[Speech Summary] Explainable AI for Safety-Critical Systems: Enhancing Trust in AI A crucial aspect of AI applications in industry is AI transparency—understanding the reasoning processes behind AI models. This concept closely aligns with the goals of explainable AI, a subfield focused on enhancing the interpretability of complex AI systems.
In the first part of my talk, I will explore various research initiatives from my lab, emphasizing key principles and techniques of explainable AI and how they are applied across diverse healthcare tasks. In the second part, I will delve into how explainable AI can foster trust and transparency in clinical AI models, showcasing our recent work in dermatology AI as an example.

[Speech Summary] Accelerating Materials Development through AI In the electronics and semiconductor industries, the development of new materials and the optimization of existing materials have the potential to improve product quality while reducing power consumption and costs. For example, the innovation of OLED technology used in smartphones and Quantum Dot technology used in TVs shows that the innovation of materials that convert electricity into light can determine the success of products and industries. In addition, the innovation of semiconductor process materials is also necessary to overcome the limits of micro processes.
However, material development is a time-consuming and difficult process that requires a lot of time and research personnel. Therefore, the utilization of computing such as quantum chemical computer calculations that express physical phenomena is increasing in addition to researchers' experience and research on physical and chemical theories. With the development of deep learning, efforts are being made globally to apply AI to the material development process to shorten the development period and improve material performance.
Samsung Electronics SAIT has been conducting deep learning research since 2012 and has successfully applied it to smartphones and TVs, innovating speech recognition, machine translation, and face recognition. Based on this successful AI research capability in consumer products and interfaces, SAIT has been focusing on developing AI technology that innovates the technology development process, such as semiconductor development and material development, in recent years.
However, compared to AI for consumer products, AI for material development requires solving difficult problems such as resolving data shortages, fusing domain knowledge, and the difficulty of technology verification. SAIT is successfully solving these difficult problems by combining AI experts and material development experts into one team and applying AI to fields such as QD and OLED. In the future, SAIT plans to expand the application of AI to semiconductor material development based on this experience.

[Speech Summary] AI Home, where the Future of an Intelligent Life meets Affectionate Intelligence The AI Home is a smart home solution that leverages artificial intelligence (AI) technology to connect and control various devices, systems, and services within the home. LG Electronics aims to develop an AI Home built on Affectionate Intelligence, which empathetically understands and relates to our customers by learning and analyzing their habits and emotions. This approach provides evolutionary assistance in everyday situations. Users can enjoy a differentiated experience, allowing them to communicate with AI in everyday language, upgrade existing appliances into AI-powered devices, and expand their control beyond appliances to include a variety of services.
In the LG AI Home, generative AI that understands contextual situations allows users to interact seamlessly in everyday language, as if conversing with family members. Powered by LG’s AI technology platform, FURON, it enables the execution of Affectionate Intelligence to deliver benefits in daily life, including convenience, safety, security, and care.
Additionally, existing appliances can be upgraded to AI-powered devices and the appliances will continuously evolve in their journey toward AI Home, going beyond simple connectivity to create new customer value.
Lastly, infinite scalability can be expected. LG Electronics plans to build a broad open ecosystem and continuously enhance connectivity by acquiring the smart home platform company 'Athom' and supporting the ThinQ Connect API of the open smart home platform ‘Home Assistant’.
LG Electronics aims to provide an 'Intelligent Space Solution' that extends the experience of AI Home beyond comprehensive care for customers' daily lives at home to various spaces including mobility. In this session, we would like to introduce 'LG AI Home: where the Future of an Intelligent Life meets Affectionate Intelligence,' marking the beginning of our journey.

[Speech Summary] If you can't measure it, you can't manage it In precision manufacturing, such as semiconductor and display manufacturing, the process of making a product involves hundreds, if not thousands, of complex steps. Sometimes, the whole process takes several months to complete. If we only check the final output, it is already too late to discover the root cause of faults and anomalies in the products and fix them. So, it is important to check the outcomes of a process in line, after each intermediate step, through measurement, or in a technical term, metrology. However, due to cost, time and space constraints of physical metrology tools, manufacturers have to sample a few products to measure, obtaining only limited information about the process.
Gauss Labs solves this challenge with Panoptes, an AI-powered metrology solution. Named after a 100-eyed giant in Greek mythology, Panoptes provides a full visibility into manufacturing processes in real time. Panoptes analyzes a variety of fab data with state-of-the-art AI technology and distills critical information across all wafers in manufacturing processes. Panoptes opens the door to endless use cases such as process control and monitoring, sampling rate optimization, equipment monitoring and matching, and yield analysis. As a result, manufacturers can optimize their manufacturing processes, increasing productivity, lowering costs, reducing cycle time, and improving yield and quality.

