43 subscribers
Go offline with the Player FM app!
Podcasts Worth a Listen
SPONSORED


1 You Are Your Longest Relationship: Artist DaQuane Cherry on Psoriasis, Art, and Self-Care 32:12
[Databite No. 161] Red Teaming Generative AI Harm
Manage episode 469479918 series 2615683
What exactly is generative AI (genAI) red-teaming? What strategies and standards should guide its implementation? And how can it protect the public interest? In this conversation, Lama Ahmad, Camille François, Tarleton Gillespie, Briana Vecchione, and Borhane Blili-Hamelin examined red-teaming’s place in the evolving landscape of genAI evaluation and governance.
Our discussion drew on a new report by Data & Society (D&S) and AI Risk and Vulnerability Alliance (ARVA), a nonprofit that aims to empower communities to recognize, diagnose, and manage harmful flaws in AI. The report, Red-Teaming in the Public Interest, investigates how red-teaming methods are being adapted to confront uncertainty about flaws in systems and to encourage public engagement with the evaluation and oversight of genAI systems. Red-teaming offers a flexible approach to uncovering a wide range of problems with genAI models. It also offers new opportunities for incorporating diverse communities into AI governance practices.
Ultimately, we hope this report and discussion present a vision of red-teaming as an area of public interest sociotechnical experimentation.
Download the report and learn more about the speakers and references at datasociety.net.
--
00:00 Opening
00:12 Welcome and Framing
04:48 Panel Introductions
09:34 Discussion Overview
10:23 Lama Ahmad on The Value of Human Red-Teaming
17:37 Tarleton Gillespie on Labor and Content Moderation Antecedents
25:03 Briana Vecchione on Participation & Accountability
28:25 Camille François on Global Policy and Open-source Infrastructure
35:09 Questions and Answers
56:39 Final Takeaways
130 episodes
Manage episode 469479918 series 2615683
What exactly is generative AI (genAI) red-teaming? What strategies and standards should guide its implementation? And how can it protect the public interest? In this conversation, Lama Ahmad, Camille François, Tarleton Gillespie, Briana Vecchione, and Borhane Blili-Hamelin examined red-teaming’s place in the evolving landscape of genAI evaluation and governance.
Our discussion drew on a new report by Data & Society (D&S) and AI Risk and Vulnerability Alliance (ARVA), a nonprofit that aims to empower communities to recognize, diagnose, and manage harmful flaws in AI. The report, Red-Teaming in the Public Interest, investigates how red-teaming methods are being adapted to confront uncertainty about flaws in systems and to encourage public engagement with the evaluation and oversight of genAI systems. Red-teaming offers a flexible approach to uncovering a wide range of problems with genAI models. It also offers new opportunities for incorporating diverse communities into AI governance practices.
Ultimately, we hope this report and discussion present a vision of red-teaming as an area of public interest sociotechnical experimentation.
Download the report and learn more about the speakers and references at datasociety.net.
--
00:00 Opening
00:12 Welcome and Framing
04:48 Panel Introductions
09:34 Discussion Overview
10:23 Lama Ahmad on The Value of Human Red-Teaming
17:37 Tarleton Gillespie on Labor and Content Moderation Antecedents
25:03 Briana Vecchione on Participation & Accountability
28:25 Camille François on Global Policy and Open-source Infrastructure
35:09 Questions and Answers
56:39 Final Takeaways
130 episodes
All episodes
×
1 Challenging AI Hype and Tech Industry Power | Book Talk 1:00:10

1 What is Work Worth? Exploring What Generative AI Means for Workers’ Lives and Labor | Keynote Event 1:00:18

1 [Live] The Cloud is Dead: Living with Legacies of Resource Extraction 59:36

1 Connective (t)Issues: Stories of Digitality, Infrastructures, and Resistance | Public Panel 1:02:06

1 [Databite No. 161] Red Teaming Generative AI Harm 1:00:09

1 Living in the Shadow of AI and Data (Code Dependent by Madhumita Murgia) | Network Book Forum 1:02:08

1 Data & Society at 10: Foreseeable Futures 1:28:51

1 [Databite 160] Black Maternal Health is in Crisis. Can Technology Help? 58:43

1 [Podcast] The Formalization of Social Precarities 1:21:13

1 [Databite 159] Doing the Work: Therapeutic Labor, Teletherapy, and the Platformization of Mental Health Care 1:00:52

1 [Databite 158] Adaptation | Generative AI's Labor Impacts 59:19

1 What's Trust Got To Do With It? | 'Trust Issues' Workshop Public Panel 1:03:05

1 Data In/Visibility (Queer Data Studies) | Network Book Forum 1:00:58

1 [Databite No. 157] Recognition | Generative AI's Labor Impacts 1:06:13

1 [Databite No. 156] Hierarchy | Generative AI's Labor Impacts 1:00:25

1 Caring for Digital Remains | Tamara Kneese and Tonia Sutherland | Network Book Forum 59:56


1 [Databite 155] Democratizing AI: Principles for Meaningful Public Participation 1:00:18

1 Network Book Forum | Disrupting DC: The Rise of Uber and the Fall of the City | Katie Wells and Kafui Attoh 1:00:19

1 Fellows Capstone Conversation: "Make a Way" | Lindsey Cameron with Sareeta Amrute 31:30

1 Fellows Capstone Conversation: "What Guides Us" | Christina Harrington with Sareeta Amrute 29:02

1 [Databite 154] The Trauma of Caste in Tech: In Conversation with Thenmozhi Soundararajan 1:01:32

1 [Databite 153] Essentially Unprotected: Health Data and Surveillance of Essential Workers during the COVID-19 Pandemic 54:13

1 Databite No. 152 Cuidado Digital—Reproductive Rights, Abortion, and Digital Networks of Care in Latin America 1:00:59

1 Databite No. 152 Cuidado Digital—Derechos Reproductivos, Aborto y Redes Digitales de Cuidado en América Latina 1:01:03

1 Databite No. 151 Power and Retail at the Digital Doorstep 1:01:05

1 Databite No. 150 AI in/from the Majority World – Unscripted Conversation 48:12
Welcome to Player FM!
Player FM is scanning the web for high-quality podcasts for you to enjoy right now. It's the best podcast app and works on Android, iPhone, and the web. Signup to sync subscriptions across devices.