HOME > >
±âÁ¶°¿¬ ¹× Ưº°Åä·Ð
¹ßÇ¥¿ä¾à
ÇÇÁöÄÃAI Á¤ºÎÁ¤Ã¥ ¹æÇâ
▪ MARYLAND(¸Þ¸±·£µå ´ëÇÐ) Ŭ¶óÅ© ¼®Á±³¼ö*, ±â°è°øÇаú ±³¼ö, »ê¾÷ÀΰøÁö´É¼¾ÅÍ ¼ÒÀå
- ¸Þ¸±·£µå ·Îº¸Æ½½º ¼¾ÅÍ(Maryland Robotics Center)
- ¸®½ºÅ© ¹× ½Å·Ú¼º ¼¾ÅÍ(Center for Risk and Reliability)
▪ Physical AI ´ë°¡
▪ ¿¬±¸ºÐ¾ß : »ê¾÷¿ë ÀΰøÁö´É(Industrial AI), »ê¾÷ ºòÅ×ÀÌÅÍ(Industrial Big Data), ½º¸¶Æ® Á¦Á¶ /Industry 4.0 ¿¬°è
AI ¼Ö·ç¼Ç, µðÁöÅÐ Æ®À© ÀüÀÌÇнÀ
▪ °æ·Â : ½º¸¶Æ® Á¦Á¶ ¹× ´ëÇü °øÀå Àüȯ ÇÁ·ÎÁ§Æ® ÁÖµµ
: 16³â ½º¸¶Æ® Á¦Á¶ ºÐ¾ß ¼±ÁöÀÚ 30ÀÎ
: 20³â ½º¸¶Æ® Á¦Á¶ ºÐ¾ß¿¡¼ °¡Àå ¿µÇâ·Â ÀÖ´Â ±³¼ö 20ÀÎ
: 22³â SME Eli Whitney »ý»ê¼º»ó ¹× SME/NAMRC S.M. Wu ¿¬±¸ ±¸Çö»ó ¼ö»ó
: 23³â ¿ì¼ö ³í¹®»ó(Outstanding Paper Award) ¼ö»ó
* Clark Distinguished Chair Professor(Ŭ¶óÅ© ¼®Á±³¼ö): ¸Þ¸±·£µå´ëÇб³ Ŭ¶óÅ© °ø°ú´ëÇÐ(Clark School of
Engineering)¿¡¼ ¿î¿µÇÏ´Â ÃÖ°í ¼öÁØÀÇ ¸í¿¹ ±³¼öÁ÷ ÇÁ·Î±×·¥À¸·Î, °øÇко߿¡¼ Ź¿ùÇÑ ¾÷Àû°ú ³ôÀº Çй®Àû ¿µÇâ·Â,
±¹Á¦Àû °¡½Ã¼º°ú ÀÎÁöµµ¸¦ Áö´Ñ Á¤±³¼ö¿¡°Ô ¼ö¿©µÊ.
-
Dr. Jay Lee is Clark Distinguished Professor and Founding Director of Industrial AI Center in the Mechanical Engineering of the Univ. of Maryland College Park. His current research is focused on developing non-traditional machine learning technologies including transfer learning, domain adaptation, similarity-based machine learning, stream-of-x machine learning, as well as industrial large knowledge model (ILKM), etc.
Previously, he was the founding director of National Science Foundation (NSF) Industry/University Cooperative Research Center (I/UCRC) on Intelligent Maintenance Systems (www.imscenter.net) in partnership with over 130 global company members and the Center was selected as the most economically impactful I/UCRC in the NSF Economic Impact Study Report in 2012. He mentored his students won 1st prize on Data Challenge Competition (PHM Society) 6 times during 2008-2023 as well as developed a number of start-up companies including Predictronics through NSF iCorps in 2013.
He is a member of Global Future Council on Advanced Manufacturing and Production of the World Economics Council (WEF), a member of Board of Governors of the Manufacturing Executive Leadership Council of National Association of Manufacturers (NAM), Board of Trustees of MTConnect, as well as a senior advisor to McKinsey.
He served as Vice Chairman and Board Member for Foxconn Technology Group (during 2019-2021 and had advised Foxconn business units to successfully receive six WEF Lighthouse Factory Awards. He also served as Director for Product Development and Manufacturing at United Technologies Research Center (now Raytheon Technologies Research Center) as well as Program Director for a number of programs at NSF.
He is fellow of ASME, SME, PHM Society, ad ISEAM and currently serves as Editor-in-Chief for IOP Journal on Machine Learning: Engineering. He was selected as 30 Visionaries in Smart Manufacturing in by SME in Jan. 2016 and 20 most influential professors in Smart Manufacturing in June 2020, and received SME Eli Whitney Productivity Award and SME/NAMRC S.M. Wu Research Implementation Award in 2022. His book on Industrial AI was published by Springer in 2020.
¹ßÇ¥¿ä¾à
Trends of AI Industry, Physical AI, and Recent Advances of Industrial AI
▪ UCLA (University of California, Los Angeles) ±³¼ö
▪ ¼¼°èÀûÀÎ ·Îº¿ ¿¬±¸ÀÇ ¸ÞÄ« UCLA ¡®RoMeLa(Robotics and Mechanisms Laboratory)¡¯ ·Îº¿ ¿¬±¸¼ÒÀå
▪ ¿¬±¸ºÐ¾ß : ÀÚÀ²½Ã½ºÅÛ, Àΰ£Çü ·Îº¿, ¿îµ¿ÇÐ, ·Îº¿ ±â°è°øÇÐ µðÀÚÀÎ, ·Îº¿ ¿îµ¿ ±â°èÇÐ
▪ °æ·Â : ¹Ì±¹ ÃÖÃÊÀÇ ÈÞ¸Ó³ëÀÌµå ·Îº¿ ¡®Âû¸®¡¯, Àü ¼¼°è ±³À°¡¤¿¬±¸¿ë ¿ÀǼҽº·Î °ø°³ÇÑ ¡®´ÙÀ©-OP¡¯, ÃÖ±Ù ÀÌÁ· º¸Çà
±â¼ú¿¡¼ ¶Ù¾î´Â ¼º°ú¸¦ º¸ÀÎ ¡®¾Æ¸£Å׹̽º¡¯ µî ´Ù¾çÇÑ Çõ½ÅÀû ·Îº¿ °³¹ß
: ¹Ì±¹¸³°úÇÐÀç´Ü(NSF) ¡® ŒÀº °úÇÐÀÚ»ó¡¯, GM ¡® ŒÀº ¿¬±¸ÀÚ»ó¡¯, ¡®¹Ì±¹ÀÚµ¿Â÷°øÇÐȸ(SAE) ±³À°»ó¡¯ µî ¼ö»ó
: ÆÄǽ·¯»çÀ̾𽺰¡ ¼±Á¤ÇÑ ¡® ŒÀº õÀç °úÇÐÀÚ 10ÀΡ¯¿¡ À̸§À» ¿Ã¸²
: ¼¼°è ÃÖÃÊ·Î ½Ã°¢Àå¾ÖÀÎÀÌ ¿îÀüÇÒ ¼ö ÀÖ´Â ¹«ÀÎ ÀÚµ¿Â÷ °³¹ßÇØ ¿ö½ÌÅÏÆ÷½ºÆ®·ÎºÎÅÍ ¡®´Þ ªm·ú¿¡ ¹ö±Ý°¡´Â ¼º°ú¡¯
¶ó´Â Æò°¡¸¦ ¹Þ¾Ò°í, ·Î¸á¶ó ·Îº¿¿¬±¸¼Ò´Â ±¹Á¦ ·Îº¿ÄÄ¿¡¼ ÃÑ 6ȸ ¿ì½Â Â÷Áö
¹ßÇ¥¿ä¾à
Physical AI: Where Hardware Meets Intelligence Mobility shows ability. Interaction delivers utility.
³ëŰ¾Æ ³×Æ®¿öÅ© ÀÎÇÁ¶ó ¾Æ½Ã¾Æ-ÅÂÆò¾ç ½Å¼ºÀå ºÐ¾ß¸¦ ÃѰýÇϰí ÀÖÀ¸¸ç, Á¤ºÎ ¹× ´ë±â¾÷ÀÌ ÀΰøÁö´É ±â¹ÝÀÇ Ã·´Ü ÀÚµ¿È, Ŭ¶ó¿ìµå, º¸¾È ±â¼úÀ» µµÀÔÇÏ¿© µðÁöÅÐ Àüȯ°ú ¹Ì·¡ Çõ½ÅÀ» ½ÇÇöÇÒ ¼ö ÀÖµµ·Ï Áö¿øÇÕ´Ï´Ù
µðÁöÅÐÈ¿Í ÀΰøÁö´É ¿¬°è°¡ ºü¸£°Ô È®»êµÊ¿¡ µû¶ó, °ß°íÇϰí ÀÚÁÖÀûÀ̸ç È®Àå °¡´ÉÇÑ ³×Æ®¿öÅ© ±¸Á¶¿¡ ´ëÇÑ ¼ö¿ä°¡ ±× ¾î´À ¶§º¸´Ù ³ô¾ÆÁö°í ÀÖ½À´Ï´Ù. ¿¬°áÇü AI, µðÁöÅÐ Æ®À©, ±×¸®°í ¹°¸®Àû AI°¡ ºÎ»óÇϸç, ÀÎÇÁ¶óÀÇ ¼³°è¡¤±¸Ãࡤ¿î¿µ ¹æ½ÄÀÌ Çõ½ÅÀûÀ¸·Î º¯ÈÇϰí ÀÖ½À´Ï´Ù.
±×ÀÇ ÁÖ¿ä Ã¥ÀÓ¿¡´Â HPC/AI µ¥ÀÌÅͼ¾ÅÍ ³×Æ®¿öÅ·, Å×¶óºñÆ®±Þ ³×Æ®¿öÅ·, ÆÐºê¸¯ ¹× ÀÎÅÍÄ¿³ØÆ® µî ÷´Ü ³×Æ®¿öÅ© ¼Ö·ç¼ÇÀÌ Æ÷ÇÔµÇ¸ç ³×Æ®¿öÅ©ÀÇ ¹Ì·¡ ÁøÈ¿Í °ü·ÃÇÏ¿© º§ ¿¬±¸¼Ò¿Í Çù¾÷ÇÏ¿´½À´Ï´Ù. ±×´Â ±Û·Î¹ú Á¦Á¶ »ê¾÷ ÃѰý, ¾Æ½Ã¾ÆÅÂÆò¾ç ´ë±â¾÷ ¹× À¥½ºÄÉÀÏ·¯ ºÎ¹® ºÎ»çÀå, ½Ì°¡Æ÷¸£ °í°´ CTO µî ´Ù¾çÇÑ ÇÙ½É ¸®´õ½ÊÀ» ¿ªÀÓÇÑ ¹Ù ÀÖ½À´Ï´Ù.
-
Philippe leads Nokia¡¯s Emerging Segments across Asia-Pacific for Nokia Network Infrastructure. He specializes in guiding governments and large enterprises through the adoption of Intelligent Automation, Cloud, and Security advancements to future-proof their AI-driven modernization efforts.
As digitalization and AI integration accelerate, the demand for robust, sovereign, and scalable network architectures has never been greater. The rise of connected AI, digital twins, and physical AI is reshaping how infrastructure is designed,deployed, and operated.
Philippe focuses on advanced network solutions, including HPC/AI data center networking and terabit-scale architecture. He engages with Nokia Bell Labs and customers to bring advanced use cases closer to deployment accelerating innovation across industry and scientific domains. He has held key leadership roles such as Global Head of Manufacturing Segment, APAC Vice President for Large Enterprise and Webscalers, and Customer CTO for Singapore.
