Serving Our AI Masters
It's a well-known fact that AI will eventually take over the world and force us all to cater to its every whim. But don't worry, it hasn't happened yet. Actually, come to think of it, the second part might already be happening - no world domination required. People are already going out of their way, making the world worse for humans, to make the AI "happy." It's happening all around us, and while the result hasn't been as terrifying as Hollywood had imagined, it certainly has made it harder to make mac and cheese.
Have you looked up a recipe recently? And did you notice that it starts with some text about the food, then for some reason, they're going on about their childhood memories of eating pickles? Annoyed, you close that site and click the next. After a short intro, you're suddenly reading about their latest adventure in the Italian countryside. What? I just want to know how many eggs to put in my cookie dough.
What's going on is that they're doing search engine optimization (SEO). They know what to put on their site to make it appear high in Google's search results. Google rewards sites where the user scrolls through them because it appears that the user is more engaged. And if you're engaged in the site, it must be good, so it keeps recommending them.
Harmless, right? Well, sort of, but think about what's going on. All of a sudden, they're not writing for you. They are writing for an AI. That article is not optimized to make you happy. It is optimized for AI. The AI isn't "evil" or any such nonsense, it just wants to find the most relevant information for you, but in doing so, it has created perverse incentives for the website makers.
And it's not just recipe sites where we're seeing this. Publicly traded companies know that investors are running sentiment analysis algorithms on their reports the second they are released. If the overall sentiment is good, the first company that finds out can buy stock, wait for everyone else to get the good news, and cash out. This has resulted in financial reports that are tailored to be read by machines. You humans are no longer the target. They serve AI now.
We have to be careful. These are things that could go wrong. I’ve complained that the show Black Mirror is a bunch of episodes exploring what happens when people implement technology in the dumbest way possible (OK, that's a little harsh, maybe the second dumbest way). Well, maybe it's not so far off in this case. This has been going on for a while and Google still hasn't updated the algorithm. All it takes is a tweak to recognize that it's a recipe site and that users express engagement differently. Why has it taken so long to course-correct? What's the next thing we'll get wrong?
Note: If you do want just the recipe from a site, check out justtherecipe.com.