AI is getting better at generating "deepfake" text on the internet, and people suck at telling it apart from comments written by real humans

An interesting article from Ars Technica (https://arstechnica.com/tech-policy/2021/01/ai-powered-text-from-this-program-could-fool-the-government/) and Wired (https://www.wired.com/story/ai-powered-text-program-could-fool-government/) that may be interesting and relevant to some Imgurians. Just a few highlights below; see the original article from either site for more details.

Max Weiss, a Harvard medical student, used openly available AI tools to generate comments for a government website as part of an experiment. He found the process surprisingly easy to set up, and the comments stayed up for months (despite the fact that he promptly notified the government about what he was doing, and included a tag in each artificial comment to make it easier to moderate).

AI is getting to the point where artificially generated content is actually quite good, potentially indistinguishable, in some contexts, from content generated by actual human beings. And this technology is only getting better at what it does.

The successor to this program, GPT-3, is already available to AI researchers, and further improvements are being made all the time. It is already possible for any random individual with a computer to do what Weiss did, and it will only get easier with each passing year.

"The volunteers did *no better than random guessing.*" I.e. these comments pass a simple Turing Test. Of course, it is easier for short comments generated by AI to be convincing, than for longer passages of text. There is less room in which to make mistakes. However, if your objective is, for example, to swamp forums or comment sections with machine-generated comments to create the false impression that there are lots of people out there sharing a particular point of view, short comments are more than adequate... And as the AI gets better, its capabilities will only improve beyond that.

And of course, if you're good at demagoguery, or have access to an army of paid human commenters, you don't even need clever AI. Its existence still opens up new and scary possibilities, however. This particular genie is already out of the bottle, and with no way of putting it back, the face of the internet may be changing significantly in short order. It's common enough to see half-joking comments about "Russian bots" on Imgur, especially under any post dealing with political topics. But the technology is real. It's already entirely possible for just one person, using freely available software, to create the impression of a groundswell of support for any position they may choose. And of course governments and large corporations have access to more resources, and more complex software and hardware. If you think the internet is being used to manipulate people and spread misinformation now... I'd say this is certainly something to worry about.