Speaking and keynotes
AI, creativity, and what makes ideas interesting
My talks connect research, working systems, musical examples, and philosophical questions. Drawing on three decades of fundamental and applied AI, I explore what these systems make possible, how people can steer them, and what remains to be understood.
Discuss a speaking invitation
Proposed talks
The focus and format can be discussed around your audience.
Can AI create? From Flow Machines to today's generative models
What does it mean for a machine to create music? Starting with the Continuator and Flow Machines, this talk follows the development of tools that learn musical styles and let musicians shape generated material. Musical examples open a discussion of intention, originality, and the part played by human collaborators.
- For
- Public audiences, creative industries, professional conferences
- Possible format
- Illustrated keynote with musical examples and Q&A
What makes an idea interesting? Beyond plausible generation
What do we mean when we call an idea interesting? How might we characterize that quality, and how do such ideas emerge? Drawing on computational creativity, music, and philosophical inquiry, this talk examines the relationships between plausibility, originality, constraints, and intention. Interestingness itself is part of the investigation.
- For
- Research and innovation teams, executives, interdisciplinary audiences
- Possible format
- Keynote or moderated discussion
Simulating people and societies: what synthetic populations can tell us
How can aggregate statistics be turned into synthetic populations, and what can we learn from them? This talk introduces probabilistic models and virtual surveys, distinguishing population constraints from assumptions about behavior. It considers the uses of simulation alongside the need for calibration, scientific validation, and a clear account of its limits.
- For
- Decision makers, data teams, social researchers, public-policy audiences
- Possible format
- Lecture with research examples and discussion
Thirty years of AI: symbols, constraints, and generative models
A personal research perspective on successive approaches to AI, from symbolic systems and Markov models to constrained generation and current generative models. Concrete projects and experience leading research teams at Sony CSL and Spotify provide a way to examine what changed, which ideas endured, and how fundamental research connects to practical applications.
- For
- Technical conferences, students, professional and leadership audiences
- Possible format
- Lecture or moderated retrospective
Watch and listen
AI and music: Is the problem solved?
A recorded lecture from Festum π. The complete recording runs 1 hour 35 minutes.
Watch the lectureSynthèse artificielle et contribution humaine
François Pachet appears alongside other speakers. This link opens the complete session, lasting about 3 hours 28 minutes.
Open the full session recordingExplore research demonstrations: Continuator · Flow Machines.
Short biography for organizers
François Pachet is an AI researcher, inventor, entrepreneur and musician, and an associate researcher at LIP6. Former director of Sony Computer Science Laboratory Paris and Spotify’s Creator Technology Research Lab, he co-founded Imagine All The People and Ynosound. His research spans interactive music, musical style and constrained generation, including MusicSpace, the Continuator and Flow Machines. He also composes, performs and writes about music and creativity.
Full biography and recognitionDiscuss your event
For a speaking invitation, please include the event, intended audience, proposed date, location or online format, preferred language, and topic of interest.
Talks in French and English
Past talks and other resources
The full historical archive remains available, including academic seminars, public lectures, interviews, and their original resources.