Appearances
Talks, panels and podcast conversations where I did the talking; the courses and lectures I taught; and a short list of reporting by others. Each entry says what kind of recording exists. Where the only trace is a program listing, it says that too. A shorter selection is on the front page.
Talks and conversations
Human questions, responsibility and institutions
-
Careers at the Frontier: Hiring the Future at OpenAI
OpenAI Forum · interview and audience questions
Video and transcript · about 60 minutes
Unconventional career paths, culture, and recruiting during rapid growth, including my own moves through Preparedness and hands-on work.
-
Building the Future of Responsible AI
Women in Big Data and LinkedIn, Responsible AI at Scale · panel with Grace Tang, Hema Raghavan and Sakshi Jain
Event page · no recording
Standards, transparency, bias, safety, and career advice for people entering the field.
-
AI Native Summit, Zetta Venture Partners, at the Computer History Museum · panel with Chloé Bakalar, Shreya Rajpal and Yaron Singer, moderated by Nick Elprin
Video · 40 minutes · program
How independent evaluation, governance and assurance can make deployed AI systems trustworthy.
-
BayLearn 2024, hosted at Apple · with Jascha Sohl-Dickstein and Stuart Russell, moderated by Nilou Salehi
Event page · no recording
A practitioner and research panel on alignment and safety.
-
Information Integrity and AI: Combatting Deepfakes and Disinformation
NVIDIA GTC · panel
Video · about 55 minutes
Policy, resilience, platform responsibility, and the opportunities and risks generative AI creates for information integrity.
-
Practical AI Fairness: Math, Ethics & Politics
International AI Cooperation and Governance Forum, HKUST and Tsinghua · fireside conversation with Chloé Bakalar
Video · 41 minutes · program (PDF)
Why fairness is not a single metric: mathematical tradeoffs, ethical judgment, politics, and what deployment actually looks like.
-
Transforming Together: Building a Blueprint for Change
Partnership on AI Partner Forum, San Francisco · panel, after opening the forum with Rebecca Finlay
Agenda · 45 minutes · no public recording
Multistakeholder approaches to socio-technical problems, consensus, and accountable institutional change.
-
Can We Meet Business Goals and Attain Responsible AI?
In AI We Trust? · podcast interview
Audio · 44 minutes
Ethics by design, organizational diversity, communicating about AI, and why responsibility and business outcomes are compatible.
-
From Optimism to Doubt (and Back) at Facebook
In Reality · podcast interview with Eric Schurenberg and Joan Donovan
Audio
Scaling Facebook AI, confronting platform harms, and building Responsible AI from inside the institution.
-
Opening keynote, Pioneer Centre for AI
University of Copenhagen · keynote at the launch of Denmark's national AI center
Announcement · no recording
Trustworthy, human-centered AI and the need for participatory governance.
-
Finding Solutions: Choice, Control, and Content Policies
Belfer Center, Harvard Kennedy School · in conversation with Karen Hao, with audience questions
Video · 60 minutes · watch on YouTube
Recommendation systems, platform harms, user choice, and the institutional limits of content-policy interventions.
-
Responsible AI: A Perspective from the Trenches
MIT Media Lab Perspectives · fireside with Andrew Lippman
Video · 60 minutes
Lessons and difficult ethical tradeoffs from fifteen years building and deploying AI at Microsoft and Facebook.
-
Building Effective and Socially Responsible Artificial Intelligence
Marconi Society, Decade of Digital Inclusion · panel with Marian Croak, Gillian Hadfield, Stuart Russell and Mark Gorenberg
Event page · program · recording not clearly linked
The pressing social questions created by pervasive AI.
-
Responsible AI Is Primarily a Human Question
HKUST Centre for Artificial Intelligence Research · virtual fireside with Pascale Fung
Video
My technical and personal trajectory as a way into what responsible deployment means for human work, life, risk and dignity.
-
Building Responsible ML Solutions in Social Media
O'Reilly AI Superstream: Responsible AI · conversation with Rumman Chowdhury
Video · O'Reilly subscription may be required
Concrete methods and organizational choices for responsible machine learning in high-scale social products.
