Showcasing the infinite possibilities of AI in the future, Foxconn Research Institute's AI NExT Forum debuts
AI experts of all levels gather together, AI 2.0, AI+ era changes industrial intelligence
Computing power replaces organizational manpower as a new definition of enterprise scale Efficiency and energy consumption will be the key to AI development
【Taipei, Taiwan - July 23, 2024】Hon Hai Research Institute today (23) held the 'AI NExT Forum' to celebrate Hon Hai's 50th anniversary. This year's forum is themed 'Hon Hai 50 Years, Smart New Era: Generative AI and Future Innovation,' inviting numerous domestic and international heavyweight experts and scholars to gather, attracting a total of 964 online and offline audiences. The discussions ranged from the transition of AI to the 2.0 era, the bottlenecks of efficiency and energy consumption in the development of generative AI, to the redefinition of enterprise scale by computing power. During the forum, Hon Hai also revealed for the first time the phased achievements of the group's three major platforms in AI application.
AI enters the AI 2.0 era Computing power represents the size of the enterprise
In the keynote speech at the 'AI NExT Forum,' Foxconn Research Institute advisory committee member, also CEO of Sinovation Ventures and Chairman of Innovation Works, Kai-Fu Lee pointed out that Generative AI is the most disruptive technological revolution in history! He emphasized, 'Generative AI has entered the AI 2.0 era and will redefine industrial intelligence. The biggest difference is what AI can do, evolving from the single-task AI of the AI 1.0 era to models trained through global data and multimodal training, capable of universally answering any question, producing any content universally, and even performing analysis and reasoning that only humans could do in the past.'
Dr. Li Kaifu further stated, 'AI 2.0 has passed the classic Turing test and is rapidly moving towards general artificial intelligence, possessing high IQ, becoming the most important productivity enhancement tool in human history, ushering in the era of AI assistants, completely changing user interfaces and business models, becoming the greatest platform revolution, and creating 10 times more value than the mobile internet era.' CEOs should not just expect to add a new assistant to the organization, but hope that the company can master large-scale model technology and make it more powerful with internal data.
The key to the popularization of large generative AI models: efficiency and energy consumption
Large-scale generative AI models, such as text generation like ChatGPT and image generation like Dall-E, are key drivers of recent AI advancements, bringing significant social and economic impacts. However, these large generative models typically require training with tens of billions of tokens and thousands of billions of parameters, leading to massive computational resources and energy consumption.
IEEE Fellow Professor Jianfei Cai from the School of Information Technology at Monash University in Australia stated, 'Leading the development of artificial intelligence, generative AI large models currently rely on further enhancing intelligence mainly by increasing scale, using larger models, and more data. Countless small and medium-sized enterprises, numerous research institutions, countless individual developers, and researchers find it difficult to cope with the huge computational resource requirements brought by large generative models. How to involve a large number of small and medium-sized enterprises, individual developers, and researchers in the development of large generative AI models is a very important issue.'
From AI to AI+ Human Collaboration with Large-scale Language will bring new opportunities to the industry
Professor Zheng Wenhuang, a distinguished lecturer at National Taiwan University and an IEEE Fellow, emphasizes that the development of artificial intelligence has evolved from the era of large-scale AI models to the era of AI agents. AI agents are no longer static applications that follow predetermined rules, but are capable of continuous learning and self-improvement using large models, and adapting flexibly to constantly changing environments and situations.
Researcher Kung-Luen Wei of the Institute of Information Science at Academia Sinica also expressed that the collaboration between humans and artificial intelligence is redefining the way we work. The latest human-LLM collaboration technology combines human creativity, judgment, and the precision and efficiency of LLM, further enhancing the practicality of the technology. In the foreseeable future, as technology continues to advance, it will assist humans in achieving higher productivity and innovation capabilities.
Foxconn Research Institute collaborates with the group's three major platforms to announce the latest research findings and innovative applications
Hon Hai's use of generative AI to accelerate smart manufacturing, smart electric vehicles, and smart city applications is also a focus of industry attention. At this year's forum, senior executives from Hon Hai Technology Group's three major platforms also shared the current stage of research results.
