The symposium will explore the most important issues of artificial intelligence over three days along four main themes: reliable and sustainable artificial intelligence, network science, healthcare and industrial automation. The symposium, which is from Thursday to Saturday, is jointly organized by the HUN-REN Hungarian Research Network and the Nanyang Technological University (NTU) in Singapore.
At the opening, President of the Republic Tamás Sulyok emphasized that he is convinced that Hungarians will be able to make good use of the opportunities offered by artificial intelligence, and that they will serve the benefit of humanity. He added that technology alone will never solve our fundamental problems, but it can make the world more attractive, sustainable and safer.
Balázs Gulyás, President of HUN-REN, recalled in his welcoming speech that Budapest has long been a city of science, innovation and the pursuit of knowledge, and that Hungary has given the world mathematicians such as - among others - János Neumann, János Kemény, Alfréd Rényi, Pál Erdős and László Lovász. The President said that "collective momentum is being forged" at the event, adding that at the symposium "the world's leading minds from various disciplines and continents will shape the future of artificial intelligence, responsibly, dedicatedly and sustainably, for the benefit of humanity".
Roland Jakab, CEO of HUN-REN, emphasized in his speech that "the research and application of artificial intelligence opens the door to opportunities for humanity that we have only dreamed of before." The CEO believed that this revolutionary technology is in the common interest of humanity, a collective cause, but at the same time it is necessary to recognize that this is also a competition that takes place between nations, economies and societies. Jakab Roland expressed his conviction that HUN-REN therefore contributes not only to global development, but also to the local economy, the standard of living of the Hungarian people, i.e. to the Hungarian common good.
The CEO added: those who fall behind in the development of artificial intelligence will inevitably be pushed into the background in the economic competition that will define the coming decades. "Therefore, it is our duty not only to be participants, but also to be key players in this transformation." From a historical perspective, AI may be humanity's most important invention to date, but it is still far from ready and even further from perfect, so "we must definitely do it well," said Roland Jakab, adding that the evolution of artificial intelligence is due to science, while AI has also become the driving force of science, as it accelerates discoveries and reveals new possibilities.
At the opening, László Palkovics, the government commissioner responsible for artificial intelligence, said that AI is now affecting our everyday lives and many segments of life. He drew attention to the need to modify the legal framework of artificial intelligence and to the fact that in the world of AI, answers are no longer primarily sought for technological, but rather for legal questions.
Luke Ong, a researcher at Nanyang Technological University in Singapore, emphasized that the aim of the symposium is to promote cooperation between Hungarian and Singaporean researchers in the field of artificial intelligence. His presentation focused on Reinforcement Learning (RL), with a special focus on the safety of AI. He demonstrated how formal methods, such as Linear Temporal Logic (LTL), can make reinforcement learning safer and more transparent. In his presentation, the Singaporean professor presented theoretical foundations, real-world applications (e.g. robotics, self-driving vehicles, large language models) and current challenges (e.g. reward errors, decision-making in unfamiliar environments), and added that increasing the safety of AI is an urgent task, if only because artificial intelligence is moving very quickly towards human-level capabilities.
Tao Dacheng, a professor at NTU in Singapore, presented deep model fusion in artificial intelligence, a method that combines multiple pre-trained AI models to improve performance, reduce training costs, and adapt more effectively to new tasks. The Singaporean researcher emphasized the importance of efficient AI development and deployment, especially in resource-constrained environments. He also highlighted experimental strategies such as task arithmetic and dictionary-based learning. These aim to optimize task-specific model fusion and thereby bridge the performance gap compared to traditional multi-task learning.
During the breakout sessions, Google DeepMind researcher Ira Ktena said that the responsible use of generative artificial intelligence could fundamentally transform healthcare. "But for this to be truly effective, these systems need to be robust, equitable, data-secure, and backed by medical expertise," the expert emphasized.
Ferenc Huszár, Professor of Computer Science and Technology at the University of Cambridge, said that large language models are capable of intelligent, rational generalizations that are not directly reflected in their training data. He said this suggests that their capabilities may go beyond simple pattern recognition and interpolation.
Bo An, Head of the Department of Artificial Intelligence in the School of Computer Science and Data Science at NTU, explained that while large language models (LLMs) are great for generating text and answering questions, AI agents can do much more: they can follow goals, use tools and interact with their environment. According to him, this requires not only LLMs, but also planning capabilities, environmental feedback and tool integration, which can be implemented using, for example, reinforcement learning (RL). The first RL-based approaches were still quite slow and had limited generalizability, but combining them with LLMs has brought significant improvements in the efficiency and applicability of AI agents.
The final keynote of the day was delivered by Lin Weisi, Deputy Dean of NTU, who discussed, among other things, how machines can help humans perceive visual and multimedia signals. According to the professor, to build intelligent and effective artificial intelligence, we need to go beyond processing raw data and model how humans perceive the world so that machines can operate more naturally and resource-efficiently in human environments.
Wen Yonggang, Head of the Department of Computer Science and Data Science at NTU, said about deep learning that artificial intelligence is both an environmental burden and a tool to reduce it. To build a sustainable digital future, we need to rethink how we develop and use AI systems. "We need to make them not only more efficient, but also environmentally friendly and in line with global sustainability goals," the professor emphasized.
Another interesting aspect of the parallel sessions was the panel discussion, which not only covered the most current topics in brain research - such as ADHD, schizophrenia, dementia and AI-based healthcare diagnostic developments - but also the participating researchers - Miklós Szócska, James Cole, Adrián Csiszárik, Jagath Chandana Rajapakse - answered questions from the audience. Among other things, the use and future of portable MRI devices were discussed, as well as the aging processes of the brain. With today's technologies, researchers are already able to predict what changes we can expect in the brain at a given age based on a single brain image. AI can also be used to determine the biological age of the brain and, using numerous large samples, even blood picture data, it is possible to predict the likelihood of a given patient having a chronic disease. Thus, medical science can predict the risk of developing diabetes, a common disease, even before any typical symptoms have appeared.