The only real limitations on making 'machines which think' are our own limitations in not knowing exactly what 'thinking' consists of.
