I’m not sure that research is a good idea

I’ve commented before that one of the reasons I left academic research is that I felt it focused too much on novelty and coolness rather than providing actual value. I don’t think most of that research is actively bad; it’s just often pointless (and thus often a waste of resources). But there is some research that may cross over into being actively bad. I personally find it hard to see the redeeming value in research on creating deepfakes.

A :”deepfake” is the result of using AI techniques (generally deep learning) for image, audio, or video synthesis, allowing the creation of fake content that is increasingly indistinguishable from real content. I’m sure the initial academic research was motivated by the challenge: hey, could we take old movie or TV footage and appear to have the actors voice different content? Maybe Hollywood sponsored some of the work as a way to repurpose old content.

But what exactly is the redeeming value of continued research on making it as easy as possible to create deepfakes? What beneficial cases will outweigh the inevitable use for spreading disinformation and fake news? Think internet trolls are bad now? Wait until they can falsify news footage with a few clicks. How well will social discourse fare when a flood of fake content overwhelms real content, and users are unable to easily tell the difference?

While at this point the technology has advanced enough that motivated amateurs will be able to make continued, if incremental, progress on their own, it seems like a fundamentally bad idea for researches to continue to push the state of the “art”. When your house catches on fire, you should probably reconsider whether you want to keep playing with matches. And you really shouldn’t pour gasoline on the blaze.