AI Model Vulnerabilities
Seth and Jeremy discuss how AI models can be manipulated through audio and image inputs, showcasing vulnerabilities through examples like a Harry Potter response and malicious link output. They explore the implications of defending against multi-modal attacks on the overall robustness of language models.In this clip
From this podcast
Last Week in AI
#132 - FraudGPT, Apple GPT, unlimited jailbreaks, RT-2, Frontier Model Forum, PhotoGuard
Related Questions