Dr. Trinh Hoang Trieu has introduced AlphaGeometry, a new theorem-proving system for Euclidean plane geometry, marking a significant advancement in human-level automated reasoning. This groundbreaking solution addresses the challenging problem of geometry, which is notoriously difficult and often lacks training data due to the complexity of converting human proofs into machine-verifiable formats.
AlphaGeometry overcomes the limitations of current machine learning methods by generating millions of theorems and proofs with varying complexity. Utilizing a neural language model trained from scratch on large-scale synthetic data, this system guides a symbolic reasoning engine through the infinite branching points of complex geometric problems. Notably, AlphaGeometry successfully solved 25 out of the 30 latest Olympic-level problems and produced human-readable proofs, a remarkable achievement in this field.
With its capability to solve all geometry problems from the 2000 and 2015 International Mathematical Olympiads (IMO) and even discover a generalized version of a theorem from IMO 2004, AlphaGeometry has approached and even surpassed the performance of some IMO gold medalists, setting a new benchmark for automated problem-solving.
During this year's Summer School 2024, the Faculty of Information Technology is honored to welcome Dr. Trinh Hoang Trieu to participate and share on the topic "AlphaGeometry: Solving IMO Geometry without Human Demonstrations." He is a senior research scientist at Google DeepMind, dedicating his career to developing AI systems capable of reasoning, with the hope that machines can one day assist humans in advancing scientific and technological progress. Dr. Trieu's presentation will not only shed light on the capabilities of AI in mathematics but also inspire deep interest in the future of this technology in education and scientific research.
Time: 08:00 - 10:00, Tuesday - July 23, 2024
Venue: Lecture Hall I, VNUHCM - University of Science (227 Nguyen Van Cu, Ward 4, District 5)
Deadline: Until 12:00 PM, Sunday, July 21, 2024
For inquiries, please contact Ms. Nguyen Tran Thuc Uyen via email at nttuyen@fit.hcmus.edu.vn