There's an interesting interview with Sam Altman here where he acknowledges the model necessarily needs to understand and define off-limit topics in order to be

There's an interesting interview with Sam Altman here where he acknowledges the model necessarily needs to understand and define off-limit topics in order to be told NOT to engage in those topics. https://www.youtube.com/watch?v=WHoWGNQRXb0 https://www.youtube.com/watch?v=WHoWGNQRXb0