VIP / Speakers

頁數: 1 2 3 - 每頁 20 筆

共有 41 位講者

All Sessions:

Hung-yi Lee

Hung-yi Lee Self-supervised Learning and its applications to speech processing

There is a trend in the machine learning community to adopt self-supervised approaches to pre-train deep networks. Self-supervised learning utilizes proxy supervised learning tasks, for example, distinguishing parts of the input signal from distractors, or generating masked input segments conditioned on the unmasked ones, to obtain training data from unlabeled corpora. These approaches make it possible to use a tremendous amount of unlabeled data on the web to train large networks and solve complicated tasks. ELMo, BERT, and GPT in NLP are famous examples in this direction. However, there is still much unexplored territory in the research direction for self-supervised learning. This talk will introduce self-supervised learning, and show how to apply this technology to speech processing tasks.
Luba Tang

Luba Tang The anchor while both hardware and software are changing

100-day demand of computer performance doubles while one-year improvement of computer performance downs to 3%. For fitting the high demand of computer performance, it comes a new golden age for computer architecture. In this talk, we will introduce the challenges coming from both hardware and software sides. And then offering our solutions addessing these great challenges.
Chia-Liang Kao

Chia-Liang Kao Productionizing AI and MLOps

As AI models are being deployed in business across industries, operationalizing the models and AI-powered software at scale becomes challenging. "MLOps" emerges to combine machine learning and the principals of the "DevOps" software practices, in a new normal that models are commodities. This talk investigates the following challenges for productionizing AI models: •unifying exploration, training, production environments •automated/continuous retraining •model deployment management: monitoring, updating
James C.-H. Song

James C.-H. Song R&D Experience of Smart Medical Products

My presentation will start with an overview of Quanta’s smart medical related products.  As an R&D leader, I will then share the development process of these products, which includes “A”: algorithm, “B”: big data, “C”: cloud, as well as “D”: device.  The presentation will conclude with how and why that we think as an IT R&D engineers, we could make some contribution to the medical industry by our AI technologies.
Allan Yang

Allan Yang Industrial Digital Transformation with AIoT and WISE-PaaS

AI has proven to be of tremendous effectiveness in solving problems across many different domains, but the AI applications in many industrial scenarios often require data from secluded environments not easily accessible, and effective model trainings often require labeling that needs professional domain know-how’s. Advantech WISE-PaaS, a cross-IaaS industrial IoT data platform for enabling industrial IoT cloud service powered with AI, is designed to facilitate both data collection and real time streaming from industrial edges and AI model trainings and deployment back to industrial fields. Through the ecosystem collaboration and Co-Creation Model, Advantech looks forward to endorsing the implementation of IIoT across a variety of industries and use the WISE-PaaS IoT software platform as a catalyst to enable more innovative win-win AIoT solutions.
David Hsu

David Hsu Tiny yet powerful: Hardware for TinyML

The convergence of AI, 5G, and IoT technologies is resulting in a need for real-time information and responses. Network latency, data privacy, and consumer expectations around user experience and environmental impact are driving a shift to computing at the endpoint. Combining on-device intelligence with AI and ML capabilities has the potential to unlock new use cases and applications. TinyML is a field of study in endpoint AI/ML that explores the types of models you can run on small, low-powered devices. This talk will touch on what makes TinyML particularly challenging, and how neural network models and hardware can be co-designed to meet these challenges. Then, there are some examples of actual tinyML hardware platforms will be introduced, and it will close with some predictions on where this is headed.
Su,Chun-Jung (James)

Su,Chun-Jung (James) AI in Government

善用資料與數位新興科技,提供政府服務、提升施政效能,是各國政府推動數位化治理的主要目的之一。本演講將透過國內稅務、關務、交通、辦公場域等面向之 AI 運用情形,以及國外相關應用案例,分享在當代人工智慧的浪潮之下,政府部門運用相關技術提升效能、創造公共價值的可能性。 Making good use of data and digital technology to provide high quality services, and to improve government efficiency, is one of the primary goals for governments which advance digital governance policies. This presentation will address the potential uses of AI in government for taxation, customs enforcement, traffic control, and other areas in Taiwan and other countries, and show the possibilities of improving efficiency and creating public value.
Ming-Shiang Wu

Ming-Shiang Wu Paradigm Shift of Medicine: from Shallow Medicine to Deep Medicine

