On the relevance of logic for AI: misunderstandings in social media, and the promise of neuro-symbolic learning