I write this while Grigory Sokolov, one of the greatest living pianists, plays a recital of some of the greatest works of all time, including Beethoven's Hammerklavier Sonata, on the TV in my living room. It is meticulously recorded and produced to show every gesture and facial expression.

And I’m treating it like background music.

As I fast-forward through the applause to get to the encores (three rarely-played pieces by Rameau -- no, make that four. Now five.), I contemplate which of the masters to put on next. Martha Argerich playing Prokofiev? Horowitz playing Rachmaninoff or Rubinstein playing Chopin? Du Pre playing Elgar? Will I listen to Jonathan Biss playing a Beethoven sonata, or just listen to Schnabel himself?

I do exaggerate my sacrilege a bit. When Sokolov begins his sixth encore, a Brahms intermezzo that I’ve heard a hundred times, his playing commands my attention, and I am enraptured. And when Amazon Prime begins to autoplay the next video, a Yuja Wang recital of Schubert and Schumann, I pause it, shamed by my own writing to give more respect to the music.

Even though I was building websites as a child since the Geocities days in the mid-90s, my early musical education was squarely in the pre-YouTube, pre-streaming era. The recordings I grew up with were the handful that my parents owned and the albums I took back with me from visits to my grandparents' house. These were mostly the popular albums of the day: Van Cliburn's Tchaikovsky Concerto, Glenn Gould's Goldberg Variations, and my grandparents' treasure, the complete Beethoven symphonies recorded by Arturo Toscanini with the NBC Symphony Orchestra. Just because these recordings were mainstream doesn't mean they weren't great – even today Toscanini's Beethoven symphonies remain among my favorites, and I may never have fallen in love with Bach if I hadn't grown up with Gould's Goldbergs. But the point is that my early musical sensibilities were very much formed by the limited music that was available to me – more nurture than nature, as it were.

The greatest source of this nurture came not from my parents and grandparents, but from my teacher, Natalie. One of the key features of her teaching was monthly performance classes where her students would play for each other. I appreciated them but can't say I really looked forward to them – except for the last one of the year. Natalie gave each student a gift at the last performance class of the year: a recording from her collection tailored to each student individually for summer listening.

Just as I was playing my first Chopin etude and feeling good about my accomplishment, she gave me a recording of Liszt's Transcendental Etudes to show me how much farther there was still left to go ("Never rest on your laurels", she'd say). One year she spent focusing on the development of the piano and gave me Jerome Lowenthal's Steinway Dynasty, featuring ten different Steinways from the mid 19th century to today. The most disappointing was the year she gave me a recording of 30 different pianists all playing the same 2-minute etude, Chopin Op. 10 No. 1, but the lesson that there are an endless number of ways to play the same piece of music still stays with me. There was Yundi Li's Liszt sonata to give me a flavor for what the piano competition world was talking about. Then there was Boris Berman's Prokofiev concerti, which blew my mind by introducing me to music past Debussy for the first time.

And so it was that I was exposed to the world of classical music. Curated based on what she felt I needed, I inhaled each of those recordings. How much each one affected my tastes and my playing is hard to say exactly. Only occasionally was it tactical. Going into high school, I was to learn the Schumann Concerto, but I had never actually heard it before, so she gave me Byron Janis's recording to get me excited about it over the summer. For most of the music I played, though, I was given no recording, just a score so that I was forced to develop my own way in to the music. Only after a piece was learned would we sit in her living room with the music in front of us and say "Let's see what Rubinstein does with this Polonaise" and then "And what does Horowitz do differently?" ("Other than the tempo", she'd always add. Don't even think about playing it that fast.)

Every recording was listened to with intent. And when there was something worth stealing from one of the great masters, it still needed justification. When I studied the Bach C Minor Fantasy (BWV 906), we listened to Gould, of course, but I wasn't allowed to follow him blindly and play it at half tempo without a musical reason of my own.

I didn't always have the discipline to resist temptation, however, even before recordings were readily available on YouTube. As a teenager I played my first Rachmaninoff Prelude, the gorgeous Op. 23 No. 4 in D Major. I found the notes not particularly difficult, but I found myself in that pattern oh-so-familiar to music students where you think you've "finished" a piece and are ready to move on, but every time you bring it into a lesson, your teacher is not quite satisfied with it and says "one more week." After a few of these lessons, even Natalie remarked to herself, almost puzzled, "Hm, I guess you just don't get this piece."

And so I left that day and bought Richter's recording. Undoubtedly a musical legend, Richter had for whatever reasono never really connected with me personally (or with Natalie). For that reason, I often chose his recordings of the pieces I played to listen to. They were guaranteed not to stray far from the score, they'd give me a different perspective, and they wouldn't tempt me to just copy what he did. But the next day at the piano, I decided to try an experiment. I put on headphones and I played along with Richter.

