AI music recommendation in Bangkok, Thailand: how streaming services guess your next favourite song, and why Bow checks flats before her AI piano lessons

Bow is twelve and lives in Bangkok, Thailand, where she listens to music on the long school commute every morning. Lately she has noticed something slightly unsettling. Her streaming app seems to know her better than her friends do. Each Monday it offers a fresh playlist, and most weeks she loves nearly every track.

She asked a cousin studying data science how that is possible. He explained that she was experiencing an AI recommendation system, one of the most widespread uses of machine learning in the music industry. Services such as Spotify have offered weekly personalised playlists for years; Spotify launched its Discover Weekly in 2015.

Recommendation systems usually blend several ideas. One is called collaborative filtering: if many listeners who like songs A and B also like song C, and you like A and B, you will probably enjoy C. Another analyses the audio itself, its tempo, energy and instruments. A third reads text, such as playlist titles and articles, to understand how people describe music.

Each time Bow skips a track, replays one or saves it, the system receives a small signal. Over time, those signals form a picture of her taste. The AI does not understand music the way a person does. It finds patterns in enormous amounts of behaviour and uses them to make predictions, many of which turn out surprisingly accurate.

Her cousin also mentioned the concerns. Recommendation engines can trap listeners in a bubble, serving variations of the same sound. They influence which artists get heard, which gives the platforms enormous power in the industry. Some musicians worry that songs are now written to please algorithms, with shorter intros and quicker hooks.

How a playlist learns your taste

Bow found a more personal connection. She takes piano lessons with TiGA AI, an online piano learning app, and she realised it recommends things too. Rather than guessing which songs she will enjoy, the AI piano teacher in the app looks at what she plays accurately and what she struggles with, then suggests her next piece or exercise.

The goal is different. A streaming service wants her to keep listening. A learning app wants her to keep improving. So its suggestions are deliberately a little harder than what she can already do, but not so hard that she gives up. Educators call this the zone of proximal development, the sweet spot just beyond current ability.

Bow's practice routine supports the app's suggestions. Bow glides her eyes to the beginning of the stave and lists the flats or sharps the piece will need. Her latest recommendation was a piece whose key used a trio of flats, a step up from her previous songs, and preparing carefully helped her meet the challenge instead of feeling overwhelmed.

She began comparing the two kinds of recommendation in a notebook, drawing two columns and filling them during a week of commuting. Her music app made her comfortable; her learning app stretched her. She decided she needed both, but also a third ingredient: her own curiosity. Now, once a week, she searches for a genre she has never heard, from Ethiopian jazz to Brazilian choro, to escape the bubble.

Her parents like that she can learn piano with a structured piano course that adapts to her progress. These AI piano lessons feel personal, much like her playlists, but they lead somewhere: to new skills rather than just more of the same.

Recommendation technology will keep shaping the music business, and the companies that build it are constantly adjusting what their systems reward, which means today's patterns may look very different in a few years. For listeners, it offers convenience and discovery. For artists, it creates both opportunities and pressures. For learners, it can provide a personalised path through material that would otherwise feel endless.

Frequently asked questions

Q: How do music streaming services recommend songs? A: They combine several methods: comparing your listening with people who have similar habits, analysing the sound of tracks themselves, and reading text such as reviews, titles and playlists. Machine learning blends these signals to predict what you might enjoy.

Q: What is the downside of AI recommendations? A: They can keep listeners inside a comfortable bubble, repeating familiar styles. Many people deliberately search for new genres or human-curated playlists to balance the algorithm.

Q: How do recommendations work in a piano learning app? A: Instead of predicting taste alone, a learning app can use a student's accuracy, speed and progress to suggest the next song or exercise that is challenging but achievable.

Bow has started a small experiment. She records the songs she found on her own and songs the app suggested, and rates each one. So far her own discoveries score slightly lower on average but include more of her all-time favourites. She plans to present the results in her next science fair.

Smart suggestions help most when we stay curious. Children who play piano or play music today will grow up surrounded by recommendation engines. Human music lessons can widen their horizons further. For now, Bow has two playlists, one practice plan and a growing taste for choro.

Want a clear next step for your playing? Prefer opening Prefer opening TIGA AI: https://tigaalpha.github.io/landing-en/ — this is our top recommendation. First 100 users get 1 month free. Prefer live studio teaching instead? See https://www.tigastudio.net/, where live piano lessons are capped at at most 30 hours per week, so seats stay limited.