Memes are mostly funny, but not always for good reasons. This is a common problem of comedic content (e.g., comedy movies, standup comedy, satires, etc.). Meme can be used to spread Information Disorder.

Meme is particularly fast and effective to convey ideas due to the fusion of image and text. Humans conceive images quickly, but each person may interprete them slightly differently. Texts are more precise in describing ideas, so it is used in memes to direct interpretation. Therefore, meme is a rapidly conveyed and precisely crafted medium for ideas.

Working with hateful memes (e.g., data labelling, content filtering) is a dangerous job, like working with radioactive elements. It spreads ideologies into your brain. We need some kind of protection for people working on that. Automation is a way to do it. Computers can be a guard for our minds (of course each person wants to expose themselves to different type of contents).

How might computers understand memes? They need to look at images, recognize atomic visual elements, connect them together and with existing world knowledge. At the same time, it needs to connect with the text, which may also have positional information on the images (e.g., a text placed on a face is actually labelling the person). AI research is active in this problem.