Ali Hatamizadeh

“Pushes back on alignment research claims, noting that ways to teach models rules and human preferences already exist.”