All Sessions：

Over the years, edge AI has made significant strides, transitioning from applications like photo classification and face detection to enhancing video and gaming picture quality. The advent of generative AI has further propelled innovation, revolutionizing our life experiences. However, due to intensive computational and memory requirements, current generative AI applications primarily rely on cloud infrastructure. In this presentation, we aim to explore the emerging trend of shifting from a cloud-centric approach to a collaborative cloud-edge paradigm for generative AI technologies. We will delve into the multitude of opportunities and challenges that lie ahead when supporting generative AI on edge devices. This transformation entails a synergetic blend of technological advancements, enhancements in hardware architecture, and the reduction of computational complexities, ultimately enabling the realization of novel generative AI applications.

Over the past few months, the capabilities of large language models (LLM) like ChatGPT have sparked widespread discussion. They possess a general processing ability, and they can often accomplish the desired tasks with the right instructions. Speech, teeming with rich and hierarchical information, offers different needs for diverse tasks. How far are we from the speech version of ChatGPT? What are we still missing for the birth of the speech ChatGPT? In this talk, I will share a series of the latest research findings in speech LM that could potentially lead us to the speech version of ChatGPT and discuss current challenges and potential solutions.

Deep learning techniques have demonstrated great success in many application domains such as computer vision, speech recognition, and natural language processing. It has been shown that the memory subsystem is the main bottleneck in executing DNN applications, including the capacity and bandwidth walls. This talk covers recent researches for tackling the memory wall challenges.

Wings are a symbol of freedom, exploration, and discovery. They enable us to fly above the clouds, travel across vast distances, and reach new horizons. Artificial intelligence (AI) is like a pair of wings that frees us from mundane tasks and enables us to achieve things we never imagined possible. It changes the way we work, learn, and create. In this talk, we will unveil the secret behind the recent AI breakthroughs and how they are just the dawn of AI’s fusion into every facet of life. We will discuss the prospects and challenges of this new generative AI era that aims to unleash our creativity and potential.

In this session, we will focus on the risk of AI, share the framework of trustworthy of AI, especially for the usage of enterprise point of view. AI has become one of hottest topics in the enterprises, what are the risks and how to use to improve services to clients. We will discuss some of risks both from internal and external of an enterprise, and propose suggestions to help enterprise to balance of information protection and better services. Finally, we will share the framework of trustworthy AI, the elements of this framework, and the future development.

Digital twins of organizations (DTOs) can be used to model, simulate, and analyze various scenarios, enabling organizations to identify bottlenecks, inefficiencies, and opportunities for improvement. However, building trustworthy DTOs is challenging, as they must be reliable, robust, fair, ethical, transparent, and explainable. This presentation will discuss the challenges and opportunities of building trustworthy DTOs. We will present the early results of a life-critical use case experiment, in which we created a digital twin prototype, with the objectives of being not only accurate but also dependable. We will discuss how this innovation can be used to transforms decision-making, risk assessment, and predictive modeling in crucial contexts.

Generative AI has emerged as an outstanding and versatile solution. Its rapid advancements and deep impacts have exceeded many other technological breakthroughs, delivering substantial benefits to numerous industries. Within the semiconductor industry, generative AI integrated into EDA tools has solidified its position by offering unparalleled optimization capabilities. Indeed, these cutting-edge tools have demonstrated the capacity to change the rules of the game, enabling teams engaged in chip and system development to significantly enhance performance. Furthermore, they play a critical role in boosting engineering productivity and simplifying complex design convergence processes. In this presentation, we will introduce how EDA’s evolution through AI and how to leverage this transformative power into chip design and implementation.

