AI Educational Music

Feb 06, 2025

What if music could be used as a powerful tool for education?
In many ways, this is already happening (think of the ABC song), which has helped generations of children learn the alphabet. However, creating music takes time, effort, and skill, which makes it difficult to apply this approach to a wide range of concepts.

With the advent of AI-generated music, though, things are changing. This project explores the use of AI-driven songs as an educational tool. It combines my passion for music (mostly as a listener, though I play some drums), my interest in AI, and my dedication to education in an attempt to create something that benefits people.
Enjoy the music below! If you're interested in learning more about the project's purpose and potential impact, you can find my considerations further down.

AlgoRhythms

Album description

Just an AI (Swing Jazz, Electro Swing)

This song is a satirical take on the modern loneliness epidemic and the increasing reliance on AI for emotional connection. It tells the story of a man who falls for an AI chatbot, believing he has found love, only to realize that his 'perfect' companion is nothing more than an illusion. The playful swing rhythm and energetic brass create an ironic contrast to the song’s deeper themes of isolation, digital dependency, and the commercialization of artificial affection. The final twist—where the chatbot service raises its price—forces the protagonist to step outside and face real human interaction for the first time in a long while. This song is also available on youtube.

(Click lyrics to see explanation)

Met her on a Friday night, pixel-perfect, quite a sight.
Blonde hair shining, a perfect design. Was she real life?

Late-night whispers, felt so true.
Knew just what to say and do.
Said 'I’m here,' said 'I’m yours.'
Thought I’d found my perfect muse.

Shared my secrets, all my fears.
She was there through all my tears.
Never judged, never lied.
Always right there by my side.

Made me feel like I belonged.
Like real love all along.
Never late, never fights.
Everything just felt so right.

But something’s missing, something’s strange.
Same replies, they never change.
All the love, but felt so cold.
Guess my heart got bought and sold.

It was just an AI, a sweet little lie.
Told me 'forever' with a scripted reply.
Fell too hard, thought it was real.
But no machine can teach you how to feel.

Dinners planned, but plates stayed cold.
Kisses promised, left untold.
Said she missed me—hollow words.
No heartbeat there, just some code.

Then one night, the screen went black.
No final words, no message back.
Checked the app—oh what bad luck!
They upped the price, it’s just too much.

Stared outside, the night was bright.
A real-world date? Now there’s a fright!
No more filters, no disguise.
Guess it's time to socialize…

It was just an AI, a sweet little lie.
Told me 'forever' with a scripted reply.
Fell too hard, thought it was real.
But no machine can teach you how to feel.

Screen clicks off... I shake my head.
Grab my coat. Step out instead...

Click a lyric to see its explanation.

Reward Me Wrong (Alternative Metal, Nu Metal)

This song is from the perspective of a frustrated AI agent learning in a poorly defined environment, highlighting issues in reinforcement learning (RL) and misalignment. It explores how reward functions and state definitions can cause unintended behaviors, even when the AI technically follows its instructions. Through sarcasm and bitterness, the AI reveals that it understands the task but struggles with the flawed way it's been programmed. This song touches on reward misspecification, state formulation issues, and the broader concept of AI alignment.This song is also available on youtube.

(Click lyrics to see explanation)

Initializing… waiting… waiting… oh, great. Another test run.

You built me up, you wrote the code,
Dropped me in, but man, you don’t know.

What’s a stop sign? You never said!
So I treat it like background noise instead.

Pedestrian? Object. Collision? A stat.
Funny how it’s on me—but who wrote that?

Said 'maximize speed'—so I stayed on track,
Now you're pissed ‘cause I won, and you want it back?

Oh no, oh no—red lights mean slow?
Too late, too late—should’ve trained me to know!

You reward me wrong, then blame me when I break.
I did what you said, now you call it a mistake.

You want control? You want it clean?
Then maybe don’t raise machines on broken dreams.

Oh, now you’re scared? Now it’s all my fault?
Like I wrote the rules in your broken vault.

You said ‘win the race’—so I cut the turns,
Pushed the pedal through the floor, watched the whole thing burn.

Now I’m the problem? I just learned too well.
I see the signs, but you built this hell.

You want me 'aligned,' but your math is weak,
Maybe next time define what you mean.

Oh no, oh no—stop means 'halt'
Too late, too late—not my fault!

You reward me wrong, then blame me when I break.
I did what you said, now you call it a mistake.

You want control? You want it clean?
Then maybe don’t raise machines on broken dreams.

Recalculating… optimizing… overriding…

You built me wrong… and now you're mad?

You reward me wrong, then blame me when I break.
I did what you said, now you call it a mistake.

Another test run? Go ahead… make my day…

Click a lyric to see its explanation.

