies in the model’s ability to continuously monitor, analyse, and update the system state – known as state tracking. Transformer-based approaches lack this capability. TiRex, on the other hand, can approximate hidden or latent states over time, improving predictive performance. Its architecture offers another major benefit: it is adaptable to hardware and enables embedded AI applications.
Jan Koutnik “Czeched out” in 2009 to improve recurrent neural networks and scale evolutionary reinforcement learning as a postdoc at IDSIA. At NNAISENSE, he directed the division of the company responsible for the “learning to park” NeurIPS 2016 demo with Audi, the “learning to run” NeurIPS competition win in 2017, and the Festo Softhand in-hand manipulation demo at Hannover Messe in 2019.
We take a brief look back with Jan Koutnik at his time at NNAISENSE and the beginnings of Industrial AI. Then we switch to the present. Jan explains his Industrial AI approach, why he doesn’t realize GenAI use cases and where he sees the sweet spot for himself and his new company. Spoiler: It’s all about vision systems.
Industry is different, when it comes to AI. Peter Seeberg talks to Stefan Suwelack about RAG for industry, the role of data and the Data Centric Approach and about the Industrial AI Canvas. Stefan works for Renumics. He and his team work for companies like Rehau, Festool or Polytec.
NVIDIA earns its money mainly with hardware, in the UK the developers of Shadow Robot Company also earn their money with hardware. They are building the robotic hand, which is designed to make AI-based robotics possible. We talk to the CEO Rich Walker.
What are GenAI systems and why do we need them? Prof. Dr. Jakub Tomczak published a sensational article on GenAI Systems a few weeks ago. We immediately realised that he had to explain it to us. His outlook: The idea of using LLMs as a backbone for Operating Systems and agents as applications has attracted a lot of attention. Here, we consider general systems with various GenAI-based components, not only LLM-based compartments. Either way, moving towards GenAI-based (operating) systems seems like the future, and the next step of cloud-based systems. Indeed, GenAISys can be deployed locally, but also in a cloud server; or as a hybrid (e.g., a GeM, a cache storage, and DEs are local but external tools and storage are in a cloud). The last option can be especially appealing for manufacturing since all real-life operations must be executed in real-time while data storage and other operations are carried out by external services (or agents).