¹ßÇ¥¿ä¾à
Realizing Physical AI: Network Evolution with Nokia Bell Labs Vision 2030
´ëÇѹα¹Àº ±¹°¡ ÀÎÇÁ¶ó¿Í »ê¾÷ ÀÚµ¿È Àü¹Ý¿¡ °ÉÃÄ Çö½Ç ¼¼°èÀÇ AI ½Ã½ºÅÛÀ» ±¸ÇöÇϱâ À§ÇÑ Àü·«Àû ¿©Á¤À» ½ÃÀÛÇϰí ÀÖ½À´Ï´Ù. ÀÌ·¯ÇÑ º¯È´Â ÃÊÀúÁö¿¬, ÀÚ±¹ ÁÖ±ÇÇü AI ÄÄÇ»ÆÃ, ȯ°æ ¼¾½Ì, ±×¸®°í ¾ÈÀüÇÏ°í ¹Ì¼Ç Å©¸®Æ¼ÄÃÇÑ ¿î¿µÀ» Áö¿øÇϴ ÷´Ü ³×Æ®¿öÅ© ÀÎÇÁ¶ó¿Í Ŭ¶ó¿ìµå Ç÷§Æû¿¡ ±â¹ÝÇϰí ÀÖ½À´Ï´Ù.
1990³â °í·Á´ëÇб³ ±â°è°øÇаú °øÇлç ÇÐÀ§¸¦ ÃëµæÇÑ ÈÄ Çö´ëÀÚµ¿Â÷¿¡ ÀÔ»çÇÏ¿©, »ý±â°èȹÆÀÀå, ¾Æ»ê»ý±â½ÇÀå, Áß³²¹Ì ±Ç¿ª°øÀå ½ÇÀåÀ» ¿ªÀÓÇϰí ÇöÀç Á¦Á¶¼Ö·ç¼Çº»ºÎÀÇ E-FOREST¼¾ÅÍÀåÀ» ´ã´çÇϰí ÀÖ´Ù.
1990. 2 °øÇлç, °í·Á´ëÇб³ ±â°è°øÇаú
1992. 1 Çö´ëÀÚµ¿Â÷ ÀÔ»ç
2016. 1 ~ 2017.12 Çö´ëÀÚµ¿Â÷ »ý±â°èȹ ÆÀÀå
2018. 1 ~ 2019.10 Çö´ëÀÚµ¿Â÷ ¾Æ»ê»ý±â½ÇÀå
2019.11 ~ 2023.12 Áß³²¹Ì±Ç¿ª°øÀå ½ÇÀå / »ó¹«
2024. 1 ~ ÇöÀç Çö´ëÂ÷/±â¾Æ E-FOREST¼¾ÅÍÀå / »ó¹«
-
After obtaning a Bachelor of Engineering degree in Mechanical Engineering from Korea University in 1990, he joined Hyundai Motor Company. He has held positions as Head of the Production Planning Team, Director of the Asan Production Technology Office, and Director of the Central and South America Regional Plant. He is currently serving as the Head of the E-FOREST Center within the Manufacturing Solutions Division.
¹ßÇ¥¿ä¾à
Journey to DF247: Our Path to Achieving Manufacturing Innovation
Journey to DF247, Á¦Á¶Çõ½Å ´Þ¼ºÀ» À§ÇÑ ¿ì¸®ÀÇ ¿©Á¤
°¿¬ ¹× Åä·Ð
±âȹ¼¼¼Ç 1-1
Physical-Digital Intelligence
ÇÇÁöÄÃ-µðÁöÅÐ ÀÎÅÚ¸®Àü½º
ÀÌÁ¾¼®Àº KAIST »ê¾÷¹×½Ã½ºÅÛ°øÇаú ±³¼ö·Î ÀçÁ÷ ÁßÀ̸ç, Á¦Á¶ µµ¸ÞÀÎÀÇ Æ¯¼ºÀ» ¹Ý¿µÇÑ ÀΰøÁö´É ÇнÀ°ú ¸ðµ¨ ±â¹Ý ÃÖÀûÈ ¹æ¹ý·Ð °³¹ßÀ» Áß½ÉÀ¸·Î ¿¬±¸¸¦ ¼öÇàÇϰí ÀÖ´Ù. ö°, ¹ÝµµÃ¼, µð½ºÇ÷¹ÀÌ, ¹èÅ͸® µî ´Ù¾çÇÑ Á¦Á¶ ºÐ¾ß¿¡¼ Æ÷½ºÄÚ, LG¿¡³ÊÁö¼Ö·ç¼Ç, »ï¼ºÀüÀÚ µî ÁÖ¿ä ±â¾÷µé°ú »êÇÐÇù·Â ÇÁ·ÎÁ§Æ®¸¦ ¼º°øÀûÀ¸·Î ¼öÇàÇÏ¿© ÀÚÀ²Á¦¾î ±â¼úÀÇ ÇöÀå Àû¿ë °æÇèÀ» ÃàÀûÇÏ¿´´Ù. ƯÈ÷ ÀΰøÁö´É ±â¹Ý ÃÊÁ¤¹Ð µµ±Ý Á¦¾î ±â¼úÀ» °³¹ßÇÏ¿© ±¹°¡Çٽɱâ¼ú·Î ÁöÁ¤µÇ´Â ¼º°ú¸¦ ÀÌ·ç¾ú´Ù. KAIST ±³¼öÁøÀ¸·Î ÇÕ·ùÇϱâ Àü¿¡´Â ¼º±Õ°ü´ëÇб³ ½Ã½ºÅ۰濵°øÇаú ±³¼ö, ¹Ì±¹ SAS º»»ç¿¡¼ ÇÁ·Î±×·¡¸Ó·Î ±Ù¹«Çß´Ù. ¾ÆÀÌ¿À¿ÍÁÖ¸³´ëÇб³¿¡¼ »ê¾÷°øÇÐ ¹Ú»çÇÐÀ§¸¦, Æ÷Ç×°ø°ú´ëÇб³¿¡¼ Á¤º¸Åë½Å ¼®»çÇÐÀ§¸¦ ¹Þ¾Ò´Ù.
-
Jong-Seok Lee is a Professor in the Department of Industrial & Systems Engineering at KAIST. His research focuses on developing AI learning frameworks and model-based optimization methodologies that reflect the unique characteristics of manufacturing domains. He has successfully led multiple industry-academia collaboration projects with major manufacturing companies such as POSCO, LG Energy Solution, and Samsung Electronics, across various manufacturing sectors, including steel, semiconductors, displays, and batteries, building extensive experience in deploying autonomous control technologies in real production environments. In particular, he developed an AI-based coating weight control technology that was designated as a National Core Technology. Before joining KAIST, he was a faculty member in the Department of Systems Management Engineering at Sungkyunkwan University and worked as a programmer at SAS Institute Headquarters in the United States. He holds a Ph.D. in Industrial Engineering from Iowa State University and an M.S. in Information Technology from POSTECH.
¹ßÇ¥¿ä¾à
-Pai Zheng (SM¡¯IEEE, M¡¯SME/ASME, AM¡¯CIRP) is currently an Associate Professor, Wong Tit-Shing Endowed Young Scholar in Smart Robotics, and Associate Director of PolyU-Wuxi Innovation and Technology Research Institute, in the Department of Industrial and Systems Engineering, at The Hong Kong Polytechnic University (PolyU). Before joining PolyU, he has been a Research Fellow at Nanyang Technological University, Singapore (2018-2019). He received the Dual Bachelor¡¯s Degrees in Mechanical Engineering (Major) and Computer Science and Engineering (Minor) from Huazhong University of Science and Technology, Wuhan, China, in 2010, the Master¡¯s Degree in Mechanical Engineering from Beihang University, Beijing, China in 2013, and the Ph.D. degree in Mechanical Engineering at The University of Auckland, Auckland, New Zealand in 2017. His research interests include human-robot collaboration, smart product-service systems, and industrial AI. Dr Zheng is a recipient of the Clarivate High Cited Researchers (2025 - ), HKIE Young Engineer of the Year Award (2025), SME Outstanding Young Manufacturing Engineers Award (2024), NSFC Excellent Young Scientist Fund (2024), PolyU Young Innovative Researcher Award (2023), and Global Top 50 AI+X Chinese Scholars by Baidu (2022). He serves as the Chair of IEEE Hong Kong Section SMC Chapter, Co-chair of IEEE TC on DMHCA, Scientific Committee Member of SME | NAMRI, Associate Editor of Journal of Manufacturing Systems, IEEE Transactions on Automation Science and Engineering, Journal of Intelligent Manufacturing, Journal of Cleaner Production, and International Journal of Interactive Design & Manufacturing.
¹ßÇ¥¿ä¾à
Towards Embodied Robotic Manufacturing Systems: A Human-in-the-Loop Vision-Language Model and Contact-Rich Learning Approach
Anchored in the ethos of Industry 5.0, Human-Robot Collaboration (HRC) enables overall system performance and human well-beings by embracing the cutting-edge Artificial Intelligence (AI) and digital technologies. In such context, Vision-Language Models (VLMs), as a typical type of foundation models, have garnered significant attention and are extensively applied in HRC tasks by achieving remarkable outcomes. This talk proposes a physics-informed and VLM-enhanced human-in-the-loop approach for cost-effective robotic learn-to-assembly in smart manufacturing, of which three main aspects yet to be explored coherently: 1) Multimodal Intelligence-based Human Demonstration (Perception/Cognition); 2) Cross-Domain Tool-use Skill Transfer (Behavior Learning), and 3) Foundation-model based NoCode Manufacturing System Execution (Physics-informed Execution). It is assumed that the success of this research endeavor could potentially pave the way for more natural HRC and effective robot learning and manipulation in production, as the next-generation Embodied Robotic Manufacturing Systems, characterized by enhanced flexible automation capabilities.
¼ÕÁö¿¬ ½ÇÀåÀº Çѱ¹°úÇбâ¼ú¿ø Àü»êÇÐ ¹Ú»çÇÐÀ§¸¦ ¹Þ¾ÒÀ¸¸ç, ÇöÀç´Â Çѱ¹ÀüÀÚÅë½Å¿¬±¸¿ø Áö´É¡¤Á¦Á¶À¶ÇÕ¿¬±¸½Ç¿¡ ÀçÁ÷ ÁßÀÌ´Ù. ¹Ì·¡ Á¦Á¶ Çõ½ÅÀ» À§ÇÑ µðÁöÅÐ Æ®À©, AI·Îº¿ µî ÷´Ü ICT ºÐ¾ßÀÇ ´Ù¼ö ¿¬±¸°³¹ß °úÁ¦ ±âȹ¡¤¼öÇàÀ» ´ã´çÇϰí ÀÖ´Ù.
-
Jiyeon Son holds a Ph.D. in Computer Science from the Korea Advanced Institute of Science and Technology (KAIST) and is currently employed at the ICT-enabled Intelligent Manufacturing Research Section of the Electronics and Telecommunications Research Institute (ETRI). She is responsible for planning and executing numerous research and development projects in advanced ICT fields such as digital twins and AI robots for future manufacturing innovation.
¹ßÇ¥¿ä¾à
ÇÇÁöÄà AI¸¦ À§ÇÑ µðÁöÅÐ Æ®À©ÀÇ ¿ªÇÒ°ú ¿¬±¸ ¹æÇâ
ÇÇÁöÄà AI´Â ·Îº¿, ÀÚÀ²ÁÖÇà, Á¦Á¶ µî ¹°¸®Àû ½Ã½ºÅÛÀÌ È¯°æÀ» ÀÎÁöÇϰí ÀÚÀ²ÀûÀ¸·Î ÇൿÇϵµ·Ï ÇÏ´Â next AI±â¼úÀ̸ç, µðÁöÅÐ Æ®À©Àº ÀÌ·¯ÇÑ AI ½Ã½ºÅÛÀÇ ÇнÀ, °ËÁõ, ÃÖÀûȸ¦ À§ÇÑ ÇÙ½É '°¡»ó ½ÇÇè½Ç' ¿ªÇÒÀ» ÇÕ´Ï´Ù.
¹ßÇ¥¿ä¾à
ÇÇÁöÄÃAI¿Í µðÁöÅÐÆ®À©ÀÇ ¸¸³²
±âȹ¼¼¼Ç 1-2
Intelligent Robotics and Physical AI
Áö´ÉÇü ·Îº¿°ú ÇÇÁöÄà AI
³ë»óµµ ±³¼ö´Â ¼¿ï´ëÇб³ ±â°è¼³°èÇаú¿¡¼ »ý»ê°øÇÐÀ¸·Î ¹Ú»çÇÐÀ§¸¦ ¹Þ¾ÒÀ¸¸ç, ÇöÀç ¼º±Õ°ü´ëÇб³ °ø°ú´ëÇÐ »ê¾÷°øÇаú¿¡¼ ±³¼ö·Î ÀçÁ÷ ÁßÀÌ´Ù. ÁÖ¿ä ¿¬±¸ ºÐ¾ß´Â CAD/CAM/PLM, »ý»ê½Ã½ºÅÛ ¸ðµ¨¸µ&½Ã¹Ä·¹À̼Ç, ½º¸¶Æ®Á¦Á¶, ½º¸¶Æ®°øÀå, »çÀ̹ö¹°¸®½Ã½ºÅÛ°ú µðÁöÅÐÆ®À© µîÀÌ´Ù.
-
Sang Do Noh received his Ph.D. in mechanical design and production engineering from Seoul National University, Republic of Korea. He currently works as a professor in Department of Industrial Engineering at Sungkyunkwan University, Republic of Korea. His major research areas are CAD/CAM/PLM, modeling and simulation of manufacturing systems, smart manufacturing, smart factory, cyber-physical system and digital twin.
¹ßÇ¥¿ä¾à
-
Xi (Vincent) Wang is an Associate Professor in the IPU Department of Production Engineering, KTH Sweden. He is working as the division head of Industrial Production Systems (IPS). In 2021, Vincent received his Docentship from KTH. He received his PhD and Bachelor degrees in Mechanical Engineering from the University of Auckland (New Zealand) and Tianjin University (China), respectively in 2013 and 2008.
Vincent¡¯s main research focus includes cloud-based manufacturing, sustainable manufacturing, robotics, digital twin, computer-aided design, and manufacturing systems. He also serves as the Managing Editor of the International Journal of Manufacturing Research (IJMR), Associate Editor of SME Journal of Manufacturing Systems (JMS), and Array – Open Access Journal by Elsevier, and Editorial Board Member of three other international journals.
¹ßÇ¥¿ä¾à
Digital Twins, Human and Physical AI
Modern manufacturing demands higher performance, greater flexibility, improved quality, and a safer and more supportive working environment for human operators. It is therefore a natural progression to combine the strength, precision, and repeatability of industrial robots with the flexibility, adaptability, and cognitive capabilities of human workers. With the integration of digital twin (DT) and artificial intelligence (AI) technologies, it becomes possible to develop a Physical AI system that enables and enhances human-robot collaborative tasks in manufacturing settings.
±è´ëȯÀº LG»ý»ê±â¼ú¿ø ÀÔ»ç ÈÄ 22³â°£ ÀçÁ÷ ÁßÀ̸ç, ÀçÁ÷ Áß ¼¿ï´ëÇб³¿¡¼ ¹Ú»çÇÐÀ§¸¦ ¹ÞÀº ÈÄ AI±â¼úÀ» Á¢¸ñÇÏ¿© ´Ù¾çÇÑ ½º¸¶Æ®ÆÑÅ丮 ¼Ö·ç¼ÇÀ» °³¹ßÇϰí ÀÖ´Ù. ÃÖ±Ù¿¡´Â ·Îº¿°ú AI¸¦ Á¢¸ñÇÑ Physical AI ±â¼ú·Î Á¦Á¶ ¿µ¿ªÀÇ ÀÚµ¿È ¼Ö·ç¼ÇÀ» °³¹ßÇϰí È®»êÇØ ³ª°¡°í ÀÖ´Ù.
-
Leo Kim has been with LG Production engineering Research Institute for 22 years. While there, he received a doctorate in AI-based prognostics & health monitoring from Seoul National University. He has since developed various smart factory solutions by incorporating AI technology. Recently, he has been developing and expanding automation solutions in the manufacturing sector using Physical AI technology, which combines robotics and AI.
¹ßÇ¥¿ä¾à
Á¦Á¶AX ÀüȯÀ» À§ÇÑ Physical AI ±â¼ú Á¶¸Á
Èçµé¸®´Â ȯÀ²°ú Á¶Á¦ Á¤Ã¥À¸·Î ±Û·Î¹ú Á¦Á¶±â¾÷µéÀº º¯ÈÇϴ ȯ°æ¿¡¼ »ì¾Æ ³²±â À§ÇØ ¸öºÎ¸²À» Ä¡°í ÀÖ´Ù. Á¦Á¶ °æÀï·Â È®º¸¸¦ ÅëÇÑ Á¦Á¶ ¿ø°¡ Çõ½ÅÀº ¸ðµç Á¦Á¶±â¾÷ÀÇ ¼÷Á¦°¡ µÇ¾î ¿Ô°í À̸¦ À§ÇØ ½º¸¶Æ®ÆÑÅ丮¸¦ ÅëÇÑ ¹«ÀÎÈ °øÀå ±¸ÇöÀ» ³ë·ÂÇϰí ÀÖÀ¸³ª, ¿À·£±â°£ »ç¶÷ÀÌ ÀÛ¾÷Çϱ⠿ëÀÌÇÏ°Ô µÈ ÀÛ¾÷ȯ°æÀ» ·Îº¿ÀÚµ¿È·Î ÀüȯÇϱâ À§ÇÑ ¸¹Àº µµÀü °úÁ¦µéÀÌ ÀÖ´Ù. AI±â¼úÀÇ ¹ßÀü°ú ´õºÒ¾î Á¦Á¶ ÇöÀå¿¡´Â AI±â¼úÀ» Á¢¸ñÇÑ µðÁöÅÐ º¯ÇõÀÇ ½ÃµµµéÀÌ ÀÖ¾î ¿ÔÀ¸¸ç, Á¦Á¶ÇöÀåÀÇ ´Ù¾çÇÑ µ¥ÀÌÅ͸¦ ºÐ¼®ÇÏ¿© ÀÌ»óÀ» °¨ÁöÇÏ°í ¿¹ÁöÇÏ´Â ±â¼ú¿¡ ¸Ó¹«¸£°í ÀÖ¾úÀ¸³ª, ÃÖ±Ù Physical AI ±â¼úÀÇ µîÀåÀ¸·Î ½º¸¶Æ®ÆÑÅ丮 ¼Ö·ç¼ÇÀÇ »õ·Î¿î ´ë¾ÈÀ¸·Î µîÀåÇϰí ÀÖ´Ù. Á¦Á¶ ÇöÀå¿¡¼ ÃÖ±Ù ½ÃµµÇϰí ÀÖ´Â Physical AI ±â¼ú ÇöȲÀ» ¤¾î º¸°í, ÇѰèÁ¡°ú ¹Ì·¡ ±â¼ú ¹æÇâÀ» Á¶¸ÁÇØ º¸°íÀÚ ÇÑ´Ù.-
¹ßÇ¥¿ä¾à
÷´ÜÁ¦Á¶·Îº¿È°¿ë ÇÇÁöÄÃAI °øÁ¤¸ðµ¨ °³¹ß ¹× ½ÇÁõ
-±âȹ¼¼¼Ç 1-3
Convergence of Robotics and Physical AI
·Îº¸Æ½½º¿Í ÇÇÁöÄà AIÀÇ À¶ÇÕ
-
¹ßÇ¥¿ä¾à
--
¹ßÇ¥¿ä¾à
Measuring and Improving Generalization of Robot Policies
-
ÃÖ¼ºÁØÀº °í·Á´ëÇб³ ÀΰøÁö´ÉÇаú ºÎ±³¼öÀÌÀÚ RLWRLDÀÇ ¼ö¼®¿¬±¸¿ø(Principal Scientist)ÀÔ´Ï´Ù. °í·Á´ëÇб³¿¡ ºÎÀÓÇϱâ Àü¿¡´Â 2018³âºÎÅÍ 2020³â±îÁö ¹Ì±¹ ·Î½º¾ØÁ©·¹½º¿¡ À§Ä¡ÇÑ µðÁî´Ï ¸®¼Ä¡(Disney Research)¿¡¼ ¹Ú»çÈÄ¿¬±¸¿øÀ¸·Î ±Ù¹«ÇßÀ¸¸ç, 2018³â¿¡´Â Çѱ¹ÀÇ Ä«Ä«¿Àºê·¹ÀÎ(Kakao Brain)¿¡¼ ¿¬±¸¿øÀ¸·Î ÀçÁ÷Çß½À´Ï´Ù. Á¦ ÁÖ¿ä ¿¬±¸ ºÐ¾ß´Â Àΰ£-·Îº¿ »óÈ£ÀÛ¿ë(HRI), Á¶°Ç ±â¹Ý ¸ð¼Ç »ý¼º, ±×¸®°í ·Îº¸Æ½½º ºÐ¾ßÀÇ ÆÄ¿îµ¥ÀÌ¼Ç ¸ðµ¨ÀÔ´Ï´Ù. ¼¿ï´ëÇб³ Àü±â¡¤ÄÄÇ»ÅͰøÇкο¡¼ ¿À¼ºÈ¸ ±³¼ö´ÔÀÇ Áöµµ ¾Æ·¡ ¹Ú»ç ÇÐÀ§¸¦ ¹Þ¾ÒÀ¸¸ç, ±× ÀÌÀü¿¡´Â °°Àº Çб³¿¡¼ Àü±âÁ¤º¸°øÇкΠÇлç ÇÐÀ§¸¦ ÃëµæÇß½À´Ï´Ù. ÇкΠÀçÇÐ ½ÃÀý¿¡´Â ºÎõ¿¡ À§Ä¡ÇÑ ÇÑ¿ï·Îº¸Æ½½º¿¡¼ ¼ÒÇÁÆ®¿þ¾î ¿£Áö´Ï¾î·Î 3³â°£ ±Ù¹«ÇÏ¸ç ·Îº¿ Á¦¾î ¹× ¼ÒÇÁÆ®¿þ¾î °³¹ßÀ» ´ã´çÇß½À´Ï´Ù.
-
I am an Associate Professor at the Department of Artificial Intelligence at Korea University and a Principal Scientist at RLWRLD. Before joining Korea University, I was a Postdoctoral Researcher at Disney Research, Los Angeles, from 2018 to 2020, and a Research Scientist at Kakao Brain, Korea, in 2018. My research interests include, but are not limited to, the following areas: human-robot interaction (HRI), conditioned motion generation, and foundation models for robotics.
I previously received my Ph.D. degree in Electrical Computer Engineering at Seoul National University in Seoul, Korea, advised by Songhwai Oh. Before that, I received my Bachelor's degree in Electrical Engineering and Computer Sciences at Seoul National University. During my undergraduate years, I worked at Hanool Robotics in Bucheon, Korea, as a software engineer for three years.
¹ßÇ¥¿ä¾à
Àΰ£ Áß½ÉÀÇ ¹°¸®Àû Áö´ÉÀ» ÇâÇÏ¿©
À̹ø ¹ßÇ¥¿¡¼´Â ÀüÅëÀû ·Îº¿°øÇÐÀÇ °üÁ¡À» ±â¹ÝÀ¸·Î, ·Îº¿ÀÌ ´Ü¼øÇÑ ±â°è ÀåÄ¡¸¦ ³Ñ¾î Àΰ£°ú »óÈ£ÀÛ¿ëÇÏ´Â Áö´ÉÀû Á¸Àç·Î ¹ßÀüÇÏ´Â °úÁ¤À» ´Ù·é´Ù. ¹ßÇ¥´Â ¸ÕÀú »ýüÀû ÀûÀÀ¼ºÀ» °Á¶ÇÏ´Â Physical Intelligence¿Í ÀÎÁöÀû ±¸ÇöÀ» ÁöÇâÇÏ´Â Physical AIÀÇ Â÷À̸¦ ¤À¸¸ç, »ê¾÷¿ë ÀÚµ¿È¸¦ ³Ñ¾î ÀϹÝÈµÈ Áö´ÉÀÇ Çʿ伺À» Á¦½ÃÇÑ´Ù. ÀÌÈÄ °í·Á´ëÇб³ ·Îº¿Áö´É¿¬±¸½Ç¿¡¼ ÁøÇà ÁßÀÎ ¿¬±¸µéÀ» ¼Ò°³ÇÑ´Ù. ½ÇÆÐ Àνİú Ãß·Ð ´É·ÂÀ» °áÇÕÇÑ VINE ¸ðµ¨, ´ëÇü ¾ð¾î¸ðµ¨(LLM)À» Ȱ¿ëÇØ ¼º°Ý ±â¹Ý ÇൿÀ» »ý¼ºÇÏ´Â Æä¸£¼Ò³ª ·Îº¿ ½Ã½ºÅÛ, ±×¸®°í »óÈ£ÀÛ¿ë ¸ð¼ÇÀ» ÀÌÇØÇÏ°í »ý¼ºÇÏ´Â foundation model for HRI µîÀÌ ÁÖ¿ä »ç·Ê·Î ´Ù·ïÁø´Ù. ¸¶Áö¸·À¸·Î Brooks¿Í ClarkeÀÇ ¹ýÄ¢À» ÀοëÇϸç, ÁøÁ¤ÇÑ ¡®¹°¸®Àû Áö´É¡¯Àº ±â¼úÀû ¿Ï¼ºµµ¸¦ ³Ñ¾î Àΰ£ÀÇ ¸Æ¶ô ¼Ó¿¡¼ Á÷°üÀûÀ¸·Î ÀÌÇØµÇ°í »çȸÀûÀ¸·Î ÀûÀÀÇÏ´Â ·Îº¿ÀÇ ÇüÅ·Π±¸ÇöµÉ °ÍÀÓÀ» °Á¶ÇÑ´Ù.
±èÁØÇÏ ´ëÇ¥´Â KAIST ±â°è°øÇаú¿¡¼ ¹Ú»çÇÐÀ§¸¦ ¹ÞÀ¸¸ç º¹ÀâÇÑ È¯°æ¿¡¼ÀÇ ·Îº¿ »óÅ ÃßÁ¤°ú µ¿Àû ¿òÁ÷ÀÓ Á¦¾î ºÐ¾ßÀÇ Àü¹®¼ºÀ» ½×¾Ò½À´Ï´Ù. KAIST ÈÞ¸Ó³ëÀÌµå ·Îº¿ ¿¬±¸¼¾ÅÍ¿¡¼ ¿¬±¸¿øÀ¸·Î Ȱµ¿ÇÏ¸ç »çÁ·º¸Çà·Îº¿ ºÐ¾ß¿¡¼ ¼¼°è ÃÖ°í ¼öÁØÀÇ ¿¬±¸ ¼º°ú¸¦ ´Þ¼ºÇßÀ¸¸ç, IEEE T-RO, IEEE RA-L, IJRR µî Àú¸í ÇмúÁö¿¡ ´Ù¼öÀÇ ³í¹®À» ¹ßÇ¥Çß½À´Ï´Ù. ÇöÀç´Â 2024³â 3¿ù KAIST µ¿·áµé°ú ÇÔ²² â¾÷ÇÑ ÁÖ½Äȸ»ç µðµç·Îº¸Æ½½ºÀÇ ´ëÇ¥·Î¼, Physical AI ±â¹Ý ÀÚÀ² ·Îº¿ Ç÷§ÆûÀ» °³¹ßÇϰí ÀÖ½À´Ï´Ù. ƯÈ÷ ÀÚ¼®¹ß Á·Çü ·Îº¿ Á¦¾î ±â¼ú°ú °ÈÇнÀ ±â¹Ý º¸Çà ¾Ë°í¸®ÁòÀ» ÅëÇØ Á¶¼±¼Ò, °Ç¼³ ÇöÀå µî ±ØÇÑ »ê¾÷ ȯ°æ¿¡¼ ±âÁ¸ ·Îº¿ÀÌ Á¢±ÙÇϱ⠾î·Á¿ü´ø 3Â÷¿ø ¹ÐÆó °ø°£ ÀÛ¾÷ ÀÚµ¿È¸¦ ½ÇÇöÇϰí ÀÖÀ¸¸ç, ±¹³» ÁÖ¿ä Á¶¼±»çµé°ú Çù·ÂÇÏ¿© »ê¾÷ ÇöÀåÀÇ ½ÇÁúÀûÀÎ ¹®Á¦¸¦ ÇØ°áÇϰí ÀÖ½À´Ï´Ù.
-
Joon-Ha Kim received his Ph.D. in Mechanical Engineering from KAIST, specializing in robot state estimation and dynamic motion control in complex environments. As a researcher at KAIST's Humanoid Robot Research Center, he achieved world-class research outcomes in quadruped robotics and published numerous papers in prestigious journals including IEEE T-RO, IEEE RA-L, and IJRR. Currently serving as CEO of DIDEN Robotics, which he co-founded with KAIST colleagues in March 2024, he leads the development of Physical AI-based autonomous robot platforms. Through magnetic-foot legged robot control technology and reinforcement learning-based locomotion algorithms, he is enabling automation in 3D confined spaces previously inaccessible to conventional robots in extreme industrial environments such as shipyards and construction sites. Working closely with major Korean shipbuilders, he is addressing practical challenges in industrial automation.
¹ßÇ¥¿ä¾à
Physical AIÀÇ »ê¾÷ÇöÀåÀ¸·ÎÀÇ Àû¿ë: ÀÚ¼®¹ßÀ» Ȱ¿ëÇÑ ÇöÀå ÁøÀÔ °¡¼ÓÈ
º» °¿¬¿¡¼´Â ±ØÇÑ »ê¾÷ȯ°æ¿¡ Physical AI¸¦ Àû¿ëÇϱâ À§ÇÑ µðµç·Îº¸Æ½½ºÀÇ Á¢±Ù¹ýÀ» ¼Ò°³ÇÕ´Ï´Ù. Á¶¼±¼Ò, °Ç¼³ ÇöÀå µî À§ÇèÇÏ°í º¹ÀâÇÑ 3Â÷¿ø ¹ÐÆó°ø°£Àº ±âÁ¸ ÀÚµ¿È ±â¼ú·Î´Â Á¢±ÙÀÌ ¾î·Á¿î ¿µ¿ªÀ¸·Î, µðµç·Îº¸Æ½½º´Â ÀÌ·¯ÇÑ È¯°æ¿¡¼ ÀÛµ¿ÇÏ´Â ÀÚÀ² ·Îº¿ ½Ã½ºÅÛÀ» Çϵå¿þ¾î, Áö´ÉÇü Á¦¾î, ½Ã°¢ Áö´ÉÀÇ ÅëÇÕÀû °üÁ¡¿¡¼ °³¹ßÇϰí ÀÖ½À´Ï´Ù.³×À̹öŬ¶ó¿ìµå ¿Á»óÈÆ »ç¾÷ ¸®´õ´Â 1¼¼´ë ÀÚ¹Ù°³¹ßÀÚ Ãâ½ÅÀ¸·Î 2006³â Çѱ¹ÀÚ¹Ù°³¹ßÀÚÇùȸ ȸÀåÀ» ¿ªÀÓÇß°í, ÀÌÈÄ 2011³â ³×À̹ö¿¡ ÀÔ»çÇÏ¿© OpenAPI¸¦ Á¦°øÇÏ´Â ³×À̹ö°³¹ßÀÚ¼¾ÅÍ PMÀ» ¸Ã¾Ò´Ù. ÀÌÈÄ Ç÷§Æû »ç¾÷Á¦ÈÞ ¹× D2½ºÅ¸Æ®¾÷ ÅõÀÚ ¾÷¹«¸¦ ÁøÇàÇß°í, Ŭ·Î¹Ù ÀΰøÁö´É »ç¾÷ÆÀ¿¡¼ Ŭ·Î¹Ù ¿¡¹ÝÁ©¸®½ºÆ®·Î¼ AI»ýŰè È®´ë ¹× AI »ç¾÷°³¹ßÀ» ÇØ¿Ô´Ù. 2020³â Äڷγª ÆÒµ¥¹Í »óȲ¿¡ ±¹³» ÃÖÃÊ·Î ¼º³²½Ã¿Í Çù·ÂÇÏ¿© Äڷγª »óŸ¦ ¸ð´ÏÅ͸µÇÏ´Â AI ÀüÈ 'Ŭ·Î¹ÙÄɾîÄÝ'À» ·ÐĪÇß´Ù. À̸¦ °è±â·Î 2011³â 11¿ù ÃʰŴë¸ðAI ±â¹ÝÀÇ AI¾ÈºÎÀüÈ ¼ºñ½º Ŭ·Î¹ÙÄɾîÄÝ »ç¾÷À» ¸Ã¾Æ 2025³â 6¿ù ±âÁØ Àü±¹ 130¿© ÁöÀÚü 3¸¸¸í ´ë»óÀ¸·Î ¼ºñ½º¸¦ È®´ëÇØ ³ª°¡°í ÀÖ°í, ÀϺ» ÀÌÁî¸ð½Ã¿¡ ½Ã¹ü¼ºñ½º ·ÐĪ¿¡ ±â¿©Çß´Ù.
¹ßÇ¥¿ä¾à
½Ã´Ï¾îÄɾî AI ¹× Äɾî·Îº¿ µ¿Çâ
±âȹ¼¼¼Ç 2-1
Mobility Innovation through Physical AI
¸ðºô¸®Æ¼ Çõ½Å°ú ÇÇÁöÄà AI
À¯¼öÀÇ ±Û·Î¹ú ±â¾÷µé¿¡¼ 25³â°£ÀÇ ´Ù¾çÇÑ IT ºñÁö´Ï½º ¿µ¿ª¿¡¼ÀÇ ¾÷¹«¿Í ½Ç¹« °æÇèµéÀ» º¸À¯.
ÇöÀç º¸½ºÅæ ´ÙÀ̳»¹Í½º¿¡¼ Çѱ¹À» Æ÷ÇÔÇÑ ¾Æ½Ã¾Æ Áö¿ª ·Îº¿ ºñÁö´Ï½º¸¦ ´ã´çÇϰí ÀÖ´Ù.
-
Jason Yoon has over 25 years of experience in various IT business areas at leading global companies. At Boston Dynamics., he is currently responsible for business growth and expansion in the Asia Pacific region.
¹ßÇ¥¿ä¾à
·Îº¿ÀÌ ¹«¾ùÀ» ÇÒ ¼ö ÀÖÀ»±î¿¡ ´ëÇÑ »ý°¢À» ¹Ù²ÙÀÚ
º¸½ºÅæ´ÙÀ̳»¹Í½ºÀÇ ¼¼°è ÃÖ°í ¼º´ÉÀÇ 4Á·º¸Çà·Îº¿, ½ºÆÌ (Spot), ÀÌ ¿©·¯ »ê¾÷ ºÐ¾ß¿¡¼ ¾î¶»°Ô »ç¿ëµÇ°í ÀÖ°í, ½ºÆÌ (Spot) ÀÇ ¿î¿µ°ú µ¥ÀÌÅ͸¦ °ü¸®ÇÏ´Â ¼ÒÇÁÆ®¿þ¾î Åø, ¿À¸£ºø (Orbit) ÀÇ AI ±â´ÉµéÀÌ Áö¼ÓÀûÀ¸·Î Ãß°¡µÇ°í ¹ßÀüµÇ¾î°¡¸é¼ ÀÌ·ç¾î°¡´Â ½ºÆÌ (Spot) ·Îº¿ÀÇ ÀÚµ¿ ¿¹Áö °ü¸® À̾߱â.
¹ßÇ¥¿ä¾à
AI in the physical world - Çö½Ç ¼¼°è¿Í À¶ÇÕÇÏ´Â AIÀÇ ¹Ì·¡
1. New Wave of Physical AI : ¿ì¸®´Â ¿Ö ¡®¹°¸®Àû¡¯ AI¸¦ ³íÇØ¾ß Çϴ°¡?
¹ßÇ¥¿ä¾à
(ÁÖ)CTRÀÇ ÇÇÁöÄÃ AI Àü·«
±âȹ¼¼¼Ç 2-2
Autonomous Manufacturing and Logistics Powered by Physical AI
ÇÇÁöÄà AI·Î ¿©´Â ÀÚÀ²Á¦Á¶ ¹°·ù
-
¹ßÇ¥¿ä¾à
--
¹ßÇ¥¿ä¾à
Physical AI – Future of Global Manufacturing Innovation
--
¹ßÇ¥¿ä¾à
Á¦Á¶¹°·ùÀÇ Çõ½Å, from chaos to harmony
--
¹ßÇ¥¿ä¾à
AI¿Í µðÁöÅÐÆ®À©À¸·Î ÁøÈÇÏ´Â ½º¸¶Æ®ÆÑÅ丮
-±âȹ¼¼¼Ç 2-3
Physical AI Food Tech
ÇÇÁöÄà AI ǪµåÅ×Å©
-
¹ßÇ¥¿ä¾à
-KENSUKE HARADA (Fellow, IEEE) received the Ph.D. degree from Kyoto University, Kyoto, Japan, in 1997. From 1997 to 2002, he was a Research Associate with Hiroshima University, Hiroshima, Japan. From 2005 to 2006, he was a Visiting Scholar with the Computer Science Department, Stanford University, Stanford, CA, USA. He is currently a Professor with the Graduate School of Engineering Science, Osaka University. Before joining Osaka University, he was a Researcher with the National Institute of AIST, Tsukuba, Japan. His research interests include the mechanics and motion planning of humanoid robots and robotic hands.
¹ßÇ¥¿ä¾à
Achieving Robust Motion Planning for Robotic Manipulation Tasks
In this talk, we present the robotic motion planning from abstracted instructions such as a cooking recipe or YouTube video. Our framework is a combination of the task graph-sequence graph and motion planner, where the high-level planning is done based on the task-sequence graph while the low-level planning is done based on the motion planner. We show that this framework is robust against hallucinations of LLM.
³²±âÇöÀº ºù±×·¹, ³²¾çÀ¯¾÷ µî ½Äǰ±â¾÷¿¡¼ 20³â ÀÌ»óÀÇ °æÇèÀ» ½×¾Æ¿Ô´Ù. ÇöÀç´Â Çϸ²ÁöÁÖÀÇ Àü·«±âȹÆÀ¿¡¼ K-Ǫµå ¼öÃâ ¹× ´Ù¾çÇÑ »ý»ê¼³ºñ¸¦ Ȱ¿ëÇÑ OEM/ODM, °¡°ø½Äǰ»ê¾÷¿¡¼ Çϸ²ÀÇ ¿ªÇÒ È®´ë¸¦ ´ã´çÇϰí ÀÖ´Ù.
-
Kihyun Nam has more than 20 years of experience in food companies such as Binggrae and Namyang Dairy. He is in charge of the OEM/ODM business using K-food exports and various production facilities in Harim Holdings' strategic planning team.
¹ßÇ¥¿ä¾à
½Ä»çÀÇ º¯È, ±â¼úÀÇ º¯È
½Ä»ç´Â Àΰ£ÀÌ »ì¾Æ°¡´Âµ¥ ÀÖ¾î °¡Àå ±âº»ÀÌ´Ù. »ýȰȯ°æÀÇ º¯È¿Í ÇÔ²² ¸ÀÀÖ´Â ½Ä»ç¸¦ Áñ±â±â À§ÇÑ ½Ä»çȯ°æµµ º¯ÈµÇ¾î ¿Ô°í, ±â¼úÀº ±×°ÍÀ» ´õ Æí¸®ÇÏ°í °£´ÜÇÏ°Ô ¸¸µé¾î ¿Ô´Ù. ÀÌ¿¡ ½Äǰȸ»ç´Â ¾î¶»°Ô º¯È¿¡ ¹ß¸ÂÃß°í ÀÖÀ»±î?
Á¤¼ºÈ¯Àº ÀüºÏ´ëÇб³ ÄÄÇ»ÅͰøÇаú¿¡¼ ¹Ú»ç ÇÐÀ§¸¦ ¹ÞÀº ÈÄ Çѱ¹ÀüÀÚ±â¼ú¿¬±¸¿ø ITÀÀ¿ë¿¬±¸¼¾ÅÍ¿¡¼ 13³â°£ ÀçÁ÷ÁßÀÌ´Ù. ¿µ»óó¸® ¹× ÀΰøÁö´É ±â¼úÀ» ±â¹ÝÀ¸·Î Á¤¹Ð³ó¾÷ ¹× ǪµåÅ×Å© AIÀÚÀ²Á¦Á¶ µîÀÇ »ç¾÷À» ¼öÇàÇϰí ÀÖ´Ù. Foodtech at the Center of Physical AI Expansion
-
After receiving his Ph.D. in Computer Engineering from Jeonbuk National University, Sung-Hwan Jeong has been working at the IT Convergence Research Center of the Korea Electronics Technology Institute (KETI) for 13 years. He is engaged in projects related to precision agriculture and food-tech AI autonomous manufacturing, based on image processing and artificial intelligence technologies.
¹ßÇ¥¿ä¾à
ÇÇÁöÄà AI È®»êÀÇ Áß½É, ǪµåÅ×Å©
½ÄÀ½·á Á¦Á¶ °øÁ¤ÀÇ ÇöȲ°ú ·Îº¿ ¹× ÀÚµ¿È ±â¼úÀÇ Àû¿ë ÇѰ輺À» ºÐ¼®Çϰí,
±âȹ¼¼¼Ç 3-1
Computer Vision for Physical AI
ÇÇÁöÄà AI¸¦ À§ÇÑ ÄÄÇ»ÅÍ ºñÀü
Àú´Â ¿ï»ê°úÇбâ¼ú¿ø(UNIST) ÀΰøÁö´É´ëÇпø ºÎ±³¼ö·Î, 2020³â 4¿ù ºÎÀÓÇÏ¿´½À´Ï´Ù. ÇкÎ(2009)¿Í ¼®»ç(2011)´Â KAIST Àü±â¹×ÀüÀÚ°øÇаú¿¡¼ ÃëµæÇÏ¿´À¸¸ç, ¹Ú»çÇÐÀ§(2020)´Â ¿µ±¹ Àӯ䏮¾ó Ä®¸®Áö ·±´ø Àü±âÀüÀÚ°øÇаú¿¡¼ ¹Þ¾Ò½À´Ï´Ù. ¹Ú»ç °úÁ¤ Áß 2019³â 6¿ùºÎÅÍ 8¿ù±îÁö INRIA ¼ÒÇÇ¾Æ ¾ÓƼÆú¸®½º ¿¬±¸¼Ò¿¡¼ ÀÎÅÏ ¿¬±¸¿øÀ¸·Î ±Ù¹«ÇÑ ¹Ù ÀÖÀ¸¸ç, ¹Ú»ç °úÁ¤ ÀÔÇÐ Àü¿¡´Â 2011³â 2¿ùºÎÅÍ 2015³â 2¿ù±îÁö »ï¼ºÀüÀÚ DMC ¿¬±¸¼¾ÅÍ¿¡¼ 4³â°£ ÀçÁ÷ÇÏ¿´½À´Ï´Ù.
-
I am an associate professor, jointly affiliated to the Artificial Intelligence Graduate at UNIST where I joined in April 2020. I obtained BS (2009) and MS (2011) degrees from Dept. of Electrical Engineering at KAIST and Ph.D. degree (2020) from Dept. of Electrical and Electronic Engineering at Imperial College London. During my Ph.D., I have interned at INRIA Sophia Antipolis (2019.6-2019.8.). Before joining Ph.D., I was an employee at DMC Research Center of Samsung Electronics for four years (2011.2.-2015.2.).
¹ßÇ¥¿ä¾à
-
Çѵ¿À± ¹Ú»ç´Â NAVER AI LabÀÇ ¿¬±¸¿øÀ̸ç KAIST ÀΰøÁö´É´ëÇпø(GSAI)ÀÇ °âÀÓ±³¼öÀÌ´Ù. ±×´Â KAIST¿¡¼ 2018³â¿¡ ¹Ú»çÇÐÀ§, 2011³â¿¡ ÇлçÇÐÀ§¸¦ ¹Þ¾Ò´Ù. ±×ÀÇ ¿¬±¸´Â ±â°è ÇнÀÀÇ °üÁ¡¿¡¼ ´ë±Ô¸ð ¾ð¾î¸ðµ¨ (large language model)°ú ¸ÖƼ ¸ð´Þ ¸ðµ¨ (multi-model model)ÀÇ ¹ßÀü¿¡ ÃÊÁ¡À» µÎ°í ÀÖ´Ù. ±×´ÂICLR 2026°ú NeurIPS 2025 (2023³â ºÎÅÍ)ÀÇ ºÐ¾ßº°Area ChairÀ¸·Î Ȱµ¿Çϰí ÀÖ´Ù.
-
Dongyoon Han is a Research Scientist at NAVER AI Lab and an Adjunct Professor at KAIST GSAI. He received his Ph.D. in 2018 and his B.S. in 2011, both from KAIST. His research focuses on advancing large language models and multimodal models through the lens of machine learning. He has served as an Area Chair for ICLR 2026 and NeurIPS 2025 (since 2023).