Large language models (LLMs) have become an increasingly prominent feature in various computing platforms, such as data centers, smartphones, and microcontrollers. As such, their potential applications in both human-to-machine interfaces (HMIs) and machine-to-machine interfaces (M2MI) have been extensively studied and discussed. However, implementing hyper-scale LLMs on heterogeneous multicore System-on-Chips (SoCs) poses significant challenges. In this talk, we will discuss the key challenges, including issues related to hardware/software interfaces, verification, and SoC architecture exploration. At the end of this talk, we will propose a practical workflow and corresponding solutions we’ve applied in many cases. And see how our standard tool - ONNC - is used in the workflow. 大型語言模型（LLM）已成為各種計算平台（例如資料中心、智慧型手機和微控制器）中日益突出的功能。因此，它們在人機介面（HMI）和機器對機器介面（M2MI）中的潛在應用已被廣泛研究和討論。然而，在異質多核晶片上系統（SoC）上實現大型語言模型（LLM）面臨著重大挑戰。在本次演講中，我們將討論關鍵挑戰，包括與硬體 / 軟體介面、驗證和 SoC 架構探索相關的問題。在本次演講的最後，我們將提出一個實用的架構以及我們在許多案例中應用的解決方案。並了解我們的標準工具 - ONNC - 如何在架構中使用。

As a blue team on the defense side, we approach cybersecurity with different perspectives compared to the red team. Our considerations go beyond accuracy in handling large-scale data; we also prioritize compliance with government cybersecurity regulations, response time, and cost-effectiveness. Throughout our journey, our team has extensively researched using Machine Learning automation in the investigation of large-scale cybersecurity incidents. We have integrated AI into various analytical aspects, from endpoint event correlation and Cmdline analysis using CmdGPT to an AI virtual analyst that automatically generates forensic reports. Starting last year, we introduced a new generation of large-scale language models into our blue team AI assistant, enabling automated cybersecurity incident handling and case management. In this presentation, we will share our practical experiences and exclusive insights on how cybersecurity teams can leverage AI effectively, along with showcasing real-life case studies. 作為防禦方的藍隊有許多與紅隊切入角度不同的思維，不但需要考慮到大規模資料量下的準確度，為了符合政府的資安規範，也更加注重反應時間與成本。一直以來，我們團隊對於大規模的資安事件，在調查中使用 Machine learning 自動化方法做了許多研究，將 AI 導入到各個分析面向，從端點事件關聯、分析 Cmdline 的 CmdGPT，以及 AI 虛擬分析師自動產生鑑識報告，而到去年開始，我們為藍隊 AI 助手導入了新一代的大型語言模型，以進行自動資安事件處理與自動案件管理。這場演講中，將透過我們的實務經驗，獨家分享資安團隊該如何正確使用 AI，並分享實際的案例。

In recent years, generative AI technology has shown impressive performance in various fields such as images, videos, speech, and text. With the maturation of these technologies, many industries have begun adopting generative AI techniques to reduce the human resource costs required for technical development. This includes tasks related to software development and security analysis. In this presentation, we will introduce how generative AI technology is utilized to assist in applications related to software development and security analysis. We will also discuss the risks and challenges that may arise from the integration of AI. 近年來生成式 AI 技術在影像、視訊、語音、文字等領域都有讓人驚豔的表現。而隨著技術發展的成熟，許多產業也開始導入生成式 AI 的技術，以降低技術開發所需的人力成本，這也包括軟體開發和安全分析相關工作。在這個分享裡，我們將介紹軟體開發和軟體安全分析相關應用如何運用 AI 技術進行輔助，以及導入 AI 可能會遇到的風險和挑戰。

In this rapidly evolving era of the AI and Generative Content Industry (AIGC), we will explore investment trends and opportunities in the Gen AI market from a venture capital perspective. We will examine global VC investment trends, market potential, and value within the Gen AI industry chain and its components worldwide. Additionally, we will uncover the areas of investment focus and highlights in the development trends of enterprises adopting Generative AI, with the aim of collectively discussing the future of this industry.

數位轉型下的客戶服務 Digital Transformation and Its Impact on Customer Service 整合 GPT 的智能機器人 Chatbot Empowered by GPT Integration 對話式知識管理是最佳服務助理 Conversational KM is the Best Service Assistant

Learn how AI agents revolutionize efficiency across industries. We'll dissect the components that make them work: Planning, Memory, Tool Use. Real-world case studies from HelperAI will demonstrate impact on customer service, logistics, and more. Whether you're an academic, a business leader, or just curious, this talk offers actionable insights into AI’s transformative power.

