BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:"Machine learning in chemistry and beyond" (ChE-651) seminar by Dr
 . Tong Xie: "From Token to Discovery: A New Paradigm in Material Discovery
 "
DTSTART:20250218T151500
DTEND:20250218T161500
DTSTAMP:20260526T012342Z
UID:e16dedaf14388445bea63a63da6f0a8c603d64183b6d0e5319dccf05
CATEGORIES:Conferences - Seminars
DESCRIPTION:Tong Xie gained his PhD from the School of Photovoltaic and Re
 newable Energy Engineering (SPREE)\, UNSW Sydney\, acclaimed as one of Aus
 tralia’s National Computational Infrastructure’s Top 10 HPC AI-Talents
 . As the CEO of GreenDynamics and the Group Lead of UNSW AI4Science\, he i
 s pioneering the use of Generative AI to accelerate the discovery and deve
 lopment of sustainable materials. His expertise extends to Natural Languag
 e Processing and Material Science. He also founded the DARWIN natural scie
 nce language model\, demonstrating his innovative approach to advancing AI
  in material sciences.\nRecent advances in artificial intelligence\, parti
 cularly large language models (LLMs)\, have opened new frontiers in chemic
 al research. While these models traditionally excel at text processing\, o
 ur research demonstrates their capability to solve practical chemistry pro
 blems through an innovative approach that combines open-source language mo
 dels with extensive chemical datasets. By training our models on over 188\
 ,000 chemistry-focused examples across 24 subdisciplines and developing th
 e Scientific Question Answering Generation (SciQAG) model for automated da
 ta preparation\, we've achieved significant breakthroughs in efficiency an
 d accuracy. Our research reveals that simultaneous training across multipl
 e chemical tasks leads to superior performance compared to single-task app
 roaches\, suggesting the models can recognize fundamental connections betw
 een different areas of chemistry. Through rigorous testing across 49 chemi
 stry-related tasks\, our open-source approach not only matches or exceeds 
 traditional machine learning methods but also ensures these advanced tools
  remain accessible to the entire scientific community\, potentially accele
 rating the pace of chemical discovery.
LOCATION:https://epfl.zoom.us/j/68447908297?pwd=OU5JUGJUSUhZc0ZNYjQ2WENvYl
 NRdz09
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR