Martin Wattenberg: ML Visualization and Interpretability

In episode 99 of The Gradient Podcast, Daniel Bashir (twitter.com/spaniel_bashir) speaks to Professor Martin Wattenberg (twitter.com/wattenberg).
Professor Wattenberg is a professor at Harvard and part-time member of Google Research’s People + AI Research (PAIR) initiative, which he co-founded. His work, with long-time collaborator Fernanda Viégas, focuses on making AI technology broadly accessible and reflective of human values. At Google, Professor Wattenberg, his team, and Professor Viégas have created end-user visualizations for products such as Search, YouTube, and Google Analytics. Note: Professor Wattenberg is recruiting PhD students through Harvard SEAS—info here (seas.harvard.edu/prospective-students/prospect…).
Have suggestions for future podcast guests (or other feedback)? Let us know here (forms.gle/YNKGko9LCGNp1rhU7) or reach us at editor@thegradient.pub
Subscribe to The Gradient Podcast: Apple Podcasts (podcasts.apple.com/us/podcast/the-gradient-pod…) | Spotify (open.spotify.com/show/6onNcSqsP6hEEqmZ6TU2g8) | Pocket Casts (pca.st/itunes/1569777340) | RSS (api.substack.com/feed/podcast/265424/s/1354/pr…)Follow The Gradient on Twitter (twitter.com/gradientpub)
Outline:
* (00:00) Intro
* (03:30) Prof. Wattenberg’s background
* (04:40) Financial journalism at SmartMoney
* (05:35) Contact with the academic visualization world, IBM
* (07:30) Transition into visualizing ML
* (08:25) Skepticism of neural networks in the 1980s
* (09:45) Work at IBM
* (10:00) Multiple scales in information graphics, organization of information
* (13:55) How much information should a graphic display to whom?
* (17:00) Progressive disclosure of complexity in interface design
* (18:45) Visualization as a rhetorical process
* (20:45) Conversation Thumbnails for Large-Scale Discussions (citeseerx.ist.psu.edu/document?repid=rep1&…)
* (21:35) Evolution of conversation interfaces—Slack, etc.
* (24:20) Path dependence — mutual influences between user behaviors and technology, takeaways for ML interface design
* (26:30) Baby Names and Social Data Analysis (bewitched.com/namevoyager.html) — patterns of interest in baby names
* (29:50) History Flow (bewitched.com/historyflow.html)
* (30:05) Why investigate editing dynamics on Wikipedia?
* (32:06) Implications of editing patterns for design and governance
* (33:25) The value of visualizations in this work, issues with Wikipedia editing
* (34:45) Community moderation, bureaucracy
* (36:20) Consensus and guidelines
* (37:10) “Neutral” point of view as an organizing principle
* (38:30) Takeaways
* PAIR
* (39:15) Tools for model understanding and “understanding” ML systems
* (41:10) Intro to PAIR (at Google)
* (42:00) Unpacking the word “understanding” and use cases
* (43:00) Historical comparisons for AI development
* (44:55) The birth of TensorFlow.js (tensorflow.org/js)
* (47:52) Democratization of ML
* (48:45) Visualizing translation (arxiv.org/abs/1611.04558) — uncovering and telling a story behind the findings
* (52:10) Shared representations in LLMs and their facility at translation-like tasks
* (53:50) TCAV (arxiv.org/abs/1711.11279)
* (55:30) Explainability and trust
* (59:10) Writing code with LMs and metaphors for using
* More recent research
* (1:01:05) The System Model and the User Model: Exploring AI Dashboard Design (arxiv.org/abs/2305.02469)
* (1:10:05) OthelloGPT (arxiv.org/abs/2210.13382) and world models, causality
* (1:14:10) Dashboards and interaction design—interfaces and core capabilities
* (1:18:07) Reactions to existing LLM interfaces
* (1:21:30) Visualizing and Measuring the Geometry of BERT (proceedings.neurips.cc/paper_files/paper/2019/…)
* (1:26:55) Note/Correction: The “Atlas of Meaning” Prof. Wattenberg mentions is called Context Atlas (storage.googleapis.com/bert-wsd-vis/demo/index…)
* (1:28:20) Language model tasks and internal representations/geometry
* (1:29:30) LLMs as “next word predictors” — explaining systems to people
* (1:31:15) The Shape of Song (bewitched.com/song.html)
* (1:31:55) What does music look like?
* (1:35:00) Levels of abstraction, emergent complexity in music and language models
* (1:37:00) What Prof. Wattenberg hopes to see in ML and interaction design
* (1:41:18) Outro
Links:
* Professor Wattenberg’s homepage (bewitched.com) and Twitter (twitter.com/wattenberg)
* Harvard SEAS application info (seas.harvard.edu/prospective-students/prospect…) — Professor Wattenberg is recruiting students!
* Research
* Earlier work
* A Fuzzy Commitment Scheme (dl.acm.org/doi/pdf/10.1145/319709.319714)
* Stacked Graphs—Geometry & Aesthetics (cg.tuwien.ac.at/courses/InfoVis/papers/InfoVis…)
* A Multi-Scale Model of Perceptual Organization in Information Graphics (citeseerx.ist.psu.edu/document?repid=rep1&…)
* Conversation Thumbnails for Large-Scale Discussions (citeseerx.ist.psu.edu/document?repid=rep1&…)
* Baby Names and Social Data Analysis (bewitched.com/namevoyager.html)
* History Flow (bewitched.com/historyflow.html) (paper (ifs.tuwien.ac.at/~silvia/wien/vu-infovis/artic…))
* At Harvard and Google / PAIR
* Tools for Model Understanding: Facets (pair-code.github.io/facets), SmoothGrad (pair-code.github.io/saliency/#home), Attacking discrimination with smarter ML (research.google.com/bigpicture/attacking-discr…)
* TensorFlow.js (tensorflow.org/js)
* Visualizing translation (arxiv.org/abs/1611.04558)
* TCAV (arxiv.org/abs/1711.11279)
* Other ML papers:
* The System Model and the User Model: Exploring AI Dashboard Design (arxiv.org/abs/2305.02469) (recent speculative essay)
* Emergent World Representations: Exploring a Sequence Model Trained on a Synthetic Task (arxiv.org/abs/2210.13382)
* Visualizing and Measuring the Geometry of BERT (proceedings.neurips.cc/paper_files/paper/2019/…)
* Artwork
* The Shape of Song (bewitched.com/song.html)

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