AI 202: Gaslighting AI Creatively
⛽ How to gaslight an AI, and why you should do it
✨ While everyone is hyping about sprinkling AI stardusts everywhere, we'll spend an afternoon imagining how we might make those dusts explosive*. 💥 Expect knowledge sharing, interactive demos, ideation, and Socratic discussions on adversarial attacks and their role in AI creativity.
This event's setting will be more university-course-like and require you to wear 🎓 critical and 🧢 creative hats *. A basic understanding of machine learning is preferred but not required. Designers, artists, creative technologists, and data/ML scientists are particularly welcome.
👥 Max 15 participants. 🌧️ Stay tuned for rescheduling arrangements if the weather goes naughty.
* metaphorically speaking, just in case 😂
What are Adversarial Attacks?
Adversarial attacks are techniques used to trick AI systems into making mistakes (e.g., prompt injection, data poisoning, adversarial noise). They exploit vulnerabilities in the AI's learning process, often with surprisingly simple methods. Besides ethical and safety implications, we could push the boundaries of these vulnerabilities for creative expressions and as design material.
About Mixed Initiative
Mixed Initiative is where creative minds inspire each other by making, doing, and reflecting on AI and authenticity. Our mission is to democratize AI literacy, promote critical discourse, and prototype new meaningful ways of being creative with AI. Learn more at http://www.mixedinitiative.com
Image credits: "Social Media" by Alan Warburton (Better Images of AI / CC BY 4.0)