AI and algorithms are everywhere. An algorithm is code which executes itself based on a written line or lines of code. Most algorithms aren’t too complex, but they usually have a specific purpose.
Monitoring algorithms exist in many places such as in social websites to monitor for abusive, foul, or explicit language, or spam; on YouTube such as those which automatically flag videos based on the content in the video, learning algorithms that suggest movies you may like such as on Netflix or Amazon Video; shopping suggestions from Amazon or EBay. These are all smart algorithms which have a specific purpose: monitor and fulfill a function.
People are beginning to learn about algorithms and how they function, algorithms are usually not too dynamic and can be outsmarted easily, since our AI technologies are still quite primitive. For instance, you can outsmart an algorithm which monitors for explicit language by typing “D A M N” or “D4MN”. I’ve seen many smart tricks to outsmart algorithms. One recent one I’ve seen is to write a profile description on into a photo or pasted to a photo, rather than have it in their description. Since the algorithm specifically sorts through text and not images, it can’t pick up anything that is written in the photo. The algorithm which sorts through text can’t detect it because it’s not dynamic enough to pick text out of a photo. I’ve seen people writing forbidden text using pictorial emoticons, which can’t be detected by the algorithm as text. These are a few ways that people can circumvent smart algorithms, either by creating media out of the range and responsibilities of the algorithm, by feeding it input that it does not recognize, or feeding input which lies outside the rules for classification as malicious.
Algorithms are implemented mostly in tedious and boring tasks, such as sorting through loads of text or sorting through hours of videos, which would take human reviewers many hours and still need to be paid. Algorithms can also act as a set of eyes. I mean, you don’t really have to pay someone to monitor stuff when you can throw a simple algorithm at it and have it do the job. Then just have one human supervise its work to make sure what it flags is good. The problem with algorithms is that they don’t have discretion and they do exactly as they are programmed. It’s possible that they flag things which aren’t really needing a flag, or vice versa.
Many YouTubers are having these algorithms flag their videos because of their content, even though it is not explicit per se, but they learn quickly how to stay above the algorithm as a way of survival of their channel. All it takes is one human flag to cause an algorithm to review your video, then if it picks up any key words or language, it can penalize you.
One notable YouTuber that was recently targeted was Psyched Substance, who provides education and harm reduction when it comes to drugs. Because of the nature of the videos, it’s easy to be flagged by any algorithms. He’s had many videos restricted based on the content.
The owner of the Psyched Substance channel has had to remake his videos and add lengthy warnings at the start of his videos detailing his purpose of the channel and videos.
Algorithms sometimes get changed by the programmer to circumvent these bypasses, in order to keep up with standards. The YouTuber then again learns to circumvent the new programming. Not only with YouTubers, but in general, algorithms work this way. Search engines and SEO work similarly.
It reminds me a lot of a predator-prey relationship where the predator gains an advantage, causing selective pressure to survive in the prey. Which causes the prey to grow a greater advantage. Then the predator evolves another advantage, and the prey, ad finitum. Humans and viruses/diseases work similarly. We create a vaccine against the disease, only to have it mutate in order to survive. Algorithms and humans are in this same race. Humans want to do things, and algorithms want to moderate it. Humans learn strategies to circumvent, and algorithms evolve to adapt.
I just wanted to point out that humans have begun a new battle, it’s man vs machine, and this war will only get more intense with time. We will see much news of this in the future as algorithms become very advanced, how humans can circumvent this like a bank robber committing a great heist.