Funniness is subjective. Some people view some things as funny, others don't. Though I do think that it should be used to judge whether the meme is NSFW or not.
What I do think it should be made is an algorithm to push memes specifically to you, according to your interests. It would be based on previously viewed memes. Memes that talk about topics that you frequently upvote and comment on should be more likely to be the base of the algorithm than memes that talk about topics that you looked for a few seconds and never again.
Speaking of that, to better optimize the proposed algorithm, views should not be counted when you simply look at the meme in the homepage. It should be counted only when you click on it. When you see the image but don't click on it, I would call that an 'impression', which should be a separate metric in my view. The reasoning for this is that the algorithm would be less likely to recommend memes that you are not interested in, as you did not engage with the meme. Imgflip should also measure how much you looked at the meme before scrolling. It can be done by unloading the meme when you scroll past it. That way, people who don't engage with the meme but clearly still liked it would still be recommended similar topics.
Also, use the algorithm for both search systems: meme search (located at the top) and template search (located in the Meme Generator). Doing so would vastly improve the search systems, and also cut down template abuse and speed up the moderation process even more for normal memes.
