Why do you care about AI Existential Safety?

I am deeply interested in ensuring the robustness and stability of DNN models, as these qualities are essential for any useful and reliable model. Additionally, AI is becoming more integrated into our daily lives and work environments. To prevent potential misuse of AI, it is crucial to prioritize its stability by addressing its origins.

Please give at least one example of your research interests related to AI existential safety:

I am very interested in how to elicit latent “knowledge” of AI models. They could fulfill human intentions and provide accurate answers. However, Do they truly meet human expectations? We need to develop tools to figure out the inner mechanism of AI models.

Efficiently conducting adversarial testings against AI models is very important since it can help us find potential “bugs”.