This special episode is an inside look at AI music from three very different vantage points: the builder, the investor, and the industry insider.
Andreas is joined by Sundar Arvind, CEO & Co-Founder at Mozart AI, building a collaborative generative audio workstation; Daniel Waterhouse, General Partner at Balderton Capital; and Ash Pournouri, Co-Founder of Belong, entrepreneur, producer, and former manager of Avicii.
Together, they unpack how AI is reshaping music creation, how serious investors underwrite risk in a litigious industry, why “one-click songs” miss the point, and whether AI expands creativity or commoditizes it.
If you want a grounded view of where the real fault lines are — rights, training data, authorship, collaboration, and the psychology of creativity — this is it.
What’s covered:
00:40 Mozart AI’s vision: a collaborative generative audio workstation
05:10 DAWs, EDM, and why tech has always expanded music creation
06:35 Why “one-prompt songs” optimise for quantity, not craft
09:20 Underwriting AI music: how VCs think about billion-dollar incumbents
13:00 Is this a new instrument or a 100x larger market?
18:45 Are professional artists already using AI tools?
21:00 Copyright, training data, and legal diligence in AI music
25:15 Philosophically: what are “rights” when machines learn from music?
33:40 Diffusion models explained simply: how AI generates sound
36:30 The return of the band? Multiplayer music creation
40:00 Ash Pournouri joins: the industry’s instinct is protection
44:10 “You can’t stop development”: why demand always wins
48:50 Packaging matters: AI as tool vs AI as replacement
51:20 Lowering thresholds and democratization across decades
53:30 Five-year predictions? We’re on the vertical part of the curve
54:20 The “vibe coding” moment for music
🎧 Listen on Apple or Spotify, or queue it for later with chapters ready to go.
Show Notes
Mozart AI’s lens: not a “one-click song,” but a new creative workstation
Sundar is clear from the start: Mozart AI is not trying to win the viral prompt race.
Yes, you can generate a song from a single prompt. But that’s not the point.
The real thesis is about preserving and upgrading the creative journey:
– reduce engineering friction
– increase time spent translating intent into sound
– keep the human in the loop
– eventually make music creation multiplayer
One-click songs optimize for novelty and volume.
Creative tooling optimizes for authorship.
And that distinction becomes central to everything that follows.
Underwriting AI music: conviction beyond incumbents
Daniel addresses the obvious question: why invest in a space where billion-dollar players already exist?
The answer is threefold:
– exceptional founder conviction
– massive, underappreciated market expansion
– fundamentally different product philosophy
The bet isn’t on prompt-based song generation.
It’s on collaborative, creative infrastructure.
Daniel frames it simply: this isn’t just a new instrument. It could become a new layer of participation in culture.
If even a fraction of music listeners become music creators, the market expands dramatically beyond today’s “music industry.”
The industry’s first instinct: protect, restrict, control
Ash’s central claim is consistent with history: the music industry is conservative.
Not because it’s irrational, but because it is structurally built around:
– rights management
– administration
– legacy power structures
– guarding monetization
When disruption arrives, especially anything that looks like extracting value from IP, the response tends to be:
– pushback
– lawsuits
– attempts to restrict
– attempts to gain ownership/control
AI music is no exception.
Ash compares this moment directly to Napster and Spotify: innovation isn’t welcomed. It’s negotiated under heavy guardrails.
“You can’t stop development” demand forces the market open
Ash takes a market-first stance.
If the audience wants something, development will happen.
Restriction may slow it, but it won’t stop it.
Illegal file-sharing revealed demand → legal streaming models emerged → the industry adapted.
The same structural pattern is playing out again.
He adds a creator-centric insight:
If your audience is already somewhere,
you either exist there or you don’t exist to them at all.
Packaging is everything: tool vs replacement
Ash’s sharpest insight isn’t technical — it’s strategic.
Adoption depends on framing.
If AI is presented as replacement, the industry treats it as an existential threat.
If AI is presented as a tool that increases productivity and creativity, acceptance rises sharply.
He uses MTV as analogy:
MTV didn’t pitch itself as monetizing artists’ IP.
It pitched itself as marketing and reach.
The same logic applies here:
Tool framing unlocks cooperation.
Replacement framing triggers war.
Lower thresholds, wider participation
Sundar and Ash converge on a historical pattern:
Music has repeatedly lowered the barrier to entry:
– you once needed instruments + studio access
– then came digital audio workstations
– then Autotune
– then bedroom producers
– then streaming distribution
Each wave triggered panic.
Each wave expanded participation.
AI is the next threshold-lowering wave.
More people can create.
Faster.
With less overhead.
And as Sundar notes, this doesn’t remove skill, it reshapes where skill matters.
Inspiration vs infringement: the philosophical fault line
The debate isn’t only legal. It’s philosophical.
Humans learn by listening. Artists borrow, reference, remix, reinterpret. If human inspiration is legal and culturally accepted, where exactly does machine learning cross the line?
Sundar argues that when AI is assistive, compensated, and commercially cleared, it functions as a tool, not a thief.
Daniel points out the nuance: copyright law was built for a different era, and this category will likely evolve through negotiation, precedent, and global coordination.
Would AI have changed historic hits?
Andreas poses the uncomfortable question:
Would iconic tracks have been AI-produced or just improved?
Ash’s answer is psychological, not technical. Creators want authorship. They don’t feel creative if they’re cheating. They want appreciation for their ideas, not for pressing a button.
AI could:
– speed up execution
– unblock ideas
– reduce friction
– remove admin
But the core creative impulse doesn’t disappear.
AI helps you get what’s in your head into the world faster.
It doesn’t replace the head.
The new skill curve: craft still matters
Sundar highlights an emerging gap:
There is already a clear difference between:
– average prompt-only users
– high-skill musicians using AI iteratively
The best users aren’t just generating.
They’re refining, exporting stems, remixing, iterating.
AI expands the floor.
But the ceiling still belongs to:
– taste
– craft
– repetition
– experience
The future is hard to predict because we’re on the vertical curve
When asked what music looks like in five years, Ash refuses to speculate. Not out of caution, but because the pace of change is exponential. We are not in linear development.
“Ten years happen in one year.”
New tools arrive daily. Two years ahead already feels unpredictable.
The slope is vertical. The unexpected upside: AI may bring creators back.
The episode ends on a surprisingly optimistic note.
Ash admits something personal:
AI makes him want to create again.
Why?
Because it strips away:
– admin
– coordination overhead
– friction
It allows ideas to ship faster without requiring an entire machine around them.
In an industry where bureaucracy exhausts artists,
that shift is not trivial.
One-line takeaway
AI won’t kill music.
It will lower the threshold to create, expand participation, force the industry to renegotiate control and leave true authorship where it has always belonged: with humans.
