Suno AI in Music School?
What Berklee Got Wrong When It Lit the Controlled Burn
I just published a video on my YouTube channel about a question I suspect a good number of readers of The Augmented Educator will care about as well: whether a college should build a course around a heavily debated generative AI tool, and, if it does so, how it ought to handle the inevitable controversy. The following essay is meant as a companion piece to that video. It presents the information in a more readable form than just a video transcript.
I should point out that, while this text has been created with the help of AI, in particular Claude 4.8, the original video transcript was written entirely by myself without AI assistance beyond a grammar check. So, if you are interested in learning more about how AI influences my writing, I invite you to check out the video as well. I would be curious to know which one you prefer. Let me know in the comments.
But back to the original question.
The starting point for this essay is a course at the Berklee College of Music, the institution most people would name first if asked where one goes to study contemporary music. In late March 2026, Berklee’s songwriting department began heavily promoting SW-303, “Bots and Beats: AI and the Future of Songwriting,” a long-standing elective that has been reoriented around the generative audio platform Suno.
Overseen by course chair Rodney Alejandro and taught by associate professor Ben Camp, the course asks students to generate lyrics, melodies, and finished recordings in direct collaboration with the AI software. The reaction was swift and unforgiving, and it was summarized by a careful video essay titled “They Teach AI Music at Music School Now...,” from Adam Neely, a bassist, Berklee graduate, and one of the most respected music theory educators on YouTube.
Neely organized his critique into four problems the course created: branding, the school’s tendency to be out of touch, the conflation of all machine learning with Suno’s particular methods, and the fact that many people simply hate AI music.
As readers of this Substack are aware, I am not a neutral party in this discussion. I use generative AI constantly, and I defend it more openly than almost anyone in my field. But I still agree with nearly everything Neely said. I highly recommend watching his video.
What is most interesting to me is how thoroughly Berklee botched the attempt to cover generative AI music within the curriculum. I think many of us can learn from their mistakes. Introducing a destructive technology into a place built on skilled craft is not unlike a controlled burn. Foresters set fire to a forest on purpose, under tight conditions, behind containment lines, with trained crews, precisely so that a far worse fire does not later arrive on its own terms.
The fire Berklee set by offering a Suno course was therefore not the mistake. The mistake was lighting it in the dry season, with no firebreaks, and handing the torch to someone with a private reason to watch it spread. That, in essence, is what the school did.
Why I can speak to this
A quick word on why I think I can add something here.
I have spent over three decades as a technology educator, most of it on exactly the kind of frontier where this fight is now playing out. In the late 1990s, I was already writing about the use of computer games in education. And as far as I know, I published the very first academic paper on professional gaming, or esports, in the early 2000s, when the field was still dismissed as frivolous.
During the pandemic, I wrote and taught about blockchain and its plausible uses in the creative economy, a subject that drew its own share of eye-rolling. Premature-seeming, contested technology is, for better or worse, a specialty I built my academic career on.
But I need to be careful. I cannot claim insider knowledge of Berklee itself, and I will not pretend otherwise. However, I have spent the last twelve years teaching at a media arts and design college with several programs ranked in the national top ten. And as a department head, I have had to make exactly the decision Berklee made: whether and how to bring a disruptive, contested tool into a curriculum that highly respects the slow, manual craft of creation.
So I know, from the inside, the difference between doing it well and doing it the way Berklee did.
The branding problem, or why you do not light the match in your own house
Neely’s first point is his strongest, and I agree with it almost without reservation. When students enroll at a college, at least in the American system, they are not only buying skills. They are buying into what the institution stands for. Berklee stands explicitly for the idea that making music is worth loving and worth paying tens of thousands of dollars to learn to do better.
The worldview of Suno, the company Berklee promoted, is entirely different, however. In an early-2025 appearance on the 20VC podcast, Suno’s chief executive, Mikey Shulman, offered a memorably bleak account of his own product’s purpose. He said it is not really enjoyable to make music, that it takes too much time and practice, and that most people do not enjoy most of the time they spend doing it.
His ambition, he explained in the same conversation, is to give “a billion people” the ability to generate music instantly, in a future where traditional skills no longer matter. To be fair, he later told Billboard he wished he had chosen different words. But by that time, the damage had already been done.
