ML Researcher at ANTHROPIC / STANFORD
At Anthropic, I work on RL for large language models (LLMs) and retrieval-augmented LLMs. It’s exciting to see the benefits of RL in the domain of human language, e.g. RLHF (RL from Human Feedback) and RLAIF (RL from AI Feedback).
Before that, I was a Staff Research Scientist and co-founder/lead of the ML for Systems team at Google Brain, where my research focused on developing deep reinforcement approaches to problems in computer systems, particularly chip design. My RL methods have been used in multiple generations of Google's flagship AI accelerator (TPU).
I graduated from MIT with a Bachelors in Computer Science, a Bachelors in Linguistics, and a Masters of Computer Science, and I am a Computer Science PhD Candidate in the Stanford NLP Group.
Ganguli et al. The Capacity for Moral Self-Correction in Large Language Models. arXiv 2023.
Bai et al. Constitutional AI: Harmlessness from AI Feedback. arXiv 2022.
Bowman et al. Measuring Progress on Scalable Oversight for Large Language Models. arXiv 2022.
Azalia Mirhoseini*, Anna Goldie* (co-first author), Mustafa Yazgan, Joe Jiang, Ebrahim Songhori, Shen Wang, Young-Joon Lee, Eric Johnson, Omkar Pathak, Azade Nazi, Jiwoo Pak, Andy Tong, Kavya Srinivasa, William Hang, Emre Tuncer, Quoc V. Le, James Laudon, Richard Ho, Roger Carpenter, Jeff Dean. A Graph Placement Methodology for Fast Chip Design (Open Access, Review Article). Nature.
Anna Goldie and Azalia Mirhoseini. "Placement Optimization with Deep Reinforcement Learning". ISPD 2020.
Yanqi Zhou, Sudip Roy, Amirali Abdolrashidi, Daniel Wong, Peter C Ma, Qiumin Xu, Ming Zhong, Hanxiao Liu, Anna Goldie, Azalia Mirhoseini, James Laudon. "A Generalized Approach to End-to-End Computational Graph Optimization". NeurIPS 2020 (Oral).
Azalia Mirhoseini*, Anna Goldie* (co-first author), Hieu Pham, Benoit Steiner, Quoc V. Le, Jeff Dean. “A Hierarchical Model for Device Placement,” ICLR 2018.
Denny Britz*, Anna Goldie* (co-first author), Thang Luong, Quoc V. Le. “Massive exploration of neural machine translation architectures,” EMNLP 2017.
Louis Shao, Stephan Gouws, Denny Britz, Anna Goldie, Brian Strope, Ray Kurzweil. “Generating high-quality and informative conversation responses with sequence-to-sequence models,” EMNLP 2017. (Co-first author on workshop paper which won best paper award at ICML 2017 Language Generation Workshop)
Azade Nazi, Will Hang, Anna Goldie, Sujith Ravi, Azalia Mirhoseini. "Clustering by Learning to Optimize Normalized Cuts". NeurIPS 2019 Workshop on Sets & Partitions.
Yushi Xu, Anna Goldie, and Stephanie Seneff, "Automatic Question Generation and Answer Judging: A Q&A Game for Language Learning," Proc. SIGSLaTE 2009.
Anna Goldie, “CHATTER: A Spoken Language Dialogue System For Language Learning Applications,” MIT Masters Thesis, June 2011.
Azade Nazi, Will Hang, Anna Goldie, Sujith Ravi, Azalia Mirhoseini. "Generalizable Approximate Graph Partitioning Framework". ICLR 2019 Workshop on Representation Learning on Graphs and Manifolds.
Jialin Song, Joe Jiang, Ebrahim Songhori, Anna Goldie, Navdeep Jaitly, Azalia Mirhoseini. "Policy Optimization By Local Improvement Through Search". ICML 2020 Workshop on Inductive Biases, Invariances and Generalization in RL.
Awards and Recognition
MIT Technology Review’s “35 Innovators Under 35” (TR35 Global 2021)
Best Paper Award at the ICML 2017 Workshop on Generating Natural Language
Organizer for NeurIPS’18, NeurIPS’19, NeurIPS'20, and ISCA’19 Workshops on Machine Learning for Systems
Organizer for West Coast NLP 2018, 2019, and 2020 (founding organizer and MC this year)
Keynote Speaker for Google Developer Day in China 2017, 2018, and 2019 (in Mandarin, livestreamed to 10M+ people, Machine Learning and TensorFlow)
Invited Speaker for the Deep Learning Session of CRA's Distinguished Lecture Series
Guest Lecturer at Caltech, UCSD, UC Berkeley
Stanford School of Engineering Fellowship 2019
AMTA Panelist for “Deploying Open Source Neural Machine Translation Toolkits in the Enterprise”
ACM Learning Center Interview: http://stephenibaraki.com/acm/interviews/v0620/a_mirhoeseini_a_goldie_acm.html