[Speech Summary] PBV, A New Concept of Space & Mobility Kia aims to be at the forefront of future mobility, creating a brighter future through PBV and its connected solutions. We would like to propose a new concept of space where you can choose where and how you wish to live and work - living rooms, offices and workshops - a shapeable space that helps you improve your life. Through innovative modularization, we aim to provide excellent flexibility and customization solutions, offering spaces that make customers' lives more convenient and prosperous, whether stationary or on the move. Kia's PBV, which is not tied to a fixed location, and goes beyond existing spatial constraints, will provide a blank canvas to reimagine lifestyles and work methods. Moving and customizing our spaces will become a natural way of life, creating a new culture of living, working, and mobility.
How will mobility change in the era of great transformation and hyper-individualization? How will people's lives change? What will mobility, or movement, mean? Through these questions, with the Kia PBV, we have pursued intelligent and thoughtful mobility with unmatched flexibility, a radical level of modularization, and extensive software integration. The true value they bring, however, is how they reshape our concepts of space and time, and how they seamlessly and naturally blend into our daily lives. Whether for private, business, or family use, the heart of PBVs lie in the people and communities they serve.
In our ever-changing world, we are witnessing a shift in mobility; cars are expanding the role they play in our lives. From a means of movement to a means of living, car interiors are developing rapidly into living and working spaces.
In this transition, as the role and scope of car designers also evolves, design can bring new values and experiences to our lives. Kia design aims to connnect and envelop a wider range of issues than it had previously, including not only styling aspects, but also business-goal driven product concepts, application of technological elements, UX/UI design, and all-encompassing storytelling, as well as their visualization. Through this, we aim to make a valued contribution to meeting the needs that come with change, making our users' lives easier and better.
As the world of mobility becomes more integrated into our everyday lives, we are building a vision and hope, that PBVs will bring a new concept of life with mobility. Though our philosophy 'Design for Society', we are bringing solutions that we believe, will make a positive impact on future society.
This is Kia’s vision for an optimistic and positive future.

[Speech Summary] Collaborative Robots: Market Insights & Future Outlook The potential market for collaborative robots is expanding as the overall global robotics market grows bigger and bigger. By answering the following three questions, Doosan Robotics seeks to examine the current state of the collaborative robot market and hopes to suggest how companies should formulate their strategies in the future:
First, what is the current state of the global robotics market? Second, what could cause a chasm in the collaborative robot market, and what specific needs/hurdles do customers have in this context? Third, what strategies are needed to overcome the chasm, and what synergetic effects can be created by combining AI with robots?
In this forum, Doosan Robotics will also discuss how collaborative robots are used across different sectors from manufacturing to service, and explore how collaborative robots will evolve in the future.

[Speech Summary] Driverless Technology: Smart, Safe, and Sustainable Mobility Mr. Se Hyuk Park is the Chief Strategy Officer at Motional, a Hyundai Motor Group joint venture and global leader in driverless technology. Mr. Park will discuss the roadmap to making autonomous vehicles a global reality, and how Motional is approaching its technology and product development. Motional is uniquely positioned through its technology expertise, commercial experience, and strategic partnership with Hyundai to deploy autonomous vehicles at scale. The adoption of autonomous vehicles has the potential to make transportation safer, more sustainable, and more accessible.

[Speech Summary] Recent Trends in Energy Industry As the AI era unfolds, the power demand for data centers is expected to more than double by 2028, driving the need for a wider range of energy sources. In response to global efforts to tackle the climate crisis and shift toward carbon neutrality, renewable energy, especially solar power, is becoming increasingly vital.
Bloomberg projects that global solar installations will grow by 33% in 2023, reaching 592GW, while in the U.S., installations are set to rise by 16% to 43GW, pointing to steady long-term growth. With the rapid expansion of the U.S. solar industry, demand for non-Chinese polysilicon is also anticipated to rise significantly. However, given the global production of non-Chinese polysilicon (around 110,000 tons per year), supply is likely to remain tight relative to demand.
OCI Holdings, which pioneered the development of polysilicon in Korea in 2008, currently produces 35,000 tons of solar-grade polysilicon annually through its Malaysian subsidiary, OCIM. By capitalizing on the premium for non-Chinese polysilicon, the company is solidifying its competitive edge among major global customers targeting the U.S. market.
This moment presents a valuable opportunity for OCI Holdings to share its business strategies with industry experts, especially in light of rapidly shifting global policies like U.S. sanctions on China, including the Inflation Reduction Act (IRA) and the Uyghur Forced Labor Prevention Act (UFLPA), as well as the globally established RE100 initiative.
Founded in 1959 as Oriental Chemical Industries Co., Ltd., OCI Holdings has long been a cornerstone of Korea’s chemical industry. Today, the company is evolving into a global leader in the renewable energy space.

[Speech Summary] Ultracapacitors as a New Solution for the Age of Electrification With recent advancements in generative AI technology, big tech companies are aggressively investing in data centers to enhance their AI capabilities. It is estimated that by 2024, the four major tech giants’ investments in AI data centers (AIDC) will reach approximately 220 trillion won, with this figure projected to rise to around 300 trillion won by 2027, reflecting an expected annual growth rate of 15%.
AIDCs, which are rapidly increasing in number, differ significantly from traditional internet data centers (IDC) for communication in terms of power load. Not only do AIDCs consume substantially more energy, but they also experience frequent and significant fluctuations in power output, with load variations occurring every few dozen seconds. These distinct power characteristics make it essential to develop power solutions tailored specifically to the needs of AIDCs.
The rapid increase in power demand driven by AI growth is leading to greater instability and shortages in power grids, especially as renewable energy becomes more prevalent in response to the rising demand for carbon emission reductions. Renewable energy is expected to expand by over 300% by 2030 compared to 2022. However, renewable energy sources such as solar and wind are inconsistent in their power generation, leading to fluctuations in frequency and voltage. If the frequency deviates from regulated levels—whether exceeding or falling short—immediate control and compensation are necessary to prevent disruptions to the power system. Therefore, a system capable of stabilizing these fluctuations is critical.
In response to this need, I introduce ultracapacitor (UC) as a key power stabilization system essential for the global megatrends of AI expansion and the explosive growth of renewable energy. With their rapid charge and discharge capabilities and excellent power output, UCs can control and compensate for large currents in a short period of time. Additionally, they have a lifespan of over 10 years, making them well-suited to play a crucial role in both of the aforementioned fields—AI and renewable energy.

[Speech Summary] Advanced Nuclear Offers Solutions for Growing Electric Demand TerraPower, a nuclear innovation company based in Washington, is a global leader in next generation nuclear. TerraPower is currently under construction with its Natrium Demonstration project; which is advancing our sodium fast reactor technology and built-in energy storage system near a retiring coal facility in Wyoming under a United States Department of Energy public-private partnership. The Natrium plant is set to be America’s next nuclear power plant, and the first commercial Generation IV nuclear reactor in the country. Through this project, TerraPower is showing how innovation-driven public private partnerships hold the key to enable the energy transition of tomorrow.
As nations seek to electrify more of the economy, including the rapid growth in transportation, industry and data centers driving the growth in AI, there is a critical need for highly reliable, baseload and carbon free technologies; and nuclear is among the only scalable and affordable solution.
Yet, it takes time and capital investment to introduce a new generation of nuclear into the commercial market. Nuclear power plants require substantial capital investment upfront but have relatively low operational costs and fuel costs. Over many decades’ nuclear power has proven to be highly cost-effective, the energy return on investment (EROI) for nuclear energy is well above other energy sources, and the lifetime of the technology can be up to 80 years.
TerraPower’s Natrium technology is designed to be inherently safe, which requires far less heavy containment and redundant safety systems compared to the current generation of nuclear.
With more intermittent generation sources on today’s power grids, and a rapidly increase demand for energy, we will need power plants like Natrium that can couple with renewables by providing dispatchable, clean power and flexible load following to balance a completely carbon-free grid.

[Speech Summary] AI Use Cases and Future Prospects in the K-Content Industry Studio Freewillusion, a pioneering AI-driven film production startup, has played a significant role in the development of the "AI film" genre. The studio's innovative work was recognized at the 1st Dubai International AI Film Festival, where it won both the Best Film Award and the Audience Choice Award. In this discussion, some of the AI-generated video content produced by Studio Freewillusion will be presented and various ways AI is currently being utilized in video content production will be explored.
We will also examine real-world examples of AI adoption in the domestic film industry and consider what the future may hold. As we reflect on how best to prepare for these changes, we will discuss the growing necessity for increased AI integration in the industry.

[Speech Summary] Could ChatGPT Write a Bestselling Novel? There are conflicting views on Large Language Models (LLMs) trained on vast amounts of text: some see them as mere stochastic parrots, while others worry they will take human jobs. This presentation examines how LLMs like ChatGPT work, their current limitations, and future development directions.
LLMs generate sentences using “autoregression”, predicting the next token based on given input and adding it back to the input. Although LLMs lack internal structures for complex thinking, they can perform multi-step inference through a "Chain of Thought" process while generating sentence. However, merely sampling high-probability tokens makes it difficult to produce results that deviate from existing probability distributions while still meeting the intended purpose.
Aristotle said that a good story ending should be "unexpected but inevitable." The novel ideas, complex structures, foreshadowing, and plot twists required for good novels cannot be generated by autoregressive language models alone. An AI agent architecture capable of high-level design, multiple attempts, review, and improvement is necessary, along with techniques to generate creative content that avoids clichés.
These AI advancements will ultimately aid not only novel writing but many tasks requiring creativity and complexity. For the foreseeable future, we will see various forms of collaboration between AI and humans.