Medicine is a science of uncertainty but art of probability. The same symptoms may result from different diagnosis. The same diagnosis may need different treatment modalities. The same therapeutic maneuver may lead to different prognosis. The uncertainty of diagnosis, therapy and prognosis stems from that we do not understand the etiology and pathogenesis. It also arises from our incomplete enrollment of patient’s data. Take peptic ulcer as an example, stress and acid are considered the etiologic factors until the discovery of H. pylori. We cannot cure peptic ulcer by anxiolytics or acid-reducing agents but we can cure it by 2 weeks of antibiotics. The progress of genomic medicine was assumed to be the key for individual variability in manifeatation of diseases and the solution for precision medicine, especially in the diagnosis and management of malignant neoplasms. However, the gut microbiota, so-called secondary genome, can influence host’s health and disease treatment through bacteria-derived metabolites, immunity and gut-brain axis. Accordingly, investigation of host genomics and microbiomics simultaneously could lead to a better understanding of health and diseases. The big data analysis is mandatory for multiomics research. Therefore, machine learning or artificial intelligence becomes important research tools for precision or smart medicine. Deep phenotyping and deep learning are the root of deep medicine. Compared with current shallow medicine, deep medicine can provide a deeper insight into the etiology and pathogenesis through comprehensive data collection and analyses. Collectively, my planned courses will be “Paradigm shift of medicine: form shallow medicine to deep medicine”. The content will include (a)Evolution of etiology, pathogenesis, diagnosis and treatment of peptic ulcer and gastric cancer. (b)Methologic progress of molecular epidemiology, clinical trial, genomic medicine, microbiomics, big data analyses and artificial intelligence.
Jyh-Shing Roger Jang

Jyh-Shing Roger Jang The Beauty and Sorrow of AI Projects in Financial Scenes

The application of AI in financial scenarios can be said to be far-reaching, and its importance is well known, but what problems will actually be encountered when implementing AI projects? We will take several AI projects as examples to illustrate the beauty (vision) and sorrow (pain points) experienced by these projects from conception, design, modeling, to implementation, and maintenance. These issues that will not be mentioned in the textbook are the key to determining whether an AI project can be successfully implemented and continuously improved in application scenarios.
James Shih

James Shih AI & IoT with Edge Computing

Edge Computing has started being in our daily life. Local machine needs to have real-time & precise responses to the changes in surrounding environment. With recent AI & IoT technology innovation, edge computing will enrich everyone’s life for work and at home. In this section, speaker will present what INTEL has invented/provided for Edge Computing with AI and how those application will empower the IoT markets.
Albert Liu

Albert Liu Deep Learning HW Design

With the Convolutional Neural Network (CNN) breakthrough in 2012, the deep learning is widely applied to our daily life, automotive, retail, healthcare and finance. In 2016, Alpha Go with Reinforcement Learning (RL) further proves new Artificial Intelligent (AI) revolution gradually changes our society, like personal computer (1977), internet (1994) and smartphone (2007) before. However, most of effort focuses on software development and seldom addresses the hardware challenges. This speech reviews various hardware designs range from CPU, GPU to NPU. New hardware can be evolved from those designs for performance and power improvement.
Wayne Lai

Wayne Lai Trust in Artificial Intelligence

AI is used in a range of applications, such as calculating the best travel route to take in real-time, predicting what customers will buy, identifying credit card fraud, helping diagnose disease, identifying people from photos, and enabling self-driving vehicles. KPMG provides insights into the key drivers of trust, community expectations and confidence in the regulation of AI, expectations of the management of societal challenges associated with AI, as well as Taiwanese current understanding and awareness of AI. Importantly, the findings provide a clear understanding of the practices and principles to responsibly develop and ethically deploy AI in society and the workplace.
Tian-Sheng Wu

Tian-Sheng Wu AI and Smart Manufacturing in AIDC

AIDC has laid out three AI-enabling paradigms—Intelligent Machinery, Intelligent Manufacturing, and Intelligent Management—to drive the applications of Artificial Intelligence in the Aerospace industry. Further, AIDC has developed in-house intelligent manufacturing and management system, iAIDC system, to integrate IIoT, Big data, CPS, Robots and other production systems. iAIDC in-house have been praised by renowned international aerospace customers, such as Boeing, Airbus, Rolls-Royce and Honeywell. In 2018 AIDC was granted Smart Machinery Golden Award by the Ministry Of Economic Affairs (MOEA). Since intelligent manufacturing was deploymend, AIDC has seen rising annual revenue and decreasing production cost. Overall enterprise performance and value-added rate have been steadily improved. To AIDC, intelligent manufacturing is not only taking place now but will continually adapt and evolve into the future as well. Most importantly, iAIDC experience and management model will be downstream replicated, driving paradigm shift in aerospace suppliers. It is envisioned that AIDC and its partners as well as suppliers will together embark on the journey of digital transformation, and grow together to create a new horizon in the years to come.
Peter Hu

