At the 2026 Berklee AI Music Summit, Jordan Rudess showed the audience what happens when one of the world’s most innovative keyboardists meets a real-time AI improvisation system built for the stage.
The performance featured JAM_BOT, a collaborative music system developed through a partnership between Jordan Rudess and researchers at the MIT Media Lab. Rather than simply generating background music or playing pre-programmed patterns, JAM_BOT listens, responds and improvises with a human performer in real time. It is designed to take part in a musical conversation.
From MIT Media Lab Research to a Live Musical Partner
JAM_BOT grew out of Jordan Rudess’s collaboration with the MIT Media Lab’s Responsive Environments Group, led by Professor Joseph Paradiso. The project brought together Rudess’s decades of musical experience, his deep interest in expressive music technology and MIT’s research into interactive systems, machine learning and responsive environments.
MIT graduate researchers Lancelot Blanchard and Perry Naseck played key roles in developing the system. Blanchard focused on the AI music system, while Naseck helped create the visual and installation elements that make JAM_BOT understandable to an audience in a live performance setting. Their work was not just about building an AI model that could generate notes. It was about creating a performance partner that could participate in the flow of improvisation.
That goal is what makes JAM_BOT so interesting for musicians. In a live improvisation, timing, phrasing, dynamics, expression and musical intent all matter. A useful AI partner cannot simply create technically correct music. It must understand enough of the performance context to respond in ways that feel coherent, surprising and musically meaningful.
How JAM_BOT Works
JAM_BOT is described as a real-time system for human-AI co-created symbolic music improvisation. In practical terms, that means the system works with structured musical information rather than only finished audio. It listens to musical input, analyzes what is happening, and generates new musical responses that can be played back through instruments and performance systems.
This is where MIDI becomes especially important. MIDI has always provided a way to represent musical performance as data: notes, timing, velocity, controllers, pitch bends and other expressive gestures. For an AI system that needs to understand and respond to live performance, structured musical data is incredibly valuable because it carries information about the musician’s intent.
JAM_BOT uses optimized music language models to accompany live performers on stage. The system is designed to follow the improvisational choices of the musician and generate responses that keep the performance moving forward. In the case of Jordan Rudess, that means responding to a performer whose playing can shift quickly between classical technique, progressive rock intensity, electronic textures and spontaneous experimentation.
Jordan Rudess as the Ideal Collaborator
Jordan Rudess has always been more than a keyboard virtuoso. As a member of Dream Theater, a solo artist, an educator and a music technology developer, he has consistently explored new ways that musicians can interact with instruments, software and audiences. His work with expressive performance apps such as GeoShred and his long history with MIDI-based instruments make him a natural collaborator for a project like JAM_BOT.
For Rudess, the challenge was not simply whether AI could generate impressive musical material. The deeper question was whether an AI system could become a genuine creative partner. Could it listen? Could it respond? Could it surprise the performer in a way that still felt musical? Could it help create a moment on stage that neither the human performer nor the system would have produced alone?
The Berklee AI Music Summit video gives viewers a chance to see that question explored in real time. Rudess does not treat JAM_BOT as a playback device. He interacts with it as part of the performance. The result is a glimpse of a future where AI is not positioned as a replacement for musicians, but as a new kind of musical collaborator.
Why the Berklee AI Music Summit Was the Right Stage
Berklee’s AI Music Summit brought together artists, educators, researchers, technologists, rights experts and companies working at the intersection of music and artificial intelligence. The event focused not only on what AI can do, but also on how musicians, educators and the music industry should shape its use.
That made the summit an ideal setting for JAM_BOT. The project sits directly at the center of many of the most important questions in AI music. How can AI systems support human creativity? How should musical data be represented? How can artists retain agency when working with intelligent tools? What does authorship mean when a live performance includes a responsive machine partner?
By putting JAM_BOT on stage with Jordan Rudess, the Berklee AI Music Summit moved those questions out of the abstract. The audience could hear and see the interaction for themselves.
A New Kind of Human-Machine Performance
One of the most compelling parts of the JAM_BOT project is that it treats performance as a relationship. Traditional music technology often gives musicians tools they can control. JAM_BOT points toward tools that can participate. That difference is significant.
In a conventional setup, a keyboardist might use MIDI to control a synthesizer, trigger a sequence, change a sound or manipulate effects. In a human-AI performance, MIDI and related musical data can become the language of interaction between the performer and the system. The musician plays. The system interprets. The system responds. The musician reacts. A feedback loop begins.
That feedback loop is what brings JAM_BOT alive. It is not only the AI model that matters. It is the complete performance system: the musician, the data, the model, the instruments, the sound, the timing and the visual environment that helps the audience understand that something interactive is happening.
Why This Matters to the MIDI Community
For the MIDI community, JAM_BOT is an important example of how the future of AI music may depend on structured, expressive and interoperable musical data. AI systems can generate audio, but live performance requires more than audio output. It requires control, timing, nuance, responsiveness and communication between devices and software.
MIDI has spent more than four decades connecting musical instruments, computers, controllers and creative software. As AI becomes part of music creation, that same role becomes even more important. AI-powered tools need ways to understand musical gestures, communicate with instruments, respond to performers and integrate into real-world creative workflows.
JAM_BOT shows one possible direction. It demonstrates how MIDI and symbolic musical data can help make AI music systems more interactive, more transparent and more performer-centered.
The Future: AI as a Creative Partner, Not a Replacement
The most exciting part of Jordan Rudess’s performance with JAM_BOT is not that a machine can play notes. Machines have been playing notes for a very long time. What is exciting is the possibility of a system that can respond to a musician in the moment, contribute ideas and help shape a performance as it unfolds.
That vision keeps the human musician at the center. Jordan Rudess brings the artistry, touch, listening, judgment and imagination. MIT’s JAM_BOT brings a new kind of responsive musical intelligence. Together, they point toward a future in which AI becomes another member of the creative ensemble.
For musicians, educators, developers and standards organizations, this is an important moment. The next generation of AI music tools will need to be built around musical values: expression, consent, creativity, interoperability and human agency. Projects like JAM_BOT help show how that future can sound.
Watch the video above to see Jordan Rudess and MIT’s JAM_BOT in action at the Berklee AI Music Summit.