Работайте офлайн с приложением Player FM !
6 - Debate and Imitative Generalization with Beth Barnes
Manage episode 289497222 series 2844728
One proposal to train AIs that can be useful is to have ML models debate each other about the answer to a human-provided question, where the human judges which side has won. In this episode, I talk with Beth Barnes about her thoughts on the pros and cons of this strategy, what she learned from seeing how humans behaved in debate protocols, and how a technique called imitative generalization can augment debate. Those who are already quite familiar with the basic proposal might want to skip past the explanation of debate to 13:00, "what problems does it solve and does it not solve".
Link to Beth's posts on the Alignment Forum: alignmentforum.org/users/beth-barnes
Link to the transcript: axrp.net/episode/2021/04/08/episode-6-debate-beth-barnes.html
32 эпизодов
Manage episode 289497222 series 2844728
One proposal to train AIs that can be useful is to have ML models debate each other about the answer to a human-provided question, where the human judges which side has won. In this episode, I talk with Beth Barnes about her thoughts on the pros and cons of this strategy, what she learned from seeing how humans behaved in debate protocols, and how a technique called imitative generalization can augment debate. Those who are already quite familiar with the basic proposal might want to skip past the explanation of debate to 13:00, "what problems does it solve and does it not solve".
Link to Beth's posts on the Alignment Forum: alignmentforum.org/users/beth-barnes
Link to the transcript: axrp.net/episode/2021/04/08/episode-6-debate-beth-barnes.html
32 эпизодов
Все серии
×Добро пожаловать в Player FM!
Player FM сканирует Интернет в поисках высококачественных подкастов, чтобы вы могли наслаждаться ими прямо сейчас. Это лучшее приложение для подкастов, которое работает на Android, iPhone и веб-странице. Зарегистрируйтесь, чтобы синхронизировать подписки на разных устройствах.