Hattie Zhou: Lottery Tickets and Algorithmic Reasoning in LLMs

Hattie Zhou: Lottery Tickets and Algorithmic Reasoning in LLMs

Author: Daniel Bashir February 16, 2023 Duration: 1:42:59

In episode 60 of The Gradient Podcast, Daniel Bashir speaks to Hattie Zhou.

Hattie is a PhD student at the Université de Montréal and Mila. Her research focuses on understanding how and why neural networks work, based on the belief that the performance of modern neural networks exceeds our understanding and that building more capable and trustworthy models requires bridging this gap. Prior to Mila, she spent time as a data scientist at Uber and did research with Uber AI Labs.

Have suggestions for future podcast guests (or other feedback)? Let us know here!

Subscribe to The Gradient Podcast:  Apple Podcasts  | Spotify | Pocket Casts | RSSFollow The Gradient on Twitter

Outline:

* (00:00) Intro

* (01:55) Hattie’s Origin Story, Uber AI Labs, empirical theory and other sorts of research

* (10:00) Intro to the Lottery Ticket Hypothesis & Deconstructing Lottery Tickets

* (14:30) Lottery tickets as lucky initialization

* (17:00) Types of masking and the “masking is training” claim

* (24:00) Type-0 masks and weight evolution over long training trajectories

* (27:00) Can you identify good masks or training trajectories a priori?

* (29:00) The role of signs in neural net initialization

* (35:27) The Supermask

* (41:00) Masks to probe pretrained models and model steerability

* (47:40) Fortuitous Forgetting in Connectionist Networks

* (54:00) Relationships to other work (double descent, grokking, etc.)

* (1:01:00) The iterative training process in fortuitous forgetting, scale and value of exploring alternatives

* (1:03:35) In-Context Learning and Teaching Algorithmic Reasoning

* (1:09:00) Learning + algorithmic reasoning, prompting strategy

* (1:13:50) What’s happening with in-context learning?

* (1:14:00) Induction heads

* (1:17:00) ICL and gradient descent

* (1:22:00) Algorithmic prompting vs discovery

* (1:24:45) Future directions for algorithmic prompting

* (1:26:30) Interesting work from NeurIPS 2022

* (1:28:20) Hattie’s perspective on scientific questions people pay attention to, underrated problems

* (1:34:30) Hattie’s perspective on ML publishing culture

* (1:42:12) Outro

Links:

* Hattie’s homepage and Twitter

* Papers

* Deconstructing Lottery Tickets: Zeros, signs, and the Supermask

* Fortuitous Forgetting in Connectionist Networks

* Teaching Algorithmic Reasoning via In-context Learning



Get full access to The Gradient at thegradientpub.substack.com/subscribe

Hosted by Daniel Bashir, The Gradient: Perspectives on AI moves beyond surface-level headlines to explore the intricate machinery and human ideas shaping artificial intelligence. Each episode is built on a foundation of deep research, leading to conversations that are both technically substantive and broadly accessible. You'll hear from researchers, engineers, and philosophers who are actively building and critiquing our technological future, discussing not just how AI systems work, but the larger implications of their integration into society. This isn't about speculative hype; it's a grounded examination of real progress, persistent challenges, and ethical considerations from those on the front lines. The discussions peel back layers on topics like model architecture, policy, and the fundamental science behind the algorithms becoming part of our daily lives. For anyone curious about the substance behind the buzz-whether you have a technical background or are simply keen to understand a defining technology of our age-this podcast offers a crucial and thoughtful resource. Tune in for a consistently detailed and nuanced take that treats artificial intelligence with the complexity it deserves.
Author: Language: English Episodes: 100

The Gradient: Perspectives on AI
Podcast Episodes
Stevan Harnad: AI's Symbol Grounding Problem [not-audio_url] [/not-audio_url]

Duration: 1:58:21
In episode 88 of The Gradient Podcast, Daniel Bashir speaks to Professor Stevan Harnad.Stevan Harnad is professor of psychology and cognitive science at Université du Québec à Montréal, adjunct professor of cognitive sci…
Terry Winograd: AI, HCI, Language, and Cognition [not-audio_url] [/not-audio_url]

Duration: 1:33:21
In episode 87 of The Gradient Podcast, Daniel Bashir speaks to Professor Terry Winograd. Professor Winograd is Professor Emeritus of Computer Science at Stanford University. His research focuses on human-computer interac…
Gil Strang: Linear Algebra and Deep Learning [not-audio_url] [/not-audio_url]

Duration: 1:00:36
In episode 86 of The Gradient Podcast, Daniel Bashir speaks to Professor Gil Strang. Professor Strang is one of the world’s foremost mathematics educators and a mathematician with contributions to finite element theory,…
Anant Agarwal: AI for Education [not-audio_url] [/not-audio_url]

Duration: 47:40
In episode 85 of The Gradient Podcast, Andrey Kurenkov speaks to Anant AgarwalAnant Agarwal is the chief platform officer of 2U, and founder of edX. Anant taught the first edX course on circuits and electronics from MIT,…
Peli Grietzer: A Mathematized Philosophy of Literature [not-audio_url] [/not-audio_url]

Duration: 2:33:33
In episode 83 of The Gradient Podcast, Daniel Bashir speaks to Peli Grietzer. Peli is a scholar whose work borrows mathematical ideas from machine learning theory to think through “ambient” and ineffable phenomena like m…
Ryan Drapeau: Battling Fraud with ML at Stripe [not-audio_url] [/not-audio_url]

Duration: 1:06:31
In episode 82 of The Gradient Podcast, Daniel Bashir speaks to Ryan Drapeau.Ryan is a Staff Software Engineer at Stripe and technical lead for Stripe’s Payment Fraud organization, which uses machine learning to help prev…
Shiv Rao: Enabling Better Patient Care with AI [not-audio_url] [/not-audio_url]

Duration: 1:00:51
In episode 81 of The Gradient Podcast, Daniel Bashir speaks to Shiv Rao.Shiv Rao, MD is the co-founder and CEO of Abridge, a healthcare conversation company that uses cutting-edge NLP and generative AI to bring context a…
Hugo Larochelle: Deep Learning as Science [not-audio_url] [/not-audio_url]

Duration: 1:48:28
In episode 80 of The Gradient Podcast, Daniel Bashir speaks to Professor Hugo Larochelle. Professor Larochelle leads the Montreal Google DeepMind team and is adjunct professor at Université de Montréal and a Canada CIFAR…
Jeremie Harris: Realistic Alignment and AI Policy [not-audio_url] [/not-audio_url]

Duration: 1:30:35
In episode 79 of The Gradient Podcast, Daniel Bashir speaks to Jeremie Harris.Jeremie is co-founder of Gladstone AI, author of the book Quantum Physics Made Me Do It, and co-host of the Last Week in AI Podcast. Jeremy pr…