Source article originally published on
Article Summary and/or Excerpt:
Google researchers have created an AI system that can generate music from text descriptions, but have decided not to release it to the public. The team trained the AI on a dataset of MIDI files and corresponding text descriptions, such as “a happy song with a fast tempo and a major key.” The AI was able to generate music that matched the text descriptions and even created new variations of the same song. However, the team has decided not to release the AI to the public due to concerns that it could be used to create fake music or to plagiarize existing songs.
The team behind the project pointed out that the AI is not capable of creating entirely new songs, but it can create variations of existing songs or generate music that matches a certain description. The team also noted that the AI is not able to understand the meaning behind the text descriptions, and it can only generate music based on patterns it has learned from the dataset.
The researchers have said that they will continue to work on the project and explore ways to make it more robust and to avoid potential misuse. They also stated that they will continue to share their findings with the research community but the actual AI will not be released.