We have witnessed a wide scope of artificial intelligence in automation, heavy calculation tasks, speeding up processing powers of a computer or even beating Go Champion Lee Sedol. Now, Google wants to take this to a higher level and wants to see if artificial intelligence can be creative and produce pieces of original art such as music. It is launching a new research project called Magenta which will test the creative capabilities of artificial intelligence.
This was revealed in an announcement made at Moogfest, a four-day music and technology festival, in Durham, North Carolina. Douglas Eck, who is a researcher in the company’s artificial intelligence research project called Google Brain, said that Magenta will focus on whether computers can truly create.
The idea is to see if a machine learning system can learn to create engaging and exciting pieces of music by exposing it to a huge amount of musical training data.
The initiative will be officially launched at the beginning of next month and will be aimed towards developing tools to help other researchers as well as its own team to explore the creative potential of computers.
Magenta will be using the machine learning engine built by Google, TensorFlow to test if artificial intelligence could create original pieces of art, music or videos.
Like TensorFlow which was recently made open source by Google, Magenta will also make its tools available to general public and researchers to enable them to carry out their own similar research.
Initially, Magenta will launch a simple program that would enable researchers to import music data from MIDI music files into TensorFlow and hence train the system to recognize and gain musical knowledge.
The goal is to design a system that can give listeners “musical chills” by regularly generating new pieces of music which they could listen from comforts of their home.
Douglas Eck also gave a glimpse of the kind of work they will be doing under Magenta during the festival. He demonstrated a digital synthesizer program powered by AI which could listen to notes that he played, and then the program played back a melody combining those notes.
However, Eck did admit during a panel discussion that we are still a long way from where AI system could actually create original art without any human intervention and said that AI systems are “very far from long narrative arcs.”
But at the same time, Eck said that we may soon be able to witness a situation where a wearable device detects that a person is stressed by measuring heart rate and an AI system will then create music to sooth that person.
Once done with music, Magenta group also plans to carry out similar tests to see if artificial intelligence can create images and videos.