Отправить по sms: Separating the “chirp” from the “chat”: self-supervised visual grounding of sound and language