OpenAI and Mistral AI's machine learning methods are known for their high computational demands, requiring extensive resources to process entire datasets before generating insights. Sepp Hochreiter, a pioneering figure in AI technology, offers a more resource-efficient approach. Operating an AI lab at Johannes Kepler University in Linz, Austria, Hochreiter focuses on teaching AI models how to forget efficiently.
Hochreiter's influence in AI dates back to the 1990s when he developed a foundational theoretical framework at the University of Munich. This framework laid the groundwork for the first generation of agile AI models used by companies like Alphabet, Apple, and Amazon. Known as Long Short-Term Memory networks (LSTM), this method allows computers to remember complex data while discerning which information to discard. After his research was published by MIT Press, LSTM became an industry standard.
As concerns over AI's massive energy consumption grow, alongside Europe's slow start in AI development, Hochreiter has re-emerged with a new AI model based on his earlier work. Last May, he and his team introduced xLSTM, claiming it outperforms generative AI in speed and energy efficiency. He likens its operation to reading a book, where readers remember key plot points without revisiting every word, highlighting the importance of distinguishing between essential and dispensable information for rapid computation.
This efficiency means xLSTM doesn't rely on expensive data centers for large data storage and processing. Hochreiter describes it as a "lighter, faster model with much lower energy consumption than existing solutions."
While the AI industry has long been dominated by U.S. tech giants, companies like China's DeepSeek are gaining traction with efficient AI solutions. Hochreiter believes tailored AI models are ideal for Europe, emphasizing the integration of existing technologies and methods.
Despite skepticism about xLSTM's scalability, Hochreiter's lab has incubated two companies collaborating with European manufacturers. NXAI GmbH, where Hochreiter serves as Chief Scientist, has raised €20 million. Emmi AI GmbH, led by former Microsoft researcher Johannes Brandstetter, began operations in March. NXAI focuses on vertical AI models tailored to industries like automotive and robotics.
Hochreiter remains confident in his approach, stating, "We have created a better product."