I need to acknowledge that this is a coherent position, and even, on its face, a democratic one. Lowering the barrier to creative expression is not a self-evidently evil goal, and I have spent years defending technologies that did precisely that. But consider Shulman’s words next to Berklee’s tuition statement, and the contradiction becomes almost comical.
A conservatory’s entire premise is that the friction Shulman wants to eliminate is not an obstacle to artistry but the central path that forges it. For an artist-first school to promote a tool engineered to dissolve that friction is to tell its own students that the thing they came to learn is a waste of their time.
The fire became a wildfire because of who was holding the torch
The instructor of record, Ben Camp, is a genuinely accomplished songwriter with more than a hundred million streams to his name, who, by some pleasing irony, teaches a Berklee course called “Stealing from the Masters.” But he is also, according to reporting the students surfaced, a paid advisor to Suno.
I think this is where the actual damage was done. Inviting Camp as a guest speaker, so that students could question him directly, would have been excellent pedagogy. But installing a paid advisor to Suno as the instructor of record, inside a tuition-funded classroom that functions as a captive market, is not a defensible call at all.
I would not permit this in my department, and I am frankly surprised it survived whatever review it received.
The students, led by organizers including Rylan Heasley, immediately identified the very obvious conflict of interest. They circulated petitions, gathering somewhere between several hundred and well over a thousand signatures, demanding that the course be disbanded. Berklee’s administration responded that an artist-first institution at the forefront of contemporary music has a responsibility to prepare students for the tools reshaping their industry.
That defense is not wrong. But it answers a question almost no one was actually asking.
Why being out of touch can be a feature
Neely’s second charge is that Berklee is hopelessly out of touch with current technology. I want to push back on the general premise a bit, though not in the direction you might expect. Being somewhat out of touch is not necessarily a bug in higher education. It is often a feature, and a load-bearing one. Colleges cannot chase every trend the moment it appears.
Imagine the disaster if universities had rebuilt their degree programs around blockchain ledgers and NFTs at the peak of that hype five years ago. The subsequent collapse would have taken every curriculum down with it. And I say this as someone who actually taught the blockchain’s plausible uses during the pandemic and still believes it has real value for the narrow set of problems that require an immutable, decentralized database.
The point of college is not to teach you which button to press in Pro Tools this year. You can get that information from YouTube tonight. Instead, college is supposed to teach you how a tool’s architecture shapes your creative decisions, and how to keep your footing when the tool inevitably changes underneath you. A useful conservatism filters out the buzzwords and keeps the durable competencies in.
But conservatism is not the same as ignorance, and a school has to understand what a technology actually does. This is exactly where Berklee’s execution fell apart. The course announcement strung together generative AI music, the blockchain, and the metaverse in a single sentence, as though they were three flavors of one thing.
They are not.
The blockchain is a decentralized ledger technology with real applications. And the metaverse, a term that comes from Neal Stephenson’s 1992 novel Snow Crash, describes a persistent virtual world with its own economy. Generative audio is a third thing entirely. Bundling them reads less like a syllabus than like a venture capital pitch deck, and it handed critics every reason to dismiss the entire enterprise as intellectually hollow.
I need to add one friendly correction to Neely’s own argument here. He treats “metaverse” as shorthand for the product Mark Zuckerberg built and abandoned at Meta. That is the colloquial meaning, and it will resonate with Neely’s audience. But it is not what the word means when it appears next to “blockchain.” In that pairing, people are almost always talking about token-based virtual economies, not Zuckerberg’s headset venture.
It is a small thing, and it does not detract from his general thesis. But on a point about being out of touch, it is a little unfortunate to be slightly out of touch oneself.
Not all machine learning is scraping
Neely’s third point is that fixating on generative audio erases the long, uncontroversial history of machine learning in music production, and on this he is plainly right.
The industry has used machine learning for assistive work for well over a decade. This includes intelligent EQ and mastering tools from companies like iZotope and Sonible, or stem-separation systems. These tools are already taught at Berklee, and rightly so, because they analyze the audio a user feeds them rather than ingesting the catalogs of artists who never consented.
A school that wants to teach AI responsibly has a rich, ethically clean set of examples sitting right there. But Berklee reached past all of that for the one platform guaranteed to start a fight.