-
ICLR 2021 Responsible AI workshop · invited talk with Isabel Kloumann and Chloé Bakalar
Video · 30 minutes · workshop schedule
Fairness as an interdisciplinary process rather than one metric, grounded in tools and cases from production systems.
-
Towards Data Science · podcast interview
Audio and transcript · about 50 minutes
Operationalizing fairness, incompatible definitions, algorithmic discrimination, and institutional governance.
-
Gradient Dissent · podcast interview with Jeremie Harris
Video and audio · 79 minutes · Apple Podcasts
A long-form bridge from production machine learning to the moral and technical conflicts inside algorithmic fairness.
-
Facebook, la justicia algorítmica y la inteligencia artificial
II Congreso Internacional de Inteligencia Artificial, Alicante · opening keynote, in Spanish
Conference archive with video
Algorithmic fairness, content systems, and the opportunities and social risks of deployed AI.
-
Fairness and Ethics in Data Science
ACM-IMS Interdisciplinary Summit on the Foundations of Data Science · panel with Alexandra Chouldechova, Andrew Gelman and Kristian Lum
Program (PDF) · 40 minutes · no recording
Statistical, social and policy perspectives on fairness.
-
Inclusive AI and fairness at F8
Facebook F8, day 2 keynote · segment
Video, embedded in Meta's F8 recap
Inclusive product design, fairness, and tools for testing AI systems across demographic groups.
-
Fairness and Artificial Intelligence
Partnership on AI All Partners Meeting, London · dinner conversation with Annette Zimmermann
Talk listing · no recording
A philosopher and a practitioner on competing notions of fairness and institutional judgment.
Building AI systems at scale
-
A fireside chat on the deployment of AI tools
Marketplace Innovation Workshop · plenary fireside with Nicolas Stier-Moses
Program (PDF) · 45 minutes · no recording
Responsible AI in labor marketplaces and the organizational realities of deploying AI tools.
-
AI and Machine Learning: A Perspective from Facebook
Berkeley Center for Law & Technology, Privacy Law Forum · keynote
Agenda · 30 minutes · no recording
The power, privacy risks and ethical obligations of large-scale AI, delivered the week the Cambridge Analytica story broke.
-
Getting a Start in ML and Applied AI at Facebook
Talking Machines · podcast interview
Audio
AI versus data science, building an applied research career, and how Facebook moved models into products.
-
Designing AI at Scale to Power Everyday Life
Machine Learning @Scale · conference talk, also given as an invited industry talk at AAAI-17
Video · 30 minutes · recap
The organizational and technical systems needed to make AI routine across products rather than the province of specialist teams.
-
Facebook F8 · keynote segment
Video
Searchable and intelligent video, and the shared AI backbone supporting production applications.
-
Facebook Applied Machine Learning Q&A
Quora Writing Session · twenty-six written answers
Article · Chinese translation at InfoQ; the original Quora page is no longer indexed
Production machine learning, research versus engineering, teams, career development, and work habits.
-
Machine Learning and the Facebook Ads Auction
ACM EC and NBER joint session · invited talk
Abstract (PDF) · no recording
Auction prediction, feedback loops, and large-scale machine learning in advertising markets.
-
How Machine Learning Is Central to Facebook
Hamilton Institute, Maynooth · research seminar
Seminar flyer (PDF) · 60 minutes
An early account of how ranking and prediction became core product infrastructure at Facebook.
Earlier technical work
-
Microsoft Research: papel de la investigación en un gigante informático
University of Granada, CITIC inauguration · public lecture, in Spanish
Listing · no recording
How industrial research produces scientific and product impact inside a large technology company.
-
Microsoft Research · short research video
Video · 2 minutes
A very short look at machine learning and the challenge of computer Go. Historical rather than substantive.
-
AdPredictor: Large Scale Bayesian Click-Through Rate Prediction in Bing
NeurIPS 2010 workshop on machine learning in online advertising · invited talk with Thore Graepel, followed by a panel
Slides (PDF) · 30 minutes · schedule
A production Bayesian model for online advertising, then a broader research and industry panel.
-
Microsoft Research · group interview and demo
Video · 20 minutes
How researchers used F# for probabilistic modeling, experimentation, and production-oriented research.
-
Probabilistic Machine Learning in Computational Advertising
NeurIPS 2009 workshop · invited talk
Workshop book (PDF) · no recording
Bayesian prediction and uncertainty in large-scale computational advertising.
