The Role and Prospects of AI in Strengthening the Safety of Nuclear Facilities
Drafting scientific articles, creating paintings, automating routine tasks, processing large amounts of data, forecasting, and solving complex problems – all of these are already being successfully done with the help of artificial intelligence (AI). The humanoid robot Sophia, created based on artificial intelligence, learns, jokes, and gives interviews in which she expresses her desire to have a family like humans. Technology companies all over the world are striving to master AI, which is why investments in AI research amounted to $21.3 billion in 2023
Artificial intelligence is an organized aggregation of information technologies that can be used to perform complex tasks by using a system of scientific research methods and algorithms for processing information received or independently created during work, as well as to create and use its own knowledge bases, decision-making models, algorithms for processing information and determine ways to achieve the tasks.
The way AI functions is similar to the way the brain works. It is a mathematical apparatus with an electronic part that can work with information like a human. Ideally, it can think, learn, and synthesize new information.
The unlimited possibilities, speed, and ability to unite and synchronize pose certain threats to humans from AI. “A person who controls artificial intelligence based on personal motivations can harm a large number of his fellow humans. It was the same with the nuclear bomb – it could not be banned. And we all seem to know that it is bad, but it is there,” said Oleksiy Reznichenko, founder and head of the Boteon Robotics Center network.
Legal Framework for Regulating Artificial Intelligence
At the end of 2020, the Cabinet of Ministers approved the Concept of Artificial Intelligence Development in Ukraine. Our country aims to significantly increase the number of qualified AI specialists, improve the quality of scientific research in this area, and spread the skills of competent use of artificial intelligence among the public.
Everyone benefits from the proper use of AI: process automation increases performance, improves the quality of administrative services provided to citizens and businesses, and reduces the impact of corruption.
European countries see AI as a serious threats, so in August 2024, the European Union enacted the Artificial Intelligence Act, which aims to guarantee the protection of people’s fundamental rights regarding AI.
From now on, chatbots must clearly notify users that they are interacting with AI. Certain content created with the help of a neural network, including deepfakes, must be marked with special labeling. The same is true for biometric categorization or emotion recognition system: a person should be informed that AI is interacting with him or her at that moment.
AI systems used for recruitment, credit assessment, or control of autonomous robots pose a high level of threats. Such AI systems must meet the requirements of a risk mitigation system and have clear information for users.
This law, in particular, prohibits applications that can manipulate people’s behavior: games with voice assistants that encourage dangerous behavior of minors, etc. In addition, the use of certain biometric systems will be prohibited, such as emotion recognition systems in the workplace or real-time remote biometric recognition for law enforcement purposes. Such AI systems are considered a serious threat to people’s fundamental rights.
Most of the rules of the AI law will be applied from August 2, 2026. However, the prohibition of AI systems that manipulate people will be applied in six months.
Potential Opportunities for AI in the Nuclear Industry
AI methods and algorithms are used to improve and optimize various aspects of nuclear technology, in particular, for the operation of nuclear reactors, radiation detection, analysis of nuclear materials, radioactive waste management, etc.
By utilizing AI’s ability to recognize, analyze complex data and make decisions, it is possible to qualitatively improve the nuclear sector by increasing efficiency, safety and resilience.
AI-powered predictive maintenance systems can monitor equipment health and predict malfunctions, reducing the likelihood of accidents and unplanned shutdowns. Robotic learning algorithms can also assist in real-time anomaly detection, helping operators quickly identify and respond to hazards.
Researchers from Argonne National Laboratory have described how nuclear power plant (NPP) operators can benefit from diagnostic information filtered through the GPT-4 large AI model.
This AI model compares real-time data from NPPs with the expected normal behavior in physics-based models to detect malfunctions.
In case of discrepancy, the system provides a probabilistic distribution of malfunctions.
The system was tested at Argonne’s Mechanisms Engineering Test Loop Facility, where it diagnosed a faulty sensor in a non-nuclear liquid sodium test facility and explained the problem to operators.
The U.S. Department of Energy released a report on the AI potential to support the growth of clean energy in America, as well as promising areas of research on AI for nuclear energy.
AI will facilitate regulatory compliance by automating data analysis and reporting, ensuring that operations meet safety and environmental standards.
Nuclear power has the potential to provide a significant portion of the world’s clean energy needs. By optimizing reactor operation and fuel utilization with AI, nuclear power plants can be made more efficient, leading to increased energy and reduced waste generation.
AI can also more accurately model waste behavior and degradation, helping to develop safer disposal methods. Robotic learning algorithms can help identify potential locations for nuclear waste storage facilities based on geological and environmental factors.
Faculty, a British artificial intelligence company, is working with Veolia Nuclear Solutions to sort radioactive waste in a $7.5 million project. By using AI and machine learning applications, radioactive waste will be sorted faster, and its management will become safer thanks to research and development programs between government and industry.
And finally, it has been identified that AI can control nuclear fusion by analyzing the vast amounts of data generated during fusion experiments, helping researchers understand the complex plasma dynamics and develop more efficient fusion reactors.
IAEA Steps to Explore the Possibilities of Artificial Intelligence
Thanks to rapid progress in computing resources and data analysis tools, the nuclear industry has already started using AI to optimize the NPP operation and maintenance. AI also supports the development of advanced nuclear power technologies, such as small modular reactors (SMRs).
The IAEA has designated the Center for Information Science at Purdue University in the United States as the first IAEA Collaborating Center to support the Agency’s activities in the application of artificial intelligence in nuclear energy, including reactor design, plant operation, training and education.
The five-year agreement of the Collaborating Center will support the IAEA’s programmatic activities and knowledge exchange in the field of AI for nuclear energy. This includes benchmarking initiatives to build confidence and acceptance of AI for nuclear power, the establishment of a “benchmarking center” for data coordination and management, as well as other activities related to the development and evaluation of AI technologies in cooperation with IAEA Member States.
The Collaboration Center Agreement is part of the IAEA’s efforts to support countries interested in using AI for nuclear science and technology. The IAEA’s International Network on Innovations to Support Operating Nuclear Power Plants is studying the regulatory and technical aspects of AI deployment. Several coordinated research projects are currently underway, one of which is related to the use of AI in SMR.
The agreement follows the Agency’s recent designation of MIT’s Center for Plasma Science and Fusion as the first collaborative center focused on accelerating fusion research, with a focus on the use of AI to advance the IAEA’s AI initiative for Fusion.
To promote the peaceful use of nuclear technology, the IAEA works with relevant institutions around the world. Through a network of cooperation centers, these organizations in Member States can assist the IAEA by conducting original research and development and training related to nuclear science, technology and their safe and secure applications.
Hazards of Using AI
Artificial intelligence algorithms require huge amounts of data to train and make accurate predictions. However, in the nuclear sector, sensitive information about reactor design, operation, and safety measures must be protected from hackers seeking to exploit vulnerabilities to steal confidential information, sabotage, or gain unauthorized access to nuclear facilities.
Ensuring effective communication between humans and AI is essential to prevent misunderstandings and errors. Implementing AI in the nuclear sector requires an operator to navigate a complex regulatory framework. AI systems can inadvertently learn biases present in training data, potentially leading to distorted results. Due to widespread biases or numerous inaccuracies, the decisions that AI makes can have serious safety and environmental implications. That is why we need the influence of a specialist and his leading role in decision-making. After all, who will be held responsible if the AI system makes a critical error and causes an accidental disaster? How transparent are the decision-making processes of AI algorithms in high-stakes nuclear operations? If so, such decisions will lead to increased global instability and a potential arms race.
Ideal Combination: Human Brain and Artificial Intelligence
When an operator uses AI-generated data, it is called human-machine interaction (HMI). This is the typical way of working with AI in a nuclear power plant, as operators receive data collected from sensors or devices and make decisions based on this data, making them a particularly important component in the nuclear power industry. A well-established HMI in the main control room (MCR) can help reduce operator errors and ensure safe operation of the NPP. Today, NPPs use simple graphical interfaces that are outdated, and no new devices are being developed.
The human factor always plays a crucial role in HMI, and to reduce human errors and increase NPP reliability, researchers propose AI methods. In particular, to use a virtual control panel to simulate a real situation and to use AI to work out the content of the text in orders to reduce errors
It is also proposed to use three-dimensional computational modeling based on virtual reality for assessment. Studies indicate that the digitalization of MCRs, including HMIs, has both positive and negative impact on the reliability of human operators. The main problems are the increase in system complexity and workload, which can be simplified with the help of AI technologies.
International partnerships can facilitate the exchange of data and best practices, ensuring responsible and safe development of nuclear AI. It is important to ensure strict regulation, oversight and responsible use of AI in the nuclear field. Researchers and stakeholders should work together to address the challenges and mitigate the risks associated with this technology. Here are some potential ways for future developments using AI:
- development of more efficient and safer nuclear reactor designs, including small modular reactors and molten salt reactors, through rapid modeling and optimization of AI algorithms;
- AI systems can autonomously manage reactor operations, maintenance and safety protocols, thus minimizing human intervention and potential errors.
Humanity is rapidly bringing the future of nuclear artificial intelligence closer, but we must remember to prioritize safety, cooperation, and ethical responsibility. Only a balanced approach will allow us to use the full potential of AI to improve and develop the nuclear industry.
Uatom.org Editorial Board