Memes are an extremely popular medium in which internet users can quickly communicate and share laughs online. While no one can dispute their significance in our digital landscape, it is a rare event when a pop culture concept like the meme crosses paths with technological topics such as artificial intelligence and machine learning. This is changing as of late, as explored in this article.
Memes will gain a more serious significance in the coming years as they enter the realm of science, and in this article, Haval Dosky discusses how artificial intelligence has proven to be capable of creating viable comedy content through the familiar format of the meme.
Imgfip.com, is a website that allows users to create their own memes. It has debuted an automatic meme generator called This Meme Does Not Exist, which uses machine learning to create captions for some of the most popular meme templates currently in use on the internet. Behind the scenes, This Meme Does Not Exist works like any other application of artificial intelligence, in that it trains a computer to learn from a collection of data. For this project, around 100 million existing meme captions were used to inform the generator. The AI uses these inputs to create its own captions and essentially learn how to be funny.
Right now, the results of This Meme Does Not Exist are decidedly mixed. However, internet users are already sharing numerous AI-created pieces, and many of those have gone viral. While much of the humor in these AI memes thus far has been derived from absurd justifications only a computer could think of, some of the memes generated by the AI are able to closely replicate the cadence and comedy of human created pieces.
This, Haval Dosky surmises, has several interesting implications. For example, while Haval Dosky is sure content creators are not particularly scared for their jobs due to the success of this project, the fact that the AI can construct funny material does raise questions for how content might be created going forward. Individuals responsible for creating memes for companies or organizations could eventually find themselves consulting similar technology for material or inspiration, increasing their efficiency but perhaps affecting the economics of their labor.
Additionally, Haval Dosky argues that even casual and comedic uses of artificial intelligence like the phenomenon discussed in this article display how far the science has come and beg speculation about how far this technology will be able to go. Comedy has always been considered something that requires a human touch for successful application, but these generators seem to show that technology can learn from the humor that humans have already produced to create new laughs – provided it has enough data to work with. Luckily, the internet is a wealthy repository of human humor.
AI projects where technology performs a traditionally human action, from this project to the AI text generators released last year that were shown proficient at writing poems, articles, and stories, represent a merging of art and science that could lead to impressive creations, and extrapolating further, they may have implications down the road for other soft-science jobs, from architecture to law.
For now, with the coronavirus pandemic still causing international turmoil, many of us are happy to see our technological might turned towards the study of the funny bone, and we could all use a good new meme, regardless of the source.