The Shadows That Follow AI Generative Models
summary
Lists recent tools like Stable Diffusion, ChatGPT, LLaMA, MidJourney and GitHub Copilot and shows concrete harms: majority of some model answers are wrong (research finds ~62% incorrect Java API replies), YouTube transcripts suffer prompt injection that inserts content, image models enable non‑consensual porn and realistic deepfakes, and image retouching can change ethnicity. Covers copyright and plagiarism risks, a Twitch comedy bot that provokes bans, and governance moves such as the EU AI Act and the US AI Bill of Rights.
AI-written summary · about
chapters (16)
- 0:00Introduction and speaker intro
- 1:08About the speaker and OpenSSF
- 2:13Content warning and AI generative models
- 3:19What are AI generative models
- 5:06Examples of generative AI outputs
- 8:36Correctness concerns in AI models
- 13:40Prompt injection risks
- 15:46Ethical issues in AI competitions
- 17:37AI-generated content and responsibility
- 19:45AI bias in image enhancement
- 21:09Gender bias in translation tools
- 23:57Serious societal impacts of AI misuse
- 26:26Non-consensual deepfake imagery
- 28:34Government regulations on AI
- 30:00Careers in AI ethics
- 31:26Closing remarks and Q&A
related talks
55mAI Showdown: The Good, The Glitchy, and The Risky
45mOverview of Risks, Vulnerabilities and Paths to Safer Use of Generative AI
42mIs your AI plotting to kill you? Why AI and Machine Learning is dangerous
36mDisinform your Surroundings: AI and disinformation campaigns
44mAI Foundations: Mitigate Risks and Boost SOC Efficiency
28mRound Table | Generative AI: Threat or Ally?
13mNavigating Generative AI and Its Risks
48mAI, truth, and the new information environment [Keynote]