Dem
51
GOP
49
When social media appeared on the scene around 2010, many people thought: "This is fun." Now many people are thinking: "This is the end of democracy." The problem of (intentional) disinformation on social media and elsewhere has become enormous. Foreign and domestic actors are pumping out lies day and night and spreading them on social media. Many people can't tell the truth from lies anymore and some don't even think there is any difference. It is like, "everybody has their own point of view." The very concept of objective reality is being questioned.
Some of the disinformation is in the form of text. Chatbots can churn out articles on just about anything. All they need is a prompt. If the prompt is: "Write an article explaining why the FBI is persecuting Donald Trump," these programs can produce a pretty convincing item. It will be full of lies and incorrect statements of "fact," but seem plausible. Many people won't be able to tell the difference between truth and fiction.
It is widely expected that lies and disinformation will flood the zone during the 2024 campaigns. Candidates will be accused of committing crimes and sins that they didn't commit. This was always possible, even without AI, but now entire page-long items full of references to nonexistent "facts" are very easy to produce. Unless the reader digs carefully, they could be convinced.
A number of Democratic secretaries of state, including Steve Hobbs in Washington and Jocelyn Benson in Michigan, are very worried about the problem. They want the tech companies to put much more effort into labeling AI-generated material as such. However, the companies are not at all keen about this. The secretaries know this and are pushing for laws that would make it a crime to publish AI-generated material without labeling it as such. So far, there has been no action on this front. It would probably pass constitutional muster since products made in China must be labeled "Made in China," even if the maker would prefer not to mention that. "Made by a computer" isn't so different from "Made in China." One thing the secretaries are doing is trying to round up "truth tellers" who will look for false (AI-generated) information and help educate voters about it. It will be tough, but at least some of them are trying.
One step beyond false text is false images. AI has now gotten so good that programs can take a text prompt like: "Make a picture of a white girl on a black horse in a field of daisies" and it can produce a lifelike "photo" of that. The image is not scarfed from the Internet, but actually generated by the program. And the images are amazingly good. Suppose you see a photo of some male politician canoodling with a woman not his wife and the source claims a paparazzo took it and the site bought it from him. It looks genuine. Would you believe it? What about a photo that shows a politician somewhere the politican denies having been, like the Capitol on Jan. 6, 2021? A picture is worth 1,000 words and it is hard to unsee such pictures.
You may be thinking: "I can easily tell an AI-generated image from a real photo, so what's the big deal?" Well, we have prepared a little test for you to see if you really can. Go take this quiz we've put together, which shows 20 images and asks you to identify each one as an actual photo or an AI-generated fake. It is not as easy as you might think. Try it and you'll see why.
We will aggregate the results and collect statistics about how many people fell for each fake and how many people didn't recognize an actual photo when they saw one. It could be illuminating. We will publish the results and explain a bit more in a few days. Please take the quiz by Wednesday at 9:00 p.m. PT to be included in the statistics. We will also discuss the process a bit more then as well. Thanks for participating. (V)