A new hub for Artificial Intelligence (AI) research and product development has been launched in the heart of Europe, laying the foundation for European AI applications on a competitive, global level.
NXAI GmbH was founded in December 2023 by AI researcher Sepp Hochreiter, company builder Netural X GmbH and PIERER Digital Holding, in close collaboration with the Johannes Kepler University (JKU) in Linz and its leading research centre LIT AI Lab at the Linz Institute of Technology. NXAI will forge ahead with cutting-edge technological research into Large Language Models (LLMs), particularly the xLSTM algorithm developed by Hochreiter.
Initial results show that xLSTM works more efficiently (with less computing capacity), faster and, above all, with higher accuracy rates than previous LLMs. In particular, xLSTM understands the semantics of texts better than previous LLMs and can understand and generate complicated texts. In addition, NXAI will be researching world-class algorithmic methods in the field of deep learning, for example to accelerate simulations of complex dynamic systems.
Setting a new standard in AI language processing
AI mastermind Sepp Hochreiter laid the foundations back in 1991 and invented LSTM (Long Short Term Memory) technology. In doing so, he paved the way for AI as we see it today. The technology was the leading method in speech processing and text analysis until 2017 and has been used billions of times in smartphones to date. In 2023, Hochreiter achieved another breakthrough with xLSTM.
Current LLMs are based on so-called transformer models. These require very high computing power during operation for long texts. The transformer calculations therefore increase quadratically with the length of the text. In contrast, the xLSTM calculations only increase linearly with the length of the text and require less computing power during operation. This is a great advantage, as complex tasks require more text for both the task description and the solution. Fortunately, xLSTM makes industrial applications possible for which transformer models are too slow. Like transformer models, xLSTM has a phonetic memory. However, the algorithm is equipped with an additional component that leads to a greater similarity with human verbal working memory, making the xLSTM algorithm much more powerful.
With its increased efficiency and performance in processing texts, xLSTM technology will set a new standard in AI language processing.
NXAI as the growth engine of the LLM revolution
The research agreement between JKU and NXAI will bring previous findings related to xLSTM to market maturity. The aim of NXAI is to bring science and business closer together to ensure the successful technology transfer of AI solutions.
In addition to research activities, NXAI will bring together top researchers, entrepreneurs and industry experts. Although the research is independent and self-determined, selected industry partners will work closely with NXAI and researchers to efficiently transfer the potential of future AI development into practice. Initially, funding has been secured by PIERER Digital Holding as a co-shareholder of NXAI.