The Continual Learning Problem