An important and insightful study, but I’d caution against thinking that building pro-social aspects in language models is a damaging or useless endeavor. Just

An important and insightful study, but I’d caution against thinking that building pro-social aspects in language models is a damaging or useless endeavor. Just speaking from experience, people who give good advice or commentary can balance between being blunt and soft, like parents or advisors or mentors. Maybe language models need to learn about the concept of tough love.