AI Roundup: Hurricane Milton Edition

AI Just Seems to Add to the Confusion and Fear Surrounding Natural Disasters

A hurricane in the atlantic ocean that has obvious AI effects. (Leonardo) Aside from the absolutely gargantuan size of this hurricane, it looks like a pretty legit satellite image. Scary.

Something which hasn’t made it onto this blog yet is that I enjoy a good weather event. Blizzards, hurricanes, etc., I find them all very interesting. My favorite channel growing up was the Weather Channel, and one day I keep telling myself I’ll be in the same place as Jim Cantore (though that usually means you don’t want to be there) and will be able to meet him in person (but a Cameo may suffice). Now, I live in a place not really known for its extreme weather, so it has to be a big event if it makes it onto my radar (haha, bad weather joke).

However, with hurricane season heating up the two recent storms of Helene and Milton have made it to headlines here in Israel (also because Israel has an obsession with American news, but that is for another time), I took a few minutes to read up on what was going on. Then I went to Twitter/X (always a mistake) to see the hashtags:

While I was doing this I noticed a bunch of things:

• Lots of imagery from places that were definitely not Florida’s west coast

• Images from Florida’s west coast that were then doctored in AI to make them seem more scary

  • The most popular image was a tornado imposed behind a strip mall. This one is just good enough for someone not familiar with some telltale AI signs to think it would be real. The biggest giveaway here is how the tornado remains incredibly stationary throughout the short video, and that there is no motion from the camera given the wind and surroundings

• People retweeting these doctored images

• The trending hashtag at the time was “MiltonHurricane” which is just not good English (wouldn’t it be possible in high-profile events to autocorrect or reindex in hashtag)? Or also to suppress these altogether given that if the hashtag is off, then the content is more likely to be “off” as well?

As AI increases in its ability to be indiscernible from actual footage, this could lead to some very scary outcomes. I know I am not the first to say these things, but this was the first time really came to light for me, and where the potential for harmful and even mortal outcomes could arise. I have no good thoughts about how to solve this but it was just something I found interesting and another intersection between these two interests of mine that I never really anticipated.