Trustworthy AI: From Principles to Practices