Credit: The Atlantic
In our world today, we often hear of artificial intelligence aiding in the business world or various fields in STEM. Deep learning AI algorithms have aided in increasing accuracy of breast cancer diagnosis. Machine learning algorithms employed by businesses are used to decide whether a certain cereal brand should be sold to children or the elderly. But what about AI in fine arts?
In January of 2020, a humanoid robot named Alter 3 conducted a professional orchestra for a live concert in the United Arab Emirates. The robot—engineered by Hiroshi Ishiguro—was able to control both the dynamics and the tempo of the orchestra. Even more astonishing, Alter 3 was able to sing while conducting the orchestra! Imagine sitting front row to this concert while a robot waves it arms fluidly in the air with intention. Watch the video below to relive the moment!
Credit: Keiichuro Shibuya
So, you may be wondering, how can such an invention—something that seems too fictional to be true—even be created? The only way modern AI can accomplish such tasks is through the use of an artificial neural network (ANN).
Credit: ResearchGate
All ANNs have the same basic characteristics. There will always be an input and output layer, with various numbers of hidden layers within the neural network. In the case of our music-conducting humanoid Alter 3, the input layer would take in the melody, lyrics, and the number of bars within a particular musical piece.
Now, the real magic happens. Whatever data that is taken as input by an ANN, whether it's text, audio, or symbols, becomes translated into numbers. Next, it travels through the neural networks through transmitting the converted numeric values from neurons (blue dots) of one layer to neurons in the next layer. These numeric values are constantly changed along the lines they travel (i.e., the black lines in the diagram above). They change by being multiplied by a "weight," which signifies how important a particular neuron is to a neuron in the next layer. In the image below, weights are represented by the black lines labeled "w."
Credit: ResearchGate
After traveling through a series of these hidden layers, the data reaches the output layer, which, as the name suggests, gives the outputted result of a neural network. For our friend Alter 3, this means that the results of the output layer are the abilities to keep proper time of a piece, wave its arms in a motion that can be interpreted by musicians, and generate vocal accompaniment alongside the orchestra.
People say AI will be able to accomplish every task imaginable in the future. However, with fascinating inventions like Alter 3, it has become quite clear that the future is now.
If you enjoyed learning about Alter 3 and the unbelievable accomplishments of AI in our modern-day world, make sure to subscribe to the Tech-Kno blog below for free to receive new posts straight to your email!