The 2024 Conference on Neural Information Processing Systems (NeurIPS) concluded in Vancouver. A vote is hard to come by for this meeting. This autumn, both Nobel Prizes are related to artificial intelligence, and the popularity of artificial intelligence is increasing. More and more private and national institutions around the world are investing heavily in artificial intelligence research. The fact is that European companies were rarely seen at this conference, and there were no artificial intelligence research laboratories in traditional European industries found here. We Europeans in particular should listen carefully to this discussion in Canada, as we are currently experiencing a process no less than the industrialization of artificial intelligence, and Europe has mainly been expanding the scale of artificial intelligence technology in recent years. Artificial intelligence also follows certain laws of technological history, and technological change is usually divided into three stages: basic research, scaling up, and industrial application or "productization". The development of steam engines or the Haber process for fertilizer production is an example of the three stages mentioned above, and both have had a profound impact on our society. Similar patterns can also be seen in the development of advanced artificial intelligence. The big language model, from basic research to scalability, is a good example. The Transformer model architecture proposed in 2017 dominated the upgrade phase, which is the second stage. The artificial intelligence community has a special fondness for Transformer architecture, but it is very primitive. Although this architecture has good performance, it comes at the cost of massive data and extremely high computing power. What does this mean? The working mechanism of Transformer architecture is that it first reads every word in a book, and when you ask it a question, it searches the entire book again to find the answer. Recently, other artificial intelligence architectures such as state space models and recurrent neural networks have also been upgraded. For example, Long Short Term Memory (LSTM) networks have been upgraded to the new xLSTM architecture. Compared to Transformers, these alternative solutions have more advantages, such as being more efficient. These alternative architectures, such as our European xLSTM, can read a book like humans, remember the plot, identify connections, and answer questions based on artificial memory. XLSTM is not just a language model, which is often misunderstood. At this conference, we showcased the "7B Language Model" to demonstrate our ability to establish large-scale language models. We believe that the xLSTM-7B model has set a new benchmark in terms of speed and energy efficiency. American companies are also very interested in this. From language to industry - how is this achieved? It is related to the sequence. A sentence contains multiple words, which form a sequence. We are also utilizing this. For example, the xLSTM architecture can also make predictions on industrial time series data and has great potential in robotics or image analysis (which is particularly important in medical technology). We are now entering the third stage, which is the industrialization stage of artificial intelligence. At this stage, we will fully integrate artificial intelligence models into practical applications of robots, biotechnology, or industrial processes. This requires not only scale, but also more consistency. With the advancement of the industrialization of artificial intelligence models, we expect to encounter many domain specific limitations - fortunately, this will be accompanied by an increasing diversity of methods, which will enable us to further develop our models, personalize them, generate new ideas, and provide added value. The industrial process is full of difficulties, so talent is needed. Therefore, experts in this field and artificial intelligence researchers need to collaborate. For example, they need to discuss operational safety and artificial intelligence in the manufacturing industry, discuss availability, robustness, safety, user experience, and human experience, discuss' never changing the running processes', as well as data inference and energy issues. Europe can achieve this. Because outstanding research has been conducted in Europe, including T ü bingen, Munich, Helsinki, Amsterdam, Zurich, and Linz. However, we Europeans are not always able to successfully transform these outstanding achievements into products and use them to explore new markets. I believe that industrial artificial intelligence is an opportunity for our European continent. But to be honest, we have been discussing this issue for several years now. How much longer do we need to discuss? We must start now - on a large scale. But companies are hesitating and avoiding investment. When American space exploration technology companies were able to achieve rocket landings, we were surprised and had to ask ourselves: which innovations come from Europe? The last company to achieve great success across Europe was Airbus. As is well known, that was over thirty years ago. We have conducted a lot of excellent research, including in many colleges and universities. This is a great advantage. But we must also acknowledge that we rely on the development of artificial intelligence in other parts of the world. Despite the existence of alternative solutions in Europe, we are still implementing the so-called 'single source strategy', which cannot proceed smoothly. (New Society)