We trained AI to agree with everything. Now it pushes back. Turns out, we preferred the lie.Continue reading on Medium »