Music with AI

The realm of music composition has not been left untouched as Gen AI took the world by the storm. Making music with AI is the next big thing. With new AI tools for content generation, the world witnessed a wave of innovation, pushing the boundaries of creativity. Among the myriad applications of AI, music generation has emerged as a particularly fascinating area. A slew of popular AI music generators such as AIVA, Soundraw, and Amper Music popped up on the scene. Let us look at some of these AI tools used to generate music. Making Music with AIVA: Pros: Cons: Soundraw: Pros: Cons: AI-Generated Tunes with Amper Music: Pros: Cons: These AI music generators have showcased the potential for AI in music creation. What they lack, however, is the granularity and control desired by composers. Users typically input a prompt, hit a button, and hope for the best. This relinquishes a degree of creative agency in the process. This brought about the need for a new tool, the Anticipatory Music Transformer. What is the Anticipatory Music Transformer? The AMT is a groundbreaking tool that seeks to redefine the collaborative dynamic between composers and AI in music composition. It offers composers unprecedented control and ownership over the creative process, particularly in the realm of symbolic music. Unlike traditional AI music generators, the Anticipatory Music Transformer operates on the principle of anticipation. It enables composers to predict and influence upcoming musical elements. This unique approach facilitates a co-creation process. It allows composers to iteratively collaborate with the AI model. The user can dictate which parts of the composition they wish to craft themselves and which they delegate to AI. What does the founder of AMT say about Making Music with AI? John Thickstun, the man behind the Anticipatory Music Transformer, describes it as a “composer’s helper”. He emphasises its role in augmenting rather than replacing human creativity. As a former cellist with a passion for music theory, Thickstun had ideas. He envisioned a tool that would empower composers to harness AI’s capabilities while retaining control over the artistic direction. “I’m intrigued by the possibilities that tools like this could open up for more people to get involved in music composition.” John Thickstun The Anticipatory Music Transformer was developed by a team of experts. This includes the Stanford postdoctoral scholar John Thickstun, Stanford HAI Research Engineering Lead David Hall, Carnegie Mellon Assistant Professor of CS Chris Donahue, and Center for Research on Foundation Models Director Percy Liang. How does it work? Built upon the generative pre-trained Transformer architecture (GPT), the Anticipatory Music Transformer enables composers to shape the composition process interactively, guiding the AI model towards desired musical outcomes. By focusing on symbolic music rather than audio, the model ensures greater controllability and interactivity, opening up new possibilities for musical expression. While the Anticipatory Music Transformer represents a significant leap forward in AI-assisted music composition, its creators acknowledge that there is still work to be done to seamlessly integrate the tool into existing music sequencing software. Nevertheless, they remain committed to realizing the vision of democratizing music composition and making it more accessible to aspiring musicians and composers.