Hans J. Berliner Lecture in Artificial Intelligence - Yejin Choi

— 6:00pm

Location:
In Person - Rashid Auditorium, Gates Hillman 4401

Speaker:
YEJIN CHOI , Dieter Schwarz Foundation Professor and Senior Fellow
Department of Computer Science, and the Institute for Human-Centered Artificial Intelligence
Stanford University
and Distinguished Research Scientist, NVIDIA
https://yejinc.github.io/

Bending Scaling Laws with Brighter Algorithms

Scaling laws tell us that "more is more" — brute-force scaling of data and compute leads to smooth improvements in AI capabilities. However, this approach is becoming increasingly unsustainable in practice, creating a need for algorithmic innovations that can bend the scaling laws and achieve more compute-efficient progress. In this talk, I will discuss our recent work in this direction, including gradient-based methods for synthetic data generation, prolonged reinforcement learning that can unlock stronger reasoning capabilities from smaller models, symbolic search algorithms for test-time reasoning, test-time training that extracts additional learning even during inference, and a new tokenization algorithm that enables better and faster inference. 

—

Yejin Choi is the Dieter Schwarz Foundation Professor and Senior Fellow at the Department of Computer Science at Stanford University and the Stanford Institute for Human-Centered Artificial Intelligence (HAI). She is also a Distinguished Research Scientist at NVIDIA, and was previously Professor at the University of Washington and Senior Director at the Allen Institute for AI (Ai2). Choi is a MacArthur Fellow (class of 2022), AI2050 Senior Fellow (class of 2024), and was named among Time100 Most Influential People in AI in 2023. She is a co-recipient of 2 Test-of-Time awards and 9 Best and Outstanding Paper Awards at top AI conferences including ACL, ICML, NeurIPS, ICCV, CVPR, and AAAI. She received the Borg Early Career Award (BECA) in 2018, won the inaugural Alexa Prize Challenge in 2017, and was named one of IEEE AI's 10 to Watch in 2016. Choi was a main stage speaker at TED 2023 and has delivered keynote talks at conferences across AI disciplines including ACL, CVPR, ICLR, MLSys, VLDB, WebConf, and AAAI. Her current research interests focus on inference-time scaling for neural language models, large and small reasoning models, symbolic methods for neural language models, alternative training recipes for language models, synthetic data generation for generative AI, and pluralistic alignment. 

Faculty Hosts:  Vincent Conitzer, Aditi Raghunathan

About the Lecture:  The Hans J. Berliner Lecture in Artificial Intelligence has been established in tribute to Hans J. Berliner, (CS'74)  in recognition of the significant and critical accomplishments as faculty, researcher, advisor, and exemplary colleague and friend to many.  This endowed lecture is presented by the Computer Science Department, in conjunction with the SCS Distinguished Lecture Series, and will let us reflect on Hans' contributions and all they enabled. 
 

For More Information:
scs-dls@cs.cmu.edu

Add event to Google
Add event to iCal