Researcher Zhan Nianyi from the Artificial Intelligence Research Institute of Foxconn stated that the intelligent simulation agent (BehaviorGPT) proposed by the institute in collaboration with City University of Hong Kong for closed-loop testing of autonomous driving algorithms won the first place in the 2024 Waymo Sim Agents Challenge. The agent simulates behaviors in autonomous driving systems, enabling rapid and safe evaluation of the planning module's capabilities through virtual simulation environments, which can assist in the development of advanced autonomous driving algorithms.
Hon Hai Research Institute's AI Research Institute researcher Liao Yibin emphasized that the institute has invested a large amount of manpower in the field of AI and autonomous driving, improving core technologies from basic models and algorithms, publishing multiple international journal and top conference papers every year, and its research results have been recognized worldwide. The ultimate goal is to use multimodal language models as the basic models of the three major platforms to support various applications within the group.
In the field of smart manufacturing, this time, Dr. Jiang Wen-Hung from the Central Intelligence Manufacturing Department/Intelligent Manufacturing Platform Empowerment Business Department of Foxconn and Central University's Ph.D. in Information Engineering took the stage. He introduced how Foxconn builds innovative applications with MFG AI and GenAI advanced technologies as the core, creating a cross-generational AI smart factory. The deep integration of MFG AI and GenAI technologies in Foxconn's AI smart factory will help Foxconn achieve the perfect combination of personalized customization and mass production, providing customers worldwide with better services and products.
Director Yang Changhua of the Smart Driving Engineering Department at Foxconn shared how Foxconn utilizes NVIDIA's automotive computing platform to analyze road environment images in real time using AI capabilities. This technology can be applied to navigation-assisted driving systems, even when facing uncommon targets on the road, such as irregularly shaped vehicles, pedestrians in various postures, road construction, medians, fences, and other elements. In different usage environments and under various weather conditions, automotive AI can adapt accordingly, ensuring the safety of drivers, passengers, and pedestrians.
Foxconn's Smart City Platform Director Guo Jinbin personally demonstrated the achievements of Foxconn's CityGPT generative AI. He emphasized that with Foxconn's 50 years of strength in software and hardware integration, it will serve as the top-level integrator in smart cities, dedicated to solving the problem of urban information silos, building a cross-disciplinary neural hub platform, and providing smart services for urban residents, businesses, and government agencies. In the future, it will drive the smart city industry chain with software, ushering in the wave of generative AI. Its smart solutions include the CityGPT platform and smart applications, as well as hardware outsourcing for smart electric vehicle equipment and computing centers. Foxconn is committed to interacting through the GenAI platform, sharing business interests, driving small and medium-sized businesses and local merchants, and forming a win-win industrial ecosystem.
AI Theme Forum Hot Participation Unprecedented
The AI NExT Forum with AI as the theme received enthusiastic responses from professionals in various fields this year, with over 964 participants in online and offline activities. The industries represented by participants include information technology, students, and education/research units, accounting for 55% of the total participants. This forum not only reveals to attendees that the AI 2.0 era is rapidly changing human work and life as never before, but also brings significant impacts on industry intelligence and corporate organizations. All industries should embrace AI technology with a new mindset of 'AI First,' master the technology of large language models, and utilize internal data to make AI an exclusive enterprise-level brain.
The establishment of the Foxconn Research Institute is one of the important development strategies for Foxconn Technology Group to move towards the F3.0 transformation and upgrade. The institute has five research institutes and one laboratory, each with an average of 40 top technical R&D talents, focusing on future 3-7 years of forward-looking technology research and development, strengthening Foxconn's technology and product innovation, upgrading the company from the past 'labor-intensive' to 'brain-intensive,' and enhancing the core competitiveness of Foxconn Technology Group's '3+3' development. For more information, please visit https://www.hh-ri.com/.