OxML reposted this
Next in the MLx Generative AI track, we had the privilege of hearing from Asma Ghandeharioun, Senior Research Scientist at Google DeepMind, on "Human and AI Alignment." 🧠 Asma delved into the complexities of AI interpretability, discussing the challenges of specification gaming, goal misgeneralization, and how interpretability can be a powerful tool in aligning AI behaviour with human intentions. She introduced innovative frameworks like Patchscopes, offering new ways to inspect and understand hidden representations in language models. Thank you for your insightful talk! 🔍 🙌 Stay tuned for more updates and highlights from #OxML2024 #OxML24 #OxML #AI4GG #Globalgoalsai #SDGs #RepresentationLearning #GenAI #AIResearch #elandiai #EdTech #GenerativeAI #UserExperience #DigitalInnovation #EducationTransformation #aiconference #MLxGenerativeAI #AI #LSE #healthtech #LLMs #AIproducts Mona Alinejad, D.Phil. (Oxon), Reza Khorshidi, D.Phil. (Oxon), Yali Du, Jane Street, CIFAR
