演講主題： Teaching GPTs to Reason

演講時間及地點：2024/11/27星期三 13:00~14:00 臺大博理館101演講廳 (同步於博理館112、113室直播演講實況，均須報名、當日報到方能入場)

講者：Mark Chen 陳信翰 (OpenAI 資深研究副總)

Abstract：

OpenAI o1 excels in reasoning-intensive tasks, ranking in the 89th percentile on Codeforces, placing in the top 500 in AIME, and surpassing PhD-level accuracy on science benchmarks. Powered by a novel reinforcement learning approach that leverages chain-of-thought reasoning, o1’s performance improves with increasing train-time and test-time compute. This talk explores a history of reasoning in LLMs and the capabilities and potential of o1.

Bio：

Mark Chen is the Senior Vice President of Research at OpenAI, where he oversees advanced AI initiatives, driving innovation in language models, reinforcement learning, multimodal models, and AI alignment. Since joining in 2018, he has played a key role in shaping the organization's most ambitious projects, including leading Codex, DALL·E, 4o, and o1. Mark is dedicated to ensuring AI developments benefit society while maintaining a focus on responsible research. Prior to joining OpenAI, Mark worked as a quantitative trader at several proprietary trading firms, including Jane Street Capital, where he built machine learning algorithms for equities and futures trading. He graduated from MIT with a bachelor’s degree in mathematics with computer science, and is also a coach for the USA Computing Olympiad team.  

主辦單位：聯發科技-臺大創新研究中心、臺大系統晶片中心

協辦單位：臺大電子工程學研究所

報名連結: https://forms.gle/RYNRfnzYVHVymagU7

因座位有限，請各位教授與同學務必提前報名，當日報到後方能入座聽講