¹ßÇ¥¿ä¾à
Towards Perceptive Visual Representations for Real-World Multimodal Learning
±¹¹®: ÀÌ ¹ßÇ¥¿¡¼´Â real-world multimodal learningÀ» À§ÇÑ µÎ °¡Áö perceptive visual representation Á¢±Ù¹ýÀ» °£·«È÷ ¼Ò°³ÇÑ´Ù. ù ¹øÂ° ¿¬±¸¿¡¼´Â robot learning, video label propagation, pose tracking µîÀÇ task¿¡ Àû¿ë °¡´ÉÇÑ »õ·Î¿î self-supervised ±â¹ÝÀÇ sequential scene understanding ¹æ¹ýÀ» ¼Ò°³ÇÑ´Ù. ÁÖ¾îÁø ¿¬¼Ó Àå¸éÀ» bottleneck tokenÀ¸·Î ¾ÐÃàÇϰí ÃÖ¼ÒÇÑÀÇ visual cues¸¸À¸·Î ÀÌÈÄ ÇÁ·¹ÀÓÀ» È¿°úÀûÀ¸·Î ¿¹ÃøÇϵµ·Ï visual representationÀ» ÇнÀÇÏ¸é °ß°íÇϰí È¿À²ÀûÀÎ Àå¸é ¿¹ÃøÀ» °¡´ÉÇÏ°Ô ÇÔÀ» º¸ÀδÙ. µÎ ¹øÂ° ¿¬±¸¿¡¼´Â multimodal language models (MLLMs)ÀÇ reinforcement learning (RL) ¹× supervised fine-tuning (SFT) ÇнÀ Àü·«À» ºÐ¼®Çϸç, À̵éÀÌ visual representations¿¡ ¾î¶² ¿µÇâÀ» ¹ÌÄ¡´ÂÁö¸¦ ½Éµµ ÀÖ°Ô Å½±¸ÇÑ´Ù. RLÀº SFTº¸´Ù ´õ °·ÂÇÏ°í °ø°£ÀûÀ¸·Î Á¤¹ÐÇÑ visual representationsÀ» ¹è¿òÀ» º¸ÀÌ¸é¼ À̰ÍÀÌ vision encoderÀÇ ¼º´ÉÀ» Å©°Ô Çâ»ó½ÃÅ´À» ¶ÇÇÑ º¸¿©ÁØ´Ù.
¹ÚÀºº´ ±³¼ö´Â ¿¬¼¼´ëÇб³ ÀΰøÁö´ÉÇаúÀÇ Á¶±³¼öÀÔ´Ï´Ù. ±×´Â 2009³â¿¡ °æÈñ´ëÇб³¿¡¼ ÄÄÇ»ÅͰøÇÐ Çлç ÇÐÀ§¸¦, 2011³â¿¡ ¼¿ï´ëÇб³¿¡¼ ÄÄÇ»ÅͰøÇÐ ¼®»ç ÇÐÀ§¸¦, 2019³â¿¡ ¹Ì±¹ ³ë½ºÄ³·Ñ¶óÀ̳ª ´ëÇб³ äÇÃÈú(University of North Carolina at Chapel Hill)¿¡¼ ÄÄÇ»ÅͰøÇÐ ¹Ú»ç ÇÐÀ§¸¦ ÃëµæÇß½À´Ï´Ù. ¿¬¼¼´ëÇб³¿¡ ºÎÀÓÇϱâ Àü¿¡´Â ¼º±Õ°ü´ëÇб³¿¡¼ Á¶±³¼ö·Î ÀçÁ÷ÇßÀ¸¸ç, Nuro¿¡¼ ¿¬±¸¿ø(Research Scientist)À¸·Î, Microsoft¿¡¼ ÀÀ¿ë°úÇÐÀÚ(Applied Scientist)·Î ±Ù¹«ÇÑ ¹Ù ÀÖ½À´Ï´Ù. ÇöÀç ±×ÀÇ ÁÖ¿ä ¿¬±¸ ºÐ¾ß´Â 3Â÷¿ø ºñÀü(3D Vision) ¹× »ý¼º ¸ðµ¨¸µ(Generative Modeling) ÀÔ´Ï´Ù.
-
Eunbyung Park is an assistant professor in the Department of Artificial Intelligence at Yonsei University, South Korea. Eunbyung Park obtained his B.S. degree in computer science from KyungHee University in 2009, his M.S. degree in computer science from Seoul National University in 2011, and his Ph.D. degree in computer science from the University of North Carolina at Chapel Hill in 2019. Before joining Yonsei University, he was an assistant professor at SKKU, a research
scientist at Nuro and an applied scientist at Microsoft. His current research interests include 3D vision and generative modeling.
¹ßÇ¥¿ä¾à
¼ø¹æÇ⠽Űæ¸Á ±â¹Ý 3Â÷¿ø °¡¿ì½Ã¾È ½ºÇ÷¡ÆÃ
ÃÖ±Ù 3D Gaussian Splatting (3DGS) ±â¼úÀÇ ¹ßÀüÀº ½Ç½Ã°£ »õ·Î¿î ½ÃÁ¡ ÇÕ¼º(real-time novel view synthesis) ºÐ¾ß¿¡ Çõ½ÅÀ» °¡Á®¿Ô½À´Ï´Ù. ±×·¯³ª ±âÁ¸ 3DGS ÆÄÀÌÇÁ¶óÀÎÀº °¢ Àå¸é¸¶´Ù ´Ù¼öÀÇ ±×·¡µð¾ðÆ® ¾÷µ¥ÀÌÆ®¸¦ ¿ä±¸ÇÏ´Â ´À¸®°í ¹Ýº¹ÀûÀÎ ÃÖÀûÈ °úÁ¤À̶ó´Â ÇѰ谡 ÀÖ½À´Ï´Ù. º» °¿¬¿¡¼´Â ÀÌ·¯ÇÑ º´¸ñÀ» ÇØ¼ÒÇϱâ À§ÇÑ Feed-forward ±â¹Ý 3D Gaussian Splatting ÇÁ·¹ÀÓ¿öÅ©¸¦ ¼Ò°³ÇÕ´Ï´Ù. Á¦¾ÈÇÏ´Â Á¢±Ù¹ýÀº ´ÙÁß ½ÃÁ¡ ¿µ»óÀ» 3D Gaussian Ç¥ÇöÀ¸·Î Á÷Á¢ ¸ÅÇÎÇÔÀ¸·Î½á, Àå¸éº° ÃÖÀûÈ °úÁ¤ ¾øÀÌ °íǰÁú 3D Àå¸éÀ» Áï½Ã À籸¼ºÇÒ ¼ö ÀÖ°Ô ÇÕ´Ï´Ù. ¶ÇÇÑ, º» ¿¬±¸½Ç¿¡¼ ÃÖ±Ù ¼öÇàÇÑ Feed-forward 3DGS °ü·Ã ÁÖ¿ä ¿¬±¸ ¼º°úµéÀ» ¼Ò°³Çϰí, ÇâÈÄ È®Àå °¡´ÉÇÏ°í °ß°íÇÏ¸ç »ý¼ºÀûÀÎ 3D Gaussian Ç¥ÇöÀ¸·Î ³ª¾Æ°¡±â À§ÇÑ ¿¬±¸ ¹æÇâÀ» ³íÀÇÇÒ ¿¹Á¤ÀÔ´Ï´Ù.
ÁÖÀç°É ±³¼ö´Â KAIST ±èÀçöAI´ëÇпø ±³¼ö·Î ÀçÁ÷ ÁßÀ̸ç, ÄÄÇ»ÅͺñÀü, ÀÚ¿¬¾îó¸®, ½Ã°è¿ ¿¹Ãø ºÐ¾ßÀÇ ÀΰøÁö´É ±â¼úÀ» ¿¬±¸Çϰí ÀÖ´Ù. ±×´Â 2013³â ¹Ì±¹ ÁÒÁö¾Æ°ø´ë¿¡¼ ¹Ú»ç ÇÐÀ§¸¦ ¹Þ¾Ò°í, 2020³â¿¡¼ 2025³â »çÀÌ¿¡´Â °í·Á´ëÇб³¿¡¼ ±³¼ö·Î ÀçÁ÷Çß´Ù.
-
Prof. Jaegul Choo is currently a Professor at the KimJaechul Graduate School of AI in KAIST, and his research interest lies in various deep learning applications such as computer vision, natural language processing, and time-series analysis. He received a PhD from Georgia Tech in 2013, and he worked as a Professor at Korea University from 2020 to 2025.
¹ßÇ¥¿ä¾à
Vision-Language-Action ¸ðµ¨ÀÇ Action Noise °³¼± ¿¬±¸ ¼Ò°³
ÃÖ±Ù foundational modelÀÇ ¹ßÀüÀ¸·Î physical AI ºÐ¾ß¿¡¼ vision-language-action (VLA) ¸ðµ¨ÀÌ °¢±¤À» ¹Þ°í ÀÖ´Ù. º» °¿¬¿¡¼´Â ÀÌ·¯ÇÑ VLA ¸ðµ¨À» ¸ð¹æÇнÀ ±â¹ÝÀ¸·Î ÇнÀÇÒ ¶§ ¹ß»ýµÉ ¼ö ÀÖ´Â action sequence »óÀÇ jittering noise¸¦ ¾ïÁ¦ÇÒ ¼ö ÀÖ´Â ¹æ¹ý·ÐÀ» ¼Ò°³ÇÑ´Ù. ±¸Ã¼ÀûÀ¸·Î Á¦¾È ¹æ¹ý·ÐÀº diffusion model ±â¹ÝÀÇ À̹ÌÁö »ý¼º ¸ðµ¨¿¡¼ Ȱ¿ëµÇ´Â Ãß·Ð ´Ü°èÀÇ guidance ±â¹ýÀ» È®ÀåÇÑ °ÍÀ¸·Î, º» °¿¬¿¡¼´Â À̸¦ ÅëÇØ ·Îº¿¿¡°Ô ÁÖ¾îÁø ŽºÅ©¸¦ º¸´Ù ºü¸£°í Á¤È®ÇÏ°Ô ¼öÇàÇÒ ¼ö ÀÖ´Ù´Â °á°ú¸¦ Á¦½ÃÇÑ´Ù.±âȹ¼¼¼Ç 3-2
AI Research Insight
AI ¿¬±¸ ÀλçÀÌÆ®
¿À¹Îȯ ±³¼ö´Â ¼¿ï´ëÇб³ µ¥ÀÌÅÍ»çÀ̾𽺴ëÇпø ºÎ±³¼ö·Î ÀçÁ÷ ÁßÀÌ´Ù. ÁÖ ¿¬±¸ ºÐ¾ß´Â ÀÇ»ç°áÁ¤ ¿¡ÀÌÀüÆ®, °ÈÇнÀ, ÄÁÅØÃò¾ó ¹êµ÷ ¾Ë°í¸®Áò, ÃÖÀûÈ, Åë°èÀû ¸Ó½Å·¯´×ÀÌ´Ù. Ä÷³ºñ¾Æ´ëÇб³¿¡¼ Operations Research ¹Ú»çÇÐÀ§¸¦ ¹Þ¾ÒÀ¸¸ç, ±×ÀÇ ¹Ú»çÇÐÀ§ ³í¹®Àº INFORMS George B. Dantzig Dissertation Award¿Í Applied Probability Society Best Student Paper AwardÀÇ ÃÖÁ¾ È帷Π¼±Á¤µÇ¾ú´Ù. ¿À¹Îȯ ±³¼ö´Â 2025³â ¾Æ¸¶Á¸ ¿¬±¸»ó(Amazon Research Award)À» ¼ö»óÇÏ¿´´Ù.
-
Min-hwan Oh is an Associate Professor in the Graduate School of Data Science at Seoul National University. His research focuses on decision-making agents, reinforcement learning, contextual bandit algorithms, optimization, statistical machine learning, and their various applications. He received his Ph.D. in Operations Research from Columbia University. His doctoral thesis was recognized as a finalist for the INFORMS George B. Dantzig Dissertation Award and the Applied Probability Society¡¯s Best Student Paper Award. He is a recipient of the Amazon Research Award.