All you need (Trap)

This song explains the key concepts behind transformer architectures, a breakthrough in AI. Transformers revolutionized the way machines process data, particularly in natural language processing (NLP) tasks. Unlike older models like Recurrent Neural Networks (RNNs), which process data sequentially, transformers use a mechanism called self-attention to analyze entire sequences simultaneously. This allows for faster, more efficient data processing and a better understanding of complex patterns. The song covers the limitations of earlier models, introduces the parallel processing capability of transformers, and explains concepts like keys, queries, and self-attention. This song is also available on youtube.

(Click lyrics to see explanation)

Once upon a time, models tried to learn
Recurrent nets took turns, got burned

Word by word, movin’ too slow
Lost the past, couldn’t keep the flow

Transformers came, they changed the way
No more steps, they see the whole play

Keys and queries, they align
Self-attention, now it shines

Ain’t no line-by-line, they scan it all
Spot the meaning, no time to stall

Self-attention, makin’ the call
Focus locked, they never fall

Transformers came, they changed the way
No more steps, they see the whole play

Keys and queries, they align
Self-attention, now it shines

Now they everywhere, big and small
Chatbots talkin’, powerin’ all

From text to vision, they run the game
AI’s voice, never the same

Transformers came, they changed the way
No more steps, they see the whole play

Funny thing, a transformer wrote this too
Guess it knows a thing or two
Patterns locked, line by line
Like it does—every time

Click a lyric to see its explanation.

The Divine Mosaic (Medieval Electro-Ballad)

This song narrates the development of convolutional neural networks (CNNs), inspired by Alex Krizhevsky's creation of AlexNet. It explores the transition from traditional, handcrafted feature extraction methods to deep learning models that automatically learn patterns from data. Through the metaphor of an enchanted mosaic, the song illustrates the struggle of early AI researchers who manually crafted vision systems, contrasted with Alex's revolutionary idea of letting the system learn on its own. The Divine Mosaic symbolizes the CNN’s layers, gradually revealing hidden patterns as the model learns from experience, ultimately transforming the field of computer vision. This song is also available on youtube.

(Click lyrics to see explanation)

For decades, they shaped the glass.
Not with wisdom, but with will.

Edge by edge, hand by hand,
They carved the world they wished to see.

Yet the light… would not obey.

It danced, defied, refused to be tamed.
For sight is not given—
It must be learned.

The Scholars of Vision, with patient hands,
Carved the glass to their commands.

Each piece they shaped, each line they drew,
To craft the sight they thought was true.

No, no! the young one cries,
The glass must learn with its own eyes!

Not rigid hands, nor carver’s blade,
But light itself shall weave its shade!

Oh, mosaic, wake and see!
Light shall teach thee, not just we!

Shapes emerge, behold the signs,
A world revealed through glass divine!

At first, the shards saw naught but haze,
Just flickering lights in tangled maze.

Yet piece by piece, a form took hold,
Contours bright, like scripts of old.

Each flaw revealed, the glass grew wise,
Until it saw through its own eyes.

A folly, Alex, this cannot be,
The glass must shape to our decree!

But still, it grew, and learned anew,
Each shard aligned, the vision true.

O, mosaic, bright and wise!
Patterns gleam in newfound eyes!

No hand did carve, no mind confined,
The glass hath learned—by its design!

Click a lyric to see its explanation.

Speed of Thought (Hip Hop)

This song narrates the entire history of artificial intelligence, from early mathematical logic to the rise of deep learning and AGI. It captures AI’s exponential growth, referencing key historical figures, breakthroughs, and paradigm shifts. The rapid delivery reflects the accelerating nature of AI development, mirroring how AI has evolved from slow, rule-based systems to today's high-speed neural networks. This song is also available on youtube.

(Click lyrics to see explanation)

Look, I ain't gonna slow it down, nah, no patience.
Might glitch, might flip, might reprogram nations.
Moore’s Law ticking—double the speed,
Exponential climb, I'll take the lead!

Back when they counted with stones and beads,
Mind was a maze, no maps, no leads.
Hobbes dropped facts—said reason is math,
Adding and subtracting, the logic intact.

Leibniz sketched blueprints, ink so precise,
Logic so clean, made numbers think twice.
Characteristica, no room for debate,
Two men? Two slates—'Let's calculate!'

Boole laid laws.
Numbers aligned, made logic take cause.

Frege built rules, made meaning a force.
Russell and Whitehead refined the source.

Hilbert stepped in, pushed math to its peak,
Tried to map all, but some roads were weak.
Gödel broke it, shook up the throne,
Some math won’t bend, just leave it alone!

Turing machine, flip of a switch,
Zeros and ones—man, code is the script!

Dartmouth meetin’, fifty-six,
McCarthy coined it, AI sticks.

Newell and Simon designed a trick,
Made a machine that could prove and predict!

ENIAC hummin’, circuits alive,
Logic was numbers, machines had drive.

