This was a great talk! I was really encouraged to hear all of the opportunities for creating new adversarial examples with current LLM systems, because it highlighted where I think the field has to go in terms of developing safe systems (and where the new "AI-related" jobs are going to be created). Michael I Jordan said a similar thing when talking about how MachineLearning/AI is moving from an ad-hoc research field into an engineering discipline in a similar way that harnessing electricity in the lab turned into Electrical Engineering. Early electrical systems were dangerous and not fully understood, so we developed rigorous methods of building and installing electrical systems so folks could turn their lights on without worrying about burning down their house. We are now at that point with large Deep-Learning systems, should we continue to deploy them in production.
![The moment we stopped understanding AI [AlexNet]](http://i.ytimg.com/vi/UZDiGooFs54/mqdefault.jpg)
![⭐️( ลอยกระทง 2024 ) Ver. แดนซ์ Remix BY [ ดีเจกิต รีมิกซ์ ]](http://i.ytimg.com/vi/rMWOLhgXUHc/mqdefault.jpg)
This was a great talk! I was really encouraged to hear all of the opportunities for creating new adversarial examples with current LLM systems, because it highlighted where I think the field has to go in terms of developing safe systems (and where the new "AI-related" jobs are going to be created).
Michael I Jordan said a similar thing when talking about how MachineLearning/AI is moving from an ad-hoc research field into an engineering discipline in a similar way that harnessing electricity in the lab turned into Electrical Engineering. Early electrical systems were dangerous and not fully understood, so we developed rigorous methods of building and installing electrical systems so folks could turn their lights on without worrying about burning down their house. We are now at that point with large Deep-Learning systems, should we continue to deploy them in production.