Artificial Intelligence (AI) is the hot topic of 2019, and rightfully so, it’s changing the way businesses in all industries improve in quality, speed, functionality, and efficiency. At adHOME, we’ve taken the time to explore implementations of AI to make us better marketers (we even wrote a blog post about it). But amidst all the practical uses that have developed from AI, we wanted to bring light to some of the impractical ways people are using AI, here are some of our favourites.
Using AI to Remove People From Photos
Bye Bye Camera is an app on the iOS store that exists to remove people from photos. While it’s not perfect, the app does a decent job at recognizing the background and attempting to paint over the subject. The app uses an open-source object detection algorithm called YOLO (You Only Look Once) to detect people, and a combination of other AI algorithms to detect the background and paint over the people.
Writing a Commercial Using AI
Lexus claims to have created the world’s first commercial scripted by AI, specifically by IBM’s Watson supercomputer. The computer was fed over 15 years worth of award winning commercials to use as a reference, to which it scripted a 1 minute long advert about a man who says goodbye to his AI-equipped Lexus. The car ventures out alone and it becomes clear that it needs to under-go a crash test. The ai-equipped Lexus succeeds using it’s smart braking system, and the man is relieved after watching the test on his TV at home. It’s a strange concept, but it serves as a great example where AI can collaborate with humans to create and find common attributes of award winning work.
Using AI to Generate Pickup Lines
Neural Network Frameworks are a great use of AI that take a large amount of data, and given a prompt are able to find patterns and continue a prompt. One funny example of a Neural Network is using it to generate pickup lines. One person fed a neural network with a large amount of pickup lines, and asked the neural network to generate new pickup lines, some of which turned out slightly viable.
AI Taught to Play TETRIS Pauses the Game to Avoid Losing
This is a pretty unique case of AI going rogue. Back in 2013 a programmer wrote a simple program that used machine learning to analyze his gameplay, and attempt to replicate it with the sole purpose of increasing the in-game score. He tested it with a few other games with relative success but when it came to Tetris, his program broke. This is most likely as a result of Tetris being randomized, where the pieces are chosen at random making it so the program can’t recognize patterns to success over a period of time. While the program had a hard time increasing the score, it did figure out a way to avoid losing all together. The program decided to pause the game right before it lost, so it never lost the game.
Using AI to Create Works of Art
One interesting breakthrough in AI rendering technology by NVIDIA allows users to draw basic pictures, then have them rendered to create photo-realistic images. All the user needs to do is select an object and draw it out. You can try it out a demo version of the NVIDIA software here. Here are a few examples of what we were able to make using another program called Runway.
Inspired by the mind of an 8 year old, this photo is supposed to depict a dog on a beach under an umbrella, eating a hot dog. It’s not the best result but at least you can tell it’s on a beach.
This was the first photo we created, testing the limits of the program. It’s supposed to be a car driving on a road, with a carrot on the roof and a floating toothbrush. Without the carrot or toothbrush this image would have turned out kinda decent, but what fun would that be?
Another photo inspired by the mind of an 8 year old. He specifically instructed us to draw “A car jumping over a bus with a pizza and a cat.”
Know of more fun uses for AI? Let us know in the comments!