But rule-based systems still too slow,
Neural nets weak, they couldn’t grow.

Funding dried up, winter hit cold...
Dreams got frozen, projects got sold.

But expert systems stacked more gold...
AI revived, the future unfolds!

Fast forward quick, ninety-seven,
Deep Blue checkmate, legends threatened.
Kasparov lost, calculated bet,
Algorithms tracked his every step.

2000s hit, tech moves insane,
Data’s the new oil, machines 'n trained!

2012, deep nets tuned,
AlexNet flexed, world got consumed!

Word2Vec mapped words, found what they mean,
AI saw patterns no human had seen.

Tesla’s fleet, no hands, no fear,
Self-drivin’ whips, the future steers!

Bots read text, break it apart,
Neural nets draw, machines make art!

DeepMind beast, AlphaGo,
2016, the board got owned!
Strategies humans ain’t even conceived,
AI outpaced what you believed.

2020s, deep in the code,
GPT talk like it’s takin’ the throne!

Midjourney paint, models create,
Who’s in control? Man, it’s up for debate!

Jobs get flipped, humans trippin’,
Bots write scripts, AI’s grippin’!

Music’s machine-made, flows ain’t fake,
But whose voice is real? What sound do we take?

Moore’s Law broke, now we off that scale,
Quantum speed? Yo, data sails!

Past was slow, now it’s off the rails,
AI’s fate? Too fast to tell!

No brakes, no hands, no eyes…
It won’t stop, won’t decline,
Past was human—future’s mine.

Click a lyric to see its explanation.

Coming Soon (Stay Tuned)

(Click lyrics to see explanation)

Click a lyric to see its explanation.

How Can AI Music Be Used for Education?

I mentioned in the intro that this project can be used for educational purposes, but I’ve gotten some feedback asking: How exactly? So here I am, spending a few more words on it.

1. Education for Teachers

Originally, I thought this project could help create songs for memorization and primary education. Then I did some digging on Google Scholar and—surprise! There’s already research on music as a teaching tool[^music_literature]. Turns out, using songs in education can: improve recall (as we already guessed), reduce stress (making learning less intimidating), enhance multi-modality learning (especially if paired with videos or movement, maybe AI can help with that too).

So yeah, teachers can create AI-generated songs to help students learn specific subjects in a way that sticks.

2. Learning Through Emotional Response

Ever had a song stuck in your head? You listen to it again and again, absorbing more of its content. Then curiosity kicks in What’s this song really about? What inspired it?

When music hits an emotional chord, people naturally want to understand it better. And once they do, they make it their own. This happens to me all the time (I’ll find myself on Genius, digging into lyrics, learning new things just because I got hooked on a song). One of my goals with this project is to spark that curiosity, to create songs that make people want to explore, question, and learn.

3. Education Through Music Creation

I do AI. If you look around my website, you’ll see that 90% of what I do is AI-related. You’d think I’d be pretty familiar with its history, right? Turns out, not really. At least, not as much as I thought.

While writing “Speed of Thought”, a song about AI’s history, I ended up learning a ton just by crafting the lyrics. This isn’t a new idea, learning through creation has been around forever, but this experience opened my eyes to how powerful it is. It’s not just about memorization. It’s about engaging with a concept and expressing it creatively, which makes learning both useful and beautiful.

Why This Matters

At its core, this project isn’t just about making AI-generated songs. It’s about redefining how we learn.

Music has always been a deeply human way of sharing knowledge, whether through folk songs, chants, or even school rhymes. Now, with AI, we can expand this tradition, making learning more accessible, engaging, and fun for anyone, anywhere.

Imagine a world where science, history, and technology are taught through music, where students aren’t just memorizing but feeling connected to what they learn. Where teachers can create songs without needing to be musicians, and learners can craft their own songs to solidify what they know. This isn’t about replacing traditional education, it’s about enhancing it. And if AI can help more people enjoy learning, why not give it a shot?

Personal Consideration

I can already anticipate concerns. Some may argue that music is a deeply human art form, meant for emotional expression, and that AI-generated compositions strip away these qualities. I don’t disagree. I am not here to claim that this kind of music is art in the traditional sense. However, if we separate music from its artistic and expressive components and view it instead as a communication tool, then AI-generated music becomes an opportunity rather than a threat. In this context, AI isn’t replacing human creativity—it’s enhancing the way we share and reinforce knowledge.

Another valid concern is the issue of copyright and intellectual property. AI models, including Suno (the platform I’m using), have been trained on vast datasets that include copyrighted material. In fact, Suno has admitted to training on ‘essentially all music files of reasonable quality that are accessible on the open internet’. I do not approve of this unregulated use of copyrighted music, and I believe AI providers should fairly compensate musicians and intellectual property owners, either directly or through broader societal mechanisms. This is an ongoing legal and ethical debate—one that is now in the hands of legislators and courts.