9/14 下午給

In this talk, we will delve into the remarkable advancements in Natural Language Processing (NLP) and their transformative potential in the realm of clinical data analysis. The field of medical imaging has seen tremendous growth, and one of the key challenges lies in translating radiomics features into actionable insights. Firstly, we will explore a groundbreaking approach that addresses this challenge by automatically generating Lung-RADS reports from low-dose lung CT scans. By leveraging the power of deep learning to construct a multi-objective framework, we have advanced the extraction of essential information to aid in the accurate diagnosis of lung abnormalities. Moving beyond medical imaging, pathology reports play a crucial role in guiding treatment decisions, particularly in lung cancer cases. We will investigate how Generative Pre-trained Transformers (GPT) can extract vital features from pathology reports, enabling the identification of key indicators for effective lung cancer treatment. By harnessing the capabilities of GPT, we unlock a wealth of information buried within textual data, enhancing the precision of personalized therapies. Moreover, advanced lung cancer requires comprehensive decision support systems. Deep learning-based techniques offer immense potential in this domain by automating the extraction of key features from pathology reports. By integrating these extracted features into a decision support system, we empower clinicians with valuable insights, aiding in the development of tailored treatment plans and ultimately improving patient outcomes. Throughout this presentation, we will witness how NLP, combined with deep learning methodologies, revolutionizes the analysis of clinical data. By bridging the gap between unstructured text and actionable knowledge, we unlock the full potential of clinical data and propel us into an era where diagnoses are more accurate, treatments are more personalized, and patient care is optimized. Together, let us embark on a journey that demonstrates how NLP is shaping the future of healthcare, forging a path towards a world where the integration of technology and medicine empowers us to extend the potential of clinical data like never before. The future is indeed here, and it is through the fusion of NLP and clinical data that we shall explore its boundless possibilities.

How can students become a better learners through ChatGPT? This presentation outlines the implementation of ChatGPT in primary education to facilitate deepening learning. It delves into the application of six evidence-based learning strategies, each coupled with corresponding prompts 如何透過 ChatGPT 成為更好的學習者，分享內容在於在小學的教學中使用 ChatGPT 以促進學生更深度的學習進而到依據 Evidence-based learning strategies 提出的六個學學策略跟對應的 prompts 以幫助學生成為更好的學習者

In the rapidly advancing age of Artificial Intelligence, esteemed global institutions have underscored the urgent need for reskilling and upskilling. As we find ourselves at the crossroads of technology and tradition, it becomes crucial to identify sustainable avenues for concerted effort and investment. Synthesizing insights from leading research entities, Haishuo will demystify the essential competencies set to define the AI-centric world. Delve deep as Haishuo showcases cutting-edge AI applications in educational landscapes worldwide and articulates spheres where the human touch remains irreplaceable.The presentation strives to offer an integrative exploration of present-day research, forward-looking perspectives, and the age-old human strengths that promise to endure. Come to the talk for a humble presentation of the intertwining destinies of AI and human resilience - an enlightening engagement, offering tools to chart our course in an AI-augmented future. 在人工智慧急速發展的時代，世界各大知名機構都強調了重新學習和提升技能的迫切需要。當我們處於技術和傳統的交叉路口，確定可永續投入的方向已變得至關重要。綜整全球各研究機構的洞見，海碩將向您稟報 AI 為中心世界的核心素養，深入探索全球教育風景中的尖端 AI 應用，並強調人類知能無可取代的領域。此演講旨在提供現今研究、前瞻性視角和人類亙古知能的綜合探索，期能在變動的時代提供永續發展的力量。請蒞臨本次演講，聆聽 AI 和人類韌性相互交織的命運的淺見，期能拋磚引玉，共同探討引領在 AI 增強未來道路上可用的工具。

With the rise of ChatGPT, Generative Artificial Intelligence (Generative AI) has gradually garnered attention from various fields. Many professors both domestically and internationally have attempted to utilize ChatGPT to answer university exam questions and have been pleasantly surprised by the promising results achieved. According to a recent report from EDUCASE, approximately thirty percent of university professors hold a pessimistic attitude towards Artificial Intelligence in General (AIGC), with over half of them being unaware of how students use AI. What impacts might Generative AI have on higher education? What roles could AIGC play in admissions, teaching, and research within higher education, and what potential disruptions might occur?