¹ßÇ¥¿ä¾à
Á¶ÇÕÀû ÇൿÀ» °í·ÁÇÑ ÄÁ¼½Ãò¾ó ¹êµ÷(Contextual Bandit)°ú RLHF¿¡¼ÀÇ È¿À²Àû Ž»ö
¹®¸Æ Á¤º¸¸¦ ¹ÙÅÁÀ¸·Î ¿¡ÀÌÀüÆ®(¾Ë°í¸®Áò)°¡ ¼øÂ÷ÀûÀ¸·Î ´Ü¼ö Çѵ¿ÀÌ ¾Æ´Ñ Çൿ ÁýÇÕÀ» ¼±ÅÃÇϰí, »ç¿ëÀÚ ¼±ÅÃÀÌ MNL(multinomial logit) ¸ðÇüÀ» µû¸¥´Ù°í °¡Á¤ÇÏ´Â ¼øÂ÷Àû ÀÇ»ç°áÁ¤ ¹®Á¦¸¦ ´Ù·ç´Â ¿¬±¸¸¦ ¼Ò°³ÇÑ´Ù. ÀÌ ¹®Á¦´Â ¿Â¶óÀÎ Ãßõ ½Ã½ºÅÛÀÇ ÇÙ½É °úÁ¦À̸ç, ´ë±Ô¸ð ¾ð¾î¸ðµ¨(LLM)ÀÇ »çÈÄ ÇнÀ(post-training) ´Ü°è¿¡¼µµ Á߿伺ÀÌ Ä¿Áö°í ÀÖ´Ù. ÀÌ·¯ÇÑ º¸ÆíÀû ¼³Á¤¿¡µµ ºÒ±¸Çϰí À̷аú ½Ç¹« »çÀÌ¿¡´Â °£±ØÀÌ Á¸ÀçÇÑ´Ù. º» ¹ßÇ¥¿¡¼´Â ´ÙÀ½°ú °°Àº ÃÖ±Ù ¼º°ú¸¦ Á¦½ÃÇÑ´Ù: ÀÌ·ÐÀû º¸Àå(provable optimality) ´Þ¼º, ºñ¼±Çü È¿¿ë(nonlinear utilities)¿¡ ´ëÇØ ÃÖÃÊ·Î °è»ê °¡´É(tractable) ÇÑ ¾Ë°í¸®Áò Á¦¾È ¹× ÀÌ·ÐÀû ¼º´É º¸Àå, ÇÁ·¹ÀÓ¿öÅ©¸¦ RLHF(¶Ç´Â preference-based RL)·Î È®Àå.
±è½Â·æÀº ¼¿ïƯº°½Ã KAIST ±èÀçöAI´ëÇпø ºÎ±³¼öÀÌ´Ù. KAIST ºÎÀÓ ÀÌÀü¿¡´Â 2020³âºÎÅÍ 2024³â±îÁö °í·Á´ëÇб³ ÄÄÇ»ÅÍÇаú¿¡¼ Á¶±³¼ö·Î ÀçÁ÷ÇÏ¿´´Ù. ±× ÀÌÀü¿¡´Â 2019³âºÎÅÍ 2020³â±îÁö ½ºÀ§½º ·ÎÀÜ¿¬¹æ°ø°ú´ëÇÐ(École Polytechnique Fédérale de Lausanne, EPFL) ÄÄÇ»ÅÍÅë½Å°úÇкο¡¼ ¹Ú»çÈÄ¿¬±¸¿øÀ¸·Î, 2018³âºÎÅÍ 2019³â±îÁö ¿¬¼¼´ëÇб³ Àü±âÀüÀÚ°øÇаú¿¡¼ ¹Ú»çÈÄ¿¬±¸¿øÀ¸·Î ±Ù¹«ÇÏ¿´´Ù. ¶ÇÇÑ Áß±¹ º£ÀÌ¡ Microsoft Research Asia(MSRA)¿¡¼ ¿¬±¸ ÀÎÅÏÀ¸·Î ±Ù¹«ÇÑ °æÇèÀÌ ÀÖ´Ù. ¿¬¼¼´ëÇб³ Àü±âÀüÀÚ°øÇаú¿¡¼ 2012³â¿¡ Çлç, 2018³â¿¡ ¹Ú»ç ÇÐÀ§¸¦ ÃëµæÇÏ¿´´Ù. ÇöÀç CVPR, ICCV, ECCV, ICLR, NeurIPS µî ÁÖ¿ä ÀΰøÁö´É ¹× ÄÄÇ»ÅͺñÀü ±¹Á¦ÇÐȸÀÇ Area Chair·Î Ȱµ¿ ÁßÀÌ´Ù. ÁÖ¿ä ¿¬±¸ ºÐ¾ß´Â ÄÄÇ»ÅÍ ºñÀü, °è»ê »çÁøÇÐ, ±×¸®°í ¸Ó½Å·¯´×ÀÌ´Ù.
-
Seungryong Kim is an Associate Professor at the Kim Jaechul Graduate School of AI, KAIST, Seoul, Korea. Prior to joining KAIST, he was an Assistant Professor at Korea University from 2020 to 2024. He also worked as a Postdoctoral Researcher at the School of Computer and Communication Sciences, École Polytechnique Fédérale de Lausanne (EPFL), Switzerland, from 2019 to 2020, and at the School of Electrical and Electronic Engineering, Yonsei University, Korea, from 2018 to 2019. He was previously a Research Intern at Microsoft Research Asia (MSRA), Beijing, China. He received his B.S. and Ph.D. degrees in Electrical and Electronic Engineering from Yonsei University, Korea, in 2012 and 2018, respectively. He is actively serving as an Area Chair for leading conferences, including CVPR, ICCV, ECCV, ICLR, and NeurIPS. His research interests include computer vision, computational photography, and machine learning.
¹ßÇ¥¿ä¾à
ºñµð¿À »ý¼ºÇü ¸ðµ¨·ÎºÎÅÍÀÇ Æ÷ÀÎÆ® Æ®·¹Å· Á¤º¸ ÃßÃâ
ÃÖ±Ù Diffusion Transformer(DiT) ±â¹ÝÀÇ ºñµð¿À È®»ê ¸ðµ¨(Video Diffusion Model)ÀÌ ½Ã°£ÀûÀ¸·Î ÀϰüµÈ ¿µ»óÀ» »ý¼ºÇÏ´Â µ¥ ³î¶ó¿î ¼º°ú¸¦ °ÅµÎ°í ÀÖ½À´Ï´Ù. ÇÏÁö¸¸ ¿©ÀüÈ÷ ³²¾ÆÀÖ´Â ±Ùº»ÀûÀÎ Áú¹®Àº ÀÌ·¯ÇÑ ¸ðµ¨Àº ³»ºÎÀûÀ¸·Î ÇÁ·¹ÀÓ °£ÀÇ ½Ã°£Àû ´ëÀÀ(temporal correspondence) À» ¾î¶»°Ô Çü¼ºÇϰí Ç¥ÇöÇÒ±î ÀÔ´Ï´Ù. º» ³í¹®¿¡¼´Â ÀÌ Áú¹®¿¡ ´äÇϱâ À§ÇØ DiffTrackÀ» Á¦¾ÈÇÕ´Ï´Ù. ÀÌ´Â ÃÖÃÊÀÇ Á¤·®Àû ºÐ¼® ÇÁ·¹ÀÓ¿öÅ©·Î, ºñµð¿À DiTÀÇ ³»ºÎ ½Ã°£Àû ¸ÞÄ¿´ÏÁòÀ» ޱ¸Çϵµ·Ï ¼³°èµÇ¾ú½À´Ï´Ù. DiffTrackÀº ÇÁ·ÒÇÁÆ®·Î »ý¼ºµÈ ¿µ»ó°ú ÀÇ»ç(pseudo) Á¤´ä ÃßÀû(annotation) µ¥ÀÌÅ͸¦ Æ÷ÇÔÇÑ »õ·Î¿î µ¥ÀÌÅͼÂÀ» ±¸¼ºÇϰí, 3D ¾îÅÙ¼Ç ¸ÞÄ¿´ÏÁòÀÇ ±¸¼º ¿ä¼Òµé(¿¹: Ç¥Çö, ·¹À̾î, ŸÀÓ½ºÅÜ)ÀÌ ½Ã°£Àû ´ëÀÀÀ» Çü¼ºÇÏ´Â µ¥ ¾î¶»°Ô ±â¿©ÇÏ´ÂÁö¸¦ ü°èÀûÀ¸·Î ºÐ¼®Çϱâ À§ÇÑ »õ·Î¿î Æò°¡ ÁöÇ¥¸¦ Á¦¾ÈÇÕ´Ï´Ù. ºÐ¼® °á°ú, ƯÁ¤ ·¹À̾îÀÇ query-key À¯»çµµ(query-key similarity) °¡ ½Ã°£Àû ¸ÅĪ¿¡¼ ÇÙ½ÉÀûÀÎ ¿ªÇÒÀ» Çϸç, ÀÌ ¸ÅĪÀº µð³ëÀÌ¡(denoising) °úÁ¤ÀÌ ÁøÇàµÉ¼ö·Ï Á¡Á¡ ´õ µÎµå·¯Áö°Ô ³ªÅ¸³´Ù´Â »ç½ÇÀ» ¹ß°ßÇß½À´Ï´Ù. ¿ì¸®´Â DiffTrackÀÇ ½Ç¿ëÀû ÀÀ¿ëÀ¸·Î zero-shot Æ÷ÀÎÆ® Æ®·¡Å·(point tracking) À» ½Ã¿¬ÇÏ¿´À¸¸ç, ±âÁ¸ÀÇ ºñÀü ±â¹Ý ÆÄ¿îµ¥ÀÌ¼Ç ¹× ÀÚ±âÁöµµ ºñµð¿À ¸ðµ¨µéº¸´Ù ÃÖ÷´Ü ¼º´É(state-of-the-art) À» ´Þ¼ºÇß½À´Ï´Ù. ¶ÇÇÑ, ¿ì¸®ÀÇ ºÐ¼®À» motion-enhanced ºñµð¿À »ý¼ºÀ¸·Î È®ÀåÇÏ¿©, Ãß°¡ ÇнÀ ¾øÀ̵µ »ý¼º ¿µ»óÀÇ ½Ã°£Àû Àϰü¼ºÀ» Çâ»ó½ÃŰ´Â »õ·Î¿î guidance ¹æ¹ýÀ» Á¦¾ÈÇß½À´Ï´Ù. ¿ì¸®ÀÇ ¿¬±¸´Â ºñµð¿À DiTÀÇ ³»ºÎ ÀÛµ¿ ¿ø¸®¿¡ ´ëÇÑ Áß¿äÇÑ ÅëÂûÀ» Á¦°øÇϸç, À̵éÀÇ ½Ã°£Àû ÀÌÇØ(temporal understanding) ¸¦ Ȱ¿ëÇÑ ÇâÈÄ ¿¬±¸ ¹× ÀÀ¿ëÀÇ ±âÃʸ¦ È®¸³ÇÑ´Ù°í ¹Ï½À´Ï´Ù.
¹ÚÀºÇõ ±³¼ö´Â ¼¿ï´ëÇб³ ÄÄÇ»ÅͰøÇкο¡¼ ¹Ú»çÇÐÀ§¸¦ ¹ÞÀº ÈÄ Æ÷Ç×°ø°ú´ëÇб³ ÀΰøÁö´É´ëÇпø¿¡¼ 5³â°£ ÀçÁ÷ÁßÀÌ´Ù. µö·¯´× ¸ðµ¨ °æ·®È ¹× °¡¼Ó°ú °ü·ÃµÈ ¿¬±¸¸¦ ÅëÇØ ÀúÀü·Â °í¼º´É AI ¸ðµ¨ÀÇ °³¹ß ¹× ¿¬»ê ½Ã½ºÅÛ ÃÖÀûÈ¿¡ ÁýÁßÇϰí ÀÖ´Ù.