Peter Hu FET on the Journey from Big Data to AI

Utilizing big data and AI technology, Far EasTone Telecommunications (FET) has been innovating and striving for breakthroughs in its internal integration and transformation experience. The "Mobile Circle" app provides personalized food, clothing, housing, transportation, entertainment, and consumption integration services, and a number of 5G smart products (Big Data, AI, and IoT) have been successfully exported and applied to a number of government and corporate smart transformation projects. FET 5G has also won three world firsts in download speed, upload speed and video experience in the recent Opensignal international authoritative global competition! FET has proven and realized that FET has the ability to continue to improve and explore new business opportunities with customers!
Brian Yang

Brian Yang Artificial Intelligence, Real Money:AI and deep learning in Martech

TenMax works with Gojek, the Indonesian super app, to leverage insights from the Gojek ecosystem which provides ride sharing, food delivery, grocery shopping and payment services, and roll out an AI-based advertisement service platform,to accommodate the marketing needs of millions of brands and merchants in the Gojek supply chain. In this talk, Brian Yang will provide a high level overview of the e-commerce and digital marketing landscape in South East Asia,the technology stacks to build an effective Martech platform, and how deep learning is applied in TenMax.
Howard Hsieh

Howard Hsieh Embedding AI as the Sustainable Core Competence for Enterprises (18:00~18:15)

模型驗證成功不會達成 ROI。AI 不僅要落地,還要能平展與維運,才能成為可維持的競爭力。本報告嘗試說明企業應該要將 AI 視為基礎架構,而不僅是應用服務,循序漸進並且環環相扣的奠定基礎。同時以群創經驗說明必須的準備,將面對的挑戰以及機會。 Successful model verification will not achieve ROI. AI must not only be implemented, but also be able to continuously upgrade and maintain online to become the sustainable core competence for enterprises. This report attempts to explain that companies should treat AI as an infrastructure, not just an application service, to lay the foundation step by step and fully connected. The necessary preparations, challenges and opportunities along the way will be discussed with the experience from Innolux as well.
Friedman Wang

Friedman Wang Opportunities and Limitations of +AI Transformation in Financial Industry (18:15~18:30)

The reformation and transformation of financial industry are usually driven by emerging technologies like mainframe technology, the Internet, and smart mobility. Every technological advance has catalyzed a new wave of innovations/opportunities and created new profitable business models. The banking services become more and more convenient and can be available anytime, everywhere. The most important technology of our era is: Artificial Intelligence (AI). After exploring +AI 1.0 within the financial industry and the baptism of competition from external FinTech players, the financial industry has moved towards a more pragmatic +AI 2.0 integration and the launch of the future expansion of +AI 3.0. In this talk, we will share the experience about opportunities and limitations on the road to +AI reformation and transformation in CTBC Financial Holding Company. 金融業 +AI 轉型的機會與限制:歷次金融業改革與轉型的根本動力一直伴隨著當代的技術創新,包括了主機技術、網際網路和行動通訊。每一個技術創新都催化了互補創新和機遇的浪潮,創造了可營利的新業務模式與方法,讓金融業持續的往無時差、無疆界與更普及的理念前進。我們這個時代最重要的通用技術是:人工智慧(AI)。金融業內部經過 +AI 1.0 的探索,以及外部金融科技業者的競爭洗禮,已走向更務實的 +AI 2.0 融合,以及啟動未來 +AI 3.0 的擴大,我們將分享中信金控經驗在這段 +AI 改革與轉型道路上所看到的機會與限制。
Darren Chen

Darren Chen Digital Transformation for Traditional Manufacturing (20:10~20:25)

1. 傳統製造業為何需要數位轉型?Why does traditional manufacturing need digital transformation? 2. 傳統製造業如何啟動數位轉型?How does traditional manufacturing initiate digital transformation? 3. 明昌數位轉型藍圖與案例 Digital transformation roadmap of Machan International Co., Ltd.
Chomin Chiu

Chomin Chiu Challenges of Developing Smart Manufacturing Factories (20:25~20:40)

1. 智慧製造階段,循序漸進,聚焦投資 Smart manufacturing stage, step by step, focusing on investment 2. 流程資源優化 Process resource optimization 3. 製造資料 / 人機協作以及產品革新 Manufacturing data/human-machine collaboration and product innovation
頁數: 1 2 3 - 每頁 20 筆