What is creativity? If you are looking in the dictionary, it is using the imagination or original ideas to create something. Of course, this has always been the preserve of humans, beings who can dream big and visualize concepts. Lately, however, researchers have argued that creativity is not only characteristic of humans, but also of artificial intelligence. AIs might not dream like we do, but in specific contexts, with key information at their fingertips, they can write, imagine, invent games, or beat others. These are some of the most extraordinary examples of creative AI.
For an AI, creativity is not exactly the power of the imagination as much as the power of creation. Artificial intelligence uses past data to translate a certain event or environment, learning just enough to generate new things. Most of the time, convolutional neural networks receive immense amounts of data and must train using it as a starting point. Algorithms need to see the patterns in the input information and then generate new examples that might be plausible in a given context.
The answer is definitely Yes, if you accept the definition of creativity proposed by science. There are dozens of examples of this, some more successful than others, starting with an AI that creates poems or stories to neural networks that can name names or bring old photos to life. .
AI technology for photos
Take, for example, this “Deep Nostalgia” AI that turns old photos into video animations. Imagine a reverse iPhone Live Photos where instead of choosing the best frame in a short video, the program takes a vintage portrait photo and sets it in motion.
Another way to bring old images to life is with colorization. So a team has trained an AI to fill in the blanks in a way that allows you to actually see the true colors of Abraham Lincoln, Albert Einstein, Frida Kahlo or Bertrand Russell. The results are incredible!
GAN’s next effort is truly mind blowing! NVIDIA has succeeded in producing new individuals, new human faces from photos of real people. Just take a look at them and tell me if you can tell who is computer generated and who is a real person!
And the AI didn’t stop there. An AI fueled by images of more than 1,000 classical sculptures succeeded in producing Dio, a unique sculpture. Ironically, Dio was built from the remnants of the computer used to generate it.
Top AI Authors
I am not joking. After trying their hand at retouching, coloring, and even creating portraits from scratch, some AI programs were updated. What? Lines, poems, love songs and even horrors.
In this case, getting it right was not the main goal. In fact, the teams that trained the neural networks were probably hoping to laugh at best.
What did they get?
Original ways of flirting, for example. Then, two-line poems written after an in-depth study of over 20 million words from 19th century poetry. Last but not least, a love song that no one should have to listen to – Depressed, tender love. It’s cranky at best.
In contrast is this love poem from a machine formed by researchers at Microsoft and professors at Kyoto University. After practicing over thousands of images juxtaposed with descriptions and poems written by humans on each image, the AI wrote a decent piece that could pass for avant-garde.
By far the most popular AI author, however, is Shelley. This time it was MIT that gave an AI the power to tell stories … and stories she wrote, from random snippets based on what she learned, to contributions on a given text. It all came to a head when Shelley broke the fourth wall and invited users on Twitter to help her write better and more.
Writing horror stories can seem like a fun, sometimes easy task. But defend your intentions with humans? It is not for the faint of heart. Fortunately, GPT-3, OpenAI’s powerful new language generator, lacks heart and was able to appeal to humanity in a deeply moving essay.
AI music generators
Going from poetry to music is child’s play. So the researchers got better and gave AI the task of composing lyrics and even entire music albums.
One of them generated some spooky music in time for Halloween. This AI-generated playlist’s eerie resemblance to horror movie soundtracks has an explanation. MIT has trained the neural network on thousands of songs from cult horror movies to produce a new sound. Scary or just unnatural? Listen here and let me know!
Another AI has been trained to create a new Swift song. To handle this, the neural network was trained on a massive amount of Tay talk. Unfortunately, his creation could not pass for a song by TSwift.
Taylor Swift wasn’t the only singer AI tried to replace on stage. Eminem and Kanye came next, although in their case it was more of a deepfake situation. The two artists changed their path and started rapping about female power. Discover their collaboration here!
Finally, this AI went beyond its musical skills. It helped an artist compose and produce an entire music album. No need for a crew!
AI in games: what can it do?
Have you heard of the God of Go? You must have. AlphaGo is the strongest Go player in history and surprise, surprise, it’s not human. DeepMind, the Alphabet subsidiary in charge of the development of artificial intelligence, is the creator of the most powerful AI player in Go. Their interpretation Zero has proven that it does not need to observe humans before challenge a Go player. In fact, he can play against his predecessor – already world champion – and beat them! Want to know more about this extraordinary program? Read about it here!
Beating humans at their own game is satisfying in itself, but bringing in a whole new game… well, that’s impressive. A neural network trained on data from over 400 existing sports has created Speedgate and its logo!
For those who prefer more static hobbies, a knitting AI might come in handy. InverseKnit has shown that he can do a more than decent job of fingerless gloves from a very specific type of acrylic yarn, but it would be easy to train him on more materials. Ultimately, the researchers would like to make InverseKnit available to the public.
AI professions: weird or awesome?
Now if you think knitter is the strangest job an AI can have, try again. A machine learning program simulated the voice and facial features of a famous Chinese TV presenter, arguing for a non-stop TV host.
In another corner of the world, a publicity contest in Belgium had an AI judge on its panel to choose the winning campaign. Surprisingly, AI made the same choice as human judges, proving its worth.
Finally, an AI took charge of naming the adoptable kittens. Sure, some of his suggestions were downright terrifying, like Warning Signs, Bones of The Master, and Kill All Humans, but the AI managed to find some ingenious ones.
This shows that artificial intelligence doesn’t have to be inventive to be creative. For AI, the sky is the limit as long as humans fill that void.