BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:Asia/Taipei BEGIN:STANDARD DTSTART:19790930T230000 TZOFFSETFROM:+0900 TZOFFSETTO:+0800 TZNAME:CST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20250602T031339Z UID:54B86A46-0568-413F-969D-84AE293FE68E DTSTART;TZID=Asia/Taipei:20250529T220000 DTEND;TZID=Asia/Taipei:20250530T120000 DESCRIPTION:Title: AI Everywhere: Past\, Present and Future\n\nRecently\, a rtificial intelligence (AI) has gone through tremendous progress spurred b y hardware and software support in machine learning (ML) computation using neural networks (NNs). Based on Komogorov-Arnold Representation Theorem (1957)\, a multivariate function can be exactly expressed as a superpositi on of a finite number of outer functions with another linear combination o f inner functions embedded within. Cybenko (1989) then developed a Univers al Approximation Theorem showing such a function can be closely approximat ed by a superposition of sigmoid functions used as cascade inner and outer structures in NN representations. Moreover\, The Nobel Prize in Physics 2 024 was awarded to Hopfield and Hinton\, two neural network pioneers. At a global level\, the US proposed an AI Manhattan Initiative and a $500B Sta rgate investment to strengthen AI infrastructures\, EU started a Stargate- like effort\, Taiwan will increase computing power to 1200 pentaflops by 2 029\, and Japanese AI is reshaping economy and industry.\n\nIn this talk\, we will begin with a discussion on three pre-AI efforts to imitate human intelligence via teaching machines to speak\, listen and translate. All th ree are related to speech and language processing\, capabilities unique to human beings. Next\, we examine AI-based applications\, such as ChatGPT\, Gemini\, Gork\, Copilot\, NLLB-200\, and DeepSeek\, that are now being pr acticed by a wide spectrum of daily users. Although most AI computation is carried out by large computing farms on clouds\, low-cost\, edge AI is sl owly emerging. Next\, we believe knowledge-driven AI and domain-specific t ransfer learning are future directions. Responsible AI and trustworthy AI will be globally governed to secure AI usage. Finally\, AI personalization will be widely deployed to deliver true values to our society.\n\nCo-spon sored by: Li-Chun Wang\n\nRoom: 108\, Bldg: ED\, NYCU\, Hsinchu\, T'ai-pei \, Taiwan\, Virtual: https://events.vtools.ieee.org/m/484264 LOCATION:Room: 108\, Bldg: ED\, NYCU\, Hsinchu\, T'ai-pei\, Taiwan\, Virtua l: https://events.vtools.ieee.org/m/484264 ORGANIZER:ie3taipeisection@gmail.com SEQUENCE:62 SUMMARY:2025 Invited Talk: Prof. Chin-Hui Leeļ¼ŒAI Everywhere: Past\, Prese nt and Future URL;VALUE=URI:https://events.vtools.ieee.org/m/484264 X-ALT-DESC:Description: <br /><p class="MsoNormal"><strong style="mso-bidi- font-weight: normal\;"><span lang="EN-US" style="font-size: 11.0pt\; font- family: 'Arial'\,sans-serif\;">Title:&nbsp\;</span></strong><em><span lang ="EN-US" style="font-size: 11.0pt\; font-family: 'Arial'\,sans-serif\;">AI Everywhere: Past\, Present and Future</span></em></p>\n<p class="MsoNorma l"><span lang="EN-US" style="font-size: 10.0pt\; font-family: 'Arial'\,san s-serif\;">Recently\, artificial intelligence (AI) has gone through tremen dous progress spurred by hardware and software support in machine learning (ML) computation using neural networks (NNs). <span style="mso-spacerun: yes\;">&nbsp\;</span>Based on Komogorov-Arnold Representation Theorem (195 7)\, a multivariate function can be exactly expressed as a superposition o f a finite number of outer functions with another linear combination of in ner functions embedded within. Cybenko (1989) then developed a Universal A pproximation Theorem showing such a function can be closely approximated b y a superposition of sigmoid functions used as cascade inner and outer str uctures in NN representations. Moreover\, The Nobel Prize in Physics 2024 was awarded to Hopfield and Hinton\, two neural network pioneers. At a glo bal level\, the US proposed an AI Manhattan Initiative and a $500B Stargat e investment to strengthen AI infrastructures\, EU started a Stargate-like effort\, Taiwan will increase computing power to 1200 pentaflops by 2029\ , and Japanese AI is reshaping economy and industry.</span><span style="fo nt-family: Arial\, sans-serif\; font-size: 10pt\; text-align: justify\;">& nbsp\;</span></p>\n<p class="MsoNormal" style="text-align: justify\; text- justify: inter-ideograph\;"><span lang="EN-US" style="font-size: 10.0pt\; font-family: 'Arial'\,sans-serif\;">In this talk\, we will begin with a di scussion on three pre-AI efforts to imitate human intelligence via teachin g machines to speak\, listen and translate. All three are related to speec h and language processing\, capabilities unique to human beings. Next\, we examine AI-based applications\, such as ChatGPT\, Gemini\, Gork\, Copilot \, NLLB-200\, and DeepSeek\, that are now being practiced by a wide spectr um of daily users. Although most AI computation is carried out by large co mputing farms on clouds\, low-cost\, edge AI is slowly emerging. Next\, we believe knowledge-driven AI and domain-specific transfer learning are fut ure directions. Responsible AI and trustworthy AI will be globally governe d to secure AI usage. Finally\, AI personalization will be widely deployed to deliver true values to our society.</span></p> END:VEVENT END:VCALENDAR