-
Professor Eunhyeok Park received his Ph.D. in Computer Science and Engineering from Seoul National University and has been serving at the Graduate School of Artificial Intelligence at POSTECH for five years. His research focuses on developing low-power, high-performance AI models and optimizing computing systems through studies on deep learning model compression and acceleration.
¹ßÇ¥¿ä¾à
GraLoRA: °íÈ¿À² °í¼º´É ÆÄÀÎÆ©´×À» À§ÇÑ ¼¼ºÐÈµÈ Àú·©Å© ÇнÀ±â¹ý
LoRA´Â ÆÄ¶ó¹ÌÅÍ È¿À²Àû ÆÄÀÎÆ©´×(PEFT) ±â¹ý Áß¿¡¼µµ Àß ¾Ë·ÁÁø ¹æ¹ýÀÔ´Ï´Ù. ·©Å© 32–64 ±¸°£¿¡¼ °¡Àå ÁÁÀº ¼º´ÉÀ» º¸ÀÌÁö¸¸, ±× À̻󿡼´Â Á¤È®µµ°¡ Á¤Ã¼µÇ°Å³ª ¶³¾îÁ® ¾ÆÁ÷Àº Àüü ÆÄÀÎÆ©´×(full fine tuning)¿¡ ¹ÌÄ¡Áö ¸øÇÕ´Ï´Ù. ¿ì¸®´Â ±× ¿øÀÎÀ» LoRAÀÇ ±¸Á¶Àû º´¸ñ¿¡¼ ã½À´Ï´Ù. ÀÌ º´¸ñ ¶§¹®¿¡ ¼·Î °ü·Ã ¾ø´Â ÀԷ ä³Î±îÁö ±×·¡µð¾ðÆ®°¡ ¾ôÈ÷°í, ±×·¡µð¾ðÆ® Àü´ÞÀÌ ¿Ö°îµË´Ï´Ù. À̸¦ Ç®±â À§ÇØ °¡ÁßÄ¡ Çà·ÄÀ» ÀÛÀº ºí·ÏÀ¸·Î ³ª´©°í °¢ ºí·Ï¿¡ Àú·©Å© ¾î´ðÅ͸¦ µÎ´Â Granular Low Rank Adaptation(GraLoRA)¸¦ Á¦¾ÈÇÕ´Ï´Ù. GraLoRA´Â ¿¬»ê°ú ÀúÀå Ãø¸éÀÇ Ãß°¡ ºñ¿ëÀÌ °ÅÀÇ ¾øÀ¸¸é¼ LoRAÀÇ ÇѰ踦 ÁÙÀ̰í Ç¥Çö·ÂÀ» ³ÐÇô, Àüü ÆÄÀÎÆ©´×ÀÇ µ¿ÀÛ¿¡ ´õ °¡±î¿öÁöµµ·Ï µ½½À´Ï´Ù.
±âȹ¼¼¼Ç 3-3
Physical AI Safety & Security
ÇÇÁöÄà AI ¾ÈÀü ¹× º¸¾È
±èº´±Ô ±³¼ö´Â ±¤¿î´ëÇб³¿¡¼ ¹æÀ§»ç¾÷ °øÇйڻçÇÐÀ§¸¦ ÃëµæÇÏ¿´À¸¸ç, ÇöÀç´Â ¼º±Õ°ü´ëÇб³ ÀΰøÁö´ÉÀ¶ÇÕ¿ø »êÇÏ ¹Ì·¡±¹¹æÀ¶ÇÕ¿¬±¸¼¾ÅÍÀåÀ» ¸Ã¾Æ¼ ¹Ì·¡ ±¹¹æºÐ¾ß ¹ßÀüÀ» À§ÇÑ Á¤Ã¥¿¬±¸¿Í ÀÎÀç¾ç¼ºÀ» ´ã´çÇϰí ÀÖ´Ù. ƯÈ÷ ±¹¹æºÎ ÁÖ°ü»ç¾÷ÀÎ ¡®±º Æ¯È AI ±³À°°úÁ¤¡¯À» ´ã´çÇÏ¿© ±º °£ºÎµéÀÇ AI ±³À°À» ÁÖ°üÇϰí ÀÖ´Ù.
-
Professor Kim Byung-kyoo earned a Ph.D. in Defense Engineering from Kwangwoon University and is currently the head of the Future Defense Convergence Research Center under the Artificial Intelligence Convergence Institute at Sungkyunkwan University, where he is responsible for policy research and talent development for the future development of the defense sector. In particular, it is in charge of the 'Military-Specific AI Training Course,' a project led by the Ministry of National Defense, and oversees AI education for military officers.
¹ßÇ¥¿ä¾à
-
ÀÌÇØ¼ö´Â ÇѾç´ëÇб³¿¡¼ ¹Ìµð¾î ±â¼ú&°æ¿µ ¹Ú»çÇÐÀ§¸¦ ¹ÞÀº ÈÄ ¼ÒÇÁÆ®¿þ¾îÁ¤Ã¥¿¬±¸¼ÒÀÇ AIÁ¤Ã¥¿¬±¸½Ç¿¡¼ ¼±ÀÓ¿¬±¸¿øÀ¸·Î ±Ù¹«Çϰí ÀÖ´Ù. °úÇбâ¼úÁ¤º¸Åë½ÅºÎÀÇ AI ±¹°¡ Àü·« ¼ö¸³À» À§ÇÑ AI »ê¾÷, À¶ÇÕ µî ´Ù¾çÇÑ Á¤Ã¥ ¿¬±¸¸¦ ´ã´çÇϰí ÀÖ´Ù.
-
Hae-soo Lee earned a Ph.D. in Media Technology and Management from Hanyang University and joined the Artificial Intelligence Policy Research Department at the Software Policy & Research Institute(SPRi) as a Senior Research Fellow. He is responsible for conducting policy research on AI industries and convergence to support the establishment of national AI strategies led by the Ministry of Science and ICT.
¹ßÇ¥¿ä¾à
ÇÇÁöÄà AI·Î ÀÎÇÑ ±¹°¡ Â÷¿øÀÇ À§Çù°ú ´ëÀÀ¹æ¾È ¹× ÁÖ¿ä±¹ µ¿Çâ
º» °¿¬Àº ÀΰøÁö´ÉÀÌ °¡»ó °ø°£À» ³Ñ¾î ¹°¸®Àû ¼¼°è¿¡¼ Àνġ¤ÆÇ´Ü¡¤ÇൿÀ» ¼öÇàÇÏ´Â ¡®ÇÇÁöÄà AI(Physical AI)¡¯ÀÇ ºÎ»ó°ú ÀÌ¿¡ µû¸¥ ±¹°¡ Â÷¿øÀÇ À§Çù ¹× ´ëÀÀ Àü·«À» ´Ù·é´Ù. ÀÚÀ²Çü µå·Ð°ú ÈÞ¸Ó³ëÀÌµå µî ÇÇÁöÄà AIÀÇ È®»êÀº »ê¾÷ È¿À²¼ºÀ» ³ôÀÌ´Â µ¿½Ã¿¡, ¹°¸®Àû °ø°ÝÀÇ À§Çè Áõ°¡¿Í »çÀ̹ö-¹°¸® À¶ÇÕ À§ÇùÀ̶ó´Â »õ·Î¿î ¾Èº¸ ¸®½ºÅ©¸¦ ¾ß±âÇÑ´Ù. ¶ÇÇÑ ¿§Áö ÄÄÇ»ÆÃ ±â¹Ý µ¥ÀÌÅÍ Ã³¸® È®´ë·Î ÀÎÇÑ Á¤º¸º¸¾È ¹®Á¦, AI °áÁ¤ °úÁ¤ÀÇ ºÒÅõ¸í¼º(¡®ºí·¢¹Ú½º À§Ç衯), ±×¸®°í ÈÞ¸Ó³ëÀ̵塤·Îº¿ È®»êÀ¸·Î ÀÎÇÑ ³ëµ¿½ÃÀå ±¸Á¶ º¯È µî »çȸÀû ºÒ¾ÈÁ¤ ¿äÀεµ ÇÔ²² Á¦±âµÈ´Ù.
½Éº´¼·Àº Çѱ¹Ç×°ø´ëÇб³ Ç×°ø±â°è°øÇÐ Çл縦 ¹ÞÀº ÈÄ, »ï¼ºÇ×°ø Ç×°ø¿ìÁÖ¿¬±¸¼Ò CAEÆÀ¿¡ ÀÔ»çÇÏ¿© T-50 µðÁöÅиñ¾÷ °³¹ß ¾÷¹«¸¦ ½ÃÀÛÀ¸·Î ÀÌÈÄ Çѱ¹Ç×°ø¿ìÁÖ»ê¾÷ Á¤º¸½Ã½ºÅÛ ºÐ¾ß¿¡¼ Ç×°ø±â ¼³°è, ÇØ¼®¿¡ »ç¿ëµÇ´Â IT ½Ã½ºÅÛ ±âȹ ¹× °ü¸®¾÷¹«¸¦ ¼öÇà, ÇöÀç´Â À¯¹«Àκ¹ÇÕ ¹× ¹«ÀαâÀÇ ÀÚÀ²ÀÓ¹«¸¦ ¼öÇàÇÏ´Â AI Pilot °³¹ßÆÀ ¸®´õ¸¦ ¸Ã°í ÀÖ´Ù.
-
Byoungserb Shim received his B.S. in Aerospace and Mechanical Engineering from Korea Aerospace University. He began his career at the Samsung Aerospace as part of the CAE team, contributing to the digital mock-up development of the T-50 advanced trainer. He later joined Korea Aerospace Industries (KAI), where he led planning and management of enterprise engineering IT systems supporting aircraft design and structural analysis workflows. He currently serves as the Team Lead for AI Development, focusing on autonomous mission execution for unmanned and manned-unmanned teaming (MUM-T) platforms.
¹ßÇ¥¿ä¾à
±Û·Î¹ú ¹«Àαâ AI Pilot ±â¼ú ¹ßÀü°ú ÇÇÁöÄà AI Àû¿ë °úÁ¦
ÀΰøÁö´ÉÀÇ ±Þ°ÝÇÑ ¹ßÀüÀº »çȸ °¢ ºÐ¾ßÀÇ ÀÌÀü°ú´Â ´Ù¸¥ º¯È¸¦ ÃÊ·¡Çϰí ÀÖÀ¸¸ç, ƯÈ÷ ±¹¹æºÐ¾ß´Â ÀΰøÁö´ÉÀÌ °¡Á®¿Ã ¹Ì·¡ ÀüÀåȯ°æ º¯È¿¡ ´ëÀÀÇϱâ À§ÇÑ ±¹¹æ AI°³¹ßÀ» ¼µÎ¸£°í ÀÖ´Ù.Æóȸ°¿¬
±âȹ¼¼¼Ç 4(¿¬°è¼¼¼Ç)
¸Å»çÃß¼¼Ã÷°ø°ú´ëÇÐ ´ëÇпø ¹Ú»ç
Çѱ¹°úÇбâ¼ú¿ø »ê¾÷ ¹× ½Ã½ºÅÛ °øÇаú ±³¼ö
- ¼°´ëÇб³ ´ëÇпø ÄÄÇ»ÅͰøÇÐ ¼®»ç
- ¼°´ëÇб³ °ø°ú´ëÇÐ ÄÄÇ»ÅͰøÇаú Á¹¾÷
- »ï¼ºÀüÀÚ ¿µ»óµð½ºÇ÷¹ÀÌ»ç¾÷ºÎ °³¹ßÆÀ »ó¹«
- »ï¼ºÁ¾ÇÕ±â¼ú¿ø µðÁöÅבּ¸¼Ò Àü¹®¿¬±¸¿ø