Where I disagree slightly with Neely is on the reasoning, not the conclusion. His case rests on the claim that the use of Suno is unethical because it was trained on copyrighted recordings scraped without permission. In the wider debate, that argument is usually compressed into a single word: “stealing.” For the models Suno is running today, that critique is well-founded. But the general picture changed in late 2025.
In June 2024, the RIAA, acting for Universal, Sony, and Warner, filed coordinated suits against Suno and its competitor Udio. Universal settled with Udio in October 2025, and Warner settled with Suno that November, in a deal that also had Suno acquire Warner’s Songkick live-music platform and commit to licensed, opt-in models built on Warner’s catalog.
Going forward, in other words, at least part of Suno’s output is moving onto legitimately licensed ground. Universal and Sony are both still in court, but a fair-use ruling expected this summer is likely to set the precedent for the whole industry.
And then there is the announcement’s single most problematic blunder. To advertise the course’s creative possibilities, the promotional material suggested students could use Suno to put the voices of Drake and Grimes into their own songs.
Whoever wrote that chose the two worst possible names. At the time of this writing, Drake records for Universal, the label suing Suno for billions and the one that once forced a viral AI imitation of Drake’s voice off the internet. Promoting his unauthorized cloning is advertising precisely the unethical conduct that the lawsuits are about. And Grimes is the mirror image, an independent artist who has openly invited fans to clone her voice and even offered to split royalties fifty-fifty on anything they make with it.
Setting the artist who fought voice cloning hardest beside the one who welcomed it most, and tying both to a settlement that covers neither, is a public demonstration that the people in charge did not think the ethics through.
Who actually hates AI music?
Neely’s fourth point is the most direct. He argues that many people just hate AI music. Within his world, he is likely right, but I think this is only part of the story.
His example is the musician and YouTuber Gabi Belle, whose video “AI has already ruined music” gathered well over a million views and became a rallying cry. But Belle’s fury has a context that explains its intensity. She was the target of a malicious deepfake campaign, with bad actors generating and circulating roughly a hundred non-consensual pornographic images of her, an experience she has described as leaving her feeling violated.
When a tool associated, in your own lived experience, with that kind of violation is then turned toward music, anger is not an overreaction. It is close to the only sane response. That intensity is obviously real. But the question Neely’s framing leaves open is whether it scales beyond the communities where it lives.
We all inhabit echo chambers, and a music theory channel’s comment section is among the more committed ones on this subject. The broader public likely looks at this differently.
The clearest evidence is a survey Deezer commissioned from Ipsos in late 2025, polling nine thousand adults across eight countries. Its headline finding is that 97 percent of listeners could not distinguish fully AI-generated tracks from human ones in a blind test. Roughly 80 percent wanted AI music clearly labeled, 73 percent considered training on copyrighted work without consent unethical, and 70 percent thought AI threatens musicians’ livelihoods.
That is the shape of public opinion. I would argue that it is not hatred. Wanting a label on something is not the same as despising what it stands for. I would like to know whether my orange juice came from concentrate, but that does not mean I hate orange juice made from concentrate. The active, organizing hate Neely describes is primarily found in engaged musicians and serious listeners. Out in the wider world, the mood reads closer to mild curiosity and a very reasonable demand for transparency.
At least not yet.
Where this leaves us
So where does this botched controlled burn leave us?
Neely is right about the thing that is most important. The way Berklee handled this, from the branding to the buzzword-stuffed announcement to the Drake example and, above all, the decision to put a paid Suno advisor in charge of the class, was a serious failure. I would argue that it did not respect what the college or its students stand for.
This would certainly not have happened in my department.
Where I would add nuance is nearly everywhere else. Being somewhat out of touch could be seen as a virtue in an institution meant to outlast trends. The ethics of a platform like Suno are messier than simple theft now that licensed models are entering the picture, even as they remain genuinely unresolved in court. And the general public is probably not as hostile as Neely’s corner of the internet suggests.
But none of that invalidates his argument.
Here is where I want to come back to the controlled-burn metaphor. You cannot keep fire away from the forest forever. AI-generated music is the marketplace these students will graduate into, and pretending otherwise serves no one. The fire is coming either way, and we need a controlled burn to mitigate the most severe consequences. AI music will need to be addressed in the curriculum, even at a school such as Berklee.
P.S. I believe transparency builds the trust that AI detection systems fail to enforce. That’s why I’ve published an ethics and AI disclosure statement, which outlines how I integrate AI tools into my intellectual work.