-
Sensible Priors from Finite Linear Models
Cambridge Machine Learning Reading Group · with Ed Snelson
Talk listing · no recording
How finite basis-function models can induce interpretable priors for Gaussian-process approximations.
-
Cambridge Machine Learning Reading Group
Talk listing · no recording
Gaussian-process latent-variable models for nonlinear dimensionality reduction.
Teaching
Courses and lectures, which ask for a different skill than a conference talk: making hard ideas legible over several hours, and helping other people get good at them.
-
Transatlantic AI student-project panel
German International School of Silicon Valley · judging and mentoring high-school teams applying AI to real problems
School recap · no recording
-
Stanford ICME, CME 500 seminar · conversation with Esteban Arcaute
Video · 60 minutes · Stanford series page
What operating real AI products teaches that benchmark and laboratory work do not.
-
Gaussian Process Summer School, University of Sheffield · technical lecture
Program · 60 minutes
A foundational lecture on Bayesian model evidence for an advanced machine-learning summer school.
-
Examples of Machine Learning and Data Science at Facebook
Gaussian Process Summer School, University of Sheffield · industry lecture
Program · 60 minutes
Early production examples of how machine learning and data science were being used across Facebook.
-
An Introduction to Probabilistic Machine Learning
Microsoft Research · technical lecture summarizing a sixteen-hour course
Video
Gaussian processes, graphical models, LDA, TrueSkill, and probabilistic reasoning in one substantial teaching session.
-
University of Cambridge · fourteen-lecture course, co-taught with Carl Edward Rasmussen
Course archive: notes, slides and coursework · about 16 hours of lectures
A graduate course spanning supervised and unsupervised probabilistic machine learning.
-
Halo, Kinect, Bing: The Magic of Research in Microsoft Products
Microsoft Research PhD Summer School · lecture
Program · no recording
TrueSkill, geometry, and large-scale machine learning as examples of research crossing into widely used products.
-
Probability and Bayesian Inference
Machine Learning Summer School · course lecture
Video · 90 minutes
An introduction to probability, Bayesian reasoning, and uncertainty for machine-learning students.
-
The EM Algorithm and Mixtures of Gaussians
Machine Learning Summer School · course lecture
Video · 90 minutes
A full derivation and practical explanation of expectation-maximization and Gaussian mixture models.
Profiles and reporting
Pieces written by others, kept short on purpose. The critical ones are here too.
-
“Unprecedented pressure to grow”
CNBC · interview and profile
Article
The organizational demands, hiring bar, and human constraints behind OpenAI's rapid expansion.
-
How Facebook Got Addicted to Spreading Misinformation
MIT Technology Review · investigative profile by Karen Hao, from extensive interviews
Article
A critical account of production incentives, the creation of Responsible AI, and the limits of internal reform.
-
“Los ingenieros no están preparados para resolver cuestiones éticas”
El País · interview, in Spanish
Article · may be metered
Why ethics cannot be delegated to engineers alone, and how an interdisciplinary Responsible AI team was built.
-
“Künstliche Intelligenz ist ein sehr mächtiges Werkzeug”
Deutschlandfunk · radio interview, in German
Audio and transcript · 11 minutes
AI-assisted content moderation, the limits of automation, and the arms race against harmful content.
-
“La inteligencia artificial nos ha ayudado a detectar los contenidos que incitaban al odio”
El Independiente · interview, in Spanish
Article
Career, hate-speech detection, moderation at scale, and the limits of fairness technology.
-
People of ACM: Joaquin Candela
Association for Computing Machinery · interview
Article
Career arc, production machine learning, research-to-product work, and growing attention to ethical AI.
-
Harvard Business Review · interview with Scott Berinato
Article · may be metered
Why “wizards don't scale,” and how platforms, tools and organizational design put production AI in everyone's hands.
-
WIRED · profile from extensive interviews
Article · may be metered
The creation of Applied Machine Learning and the strategy for putting machine learning into every Facebook product.
-
Artificial Intelligence 2017 Trends: interview
Zenith · report interview
Report (PDF)
Facebook's AI roadmap across vision, speech, language, connectivity and VR, framed for business and media audiences.