It's one thing to play along with a Mozart concerto that for the most part rolls along with the metronome, another altogether to try to play along with a lyrical Rachmaninoff prelude, which leaves so much room for rubato, when the performer might take some extra time here and give it back there. Luckily, Richter plays this prelude straighter than many pianists, but even so, it took a lot of practice to match the timing exactly, and then the dynamics and the voicing and the pedaling and so on. Of course the sound would never be 100% right, by the end of the week, I could play the piece with Richter note-for-note.

I had no rationale for the musical choices that were made. I didn't think about it at all. I just tried to play exactly like Richter. I figured, if Natalie still didn't like it, then she was saying she didn't like Richter either. In hindsight, it was a bit like Artificial Intelligence's Chinese Room argument. If my "output" was the same as Richter's, would an educated listener be able to tell that there was no substance backing it up, or would it just sound beautiful?

When I finished playing for Natalie the following week, she immediately exclaimed, "Wow! You really got it!"

That moment was the most ashamed I ever felt in a piano lesson, and for the rest of her life, I was never able to tell her how I had cheated.

There's an argument that could be made that I shouldn't have felt ashamed, and that making beautiful music is the only thing that matters. There's nothing to feel bad about for playing like one of the greatest pianists ever, some might say. At one point I might have been convinced of that logic, but the feeling of shame I felt from that compliment put that argument to rest for me. People speak of plagiarism in academia and forgery in art, but regardless of whether "great artists steal", there should be a parallel concept in music. There's already a way to hear Richter's recording, and if you say it doesn't feel the same as coming from a real piano, I imagine that's just a temporary technological shortcoming. That's not the why of music-making.

After that lesson, I "finished" that piece. I did tape myself playing it a couple months later – I backed off of Richter's interpretation a bit, but it's still there, and I never did feel like I really "got" this beautiful prelude.

When I was 22, fresh out of college, I found myself, for the first time in 19 years of studying piano, without a teacher. So, I chose a piece that I had grown up listening to and always wanted to play, but never had the time to learn with deadlines for performances, juries, etc. – the Ravel Ondine.

Natalie's Ondine was the Ondine that I grew up listening to. The first CD I was given was Natalie's own, one of the first ever four-channel recordings, with music from as early as 18th century Rameau and ending with the 20th century Prokofiev Toccata. But the gem for me growing up was always the Ravel Ondine, and I've never lost my bias for her performance, even among greats like Argerich, De Larrocha, Michaelangeli.

So here was a piece that I had an overall picture of before I looked at the music, but which I had never taken apart, and I was completely on my own. How would I approach it?

After the initial stage of getting the notes, I managed to hold my discipline for a bit and do a bit of analysis. I took apart the themes and motifs, mapped how they developed, and came up with potential ways to map Ondine's story to the music. I was fairly happy with the way I played. But I wanted more and better and I gave in. I watched every masterclass I could find on YouTube. I listened to recordings from Perlemuter and from a dozen different Juilliard and Curtis students in piano competitions. As I once heard Menahem Pressler say in a masterclass, "Be like a bee. Take from everywhere you can, and then produce your honey." Like so many amateur pianists today, YouTube became my teacher. I did listen with intent, and justified everything I stole, but I stole a lot. I took something I liked from Pogorelich, something from Casadesus, something from Lucas Debargue. Or at least I think I did. Nobody cites their sources during a piano performance.

Some of these ideas were undoubtedly better than the ones I had before. But it still didn't feel right. The biggest compliment I had ever received as a pianist was that my playing was authentic. What was it now? Was it still me?

In all likelihood, the "me" was still in there, and more prominent than I thought it was, but I could no longer sort out what it was and what it wasn't. Add to this that Ravel didn't even want his music "interpreted" in the first place, just played "straight", and I was thoroughly confused. I turned the corner when I decided to send a copy of my performance to a friend, and ended up listening to one more recording – my own. I spent hours going over my own performance and analyzing it and changing it and re-recording it. Finally, this last pass clarified for me what exactly I was trying to do with the piece and what every accent, every rubato, and every note was in service to.

I've played the piece often since I made that first recording on a practice piano at my office late one night, and it evolves with every performance I give of it. Ondine's story tells of a water nymph who, like a Siren, attempts to lure a mortal into her kingdom underwater. Finally, the mortal speaks, and says they love someone else and will not follow. Ondine cackles and disappears. Depending on the piano, the audience, and how I'm feeling, it may be quicker, darker, angrier, or more seductive. And if I'm in the mood to listen to Ondine, I still usually listen to Natalie's recording, not my own, despite everything I put into it. But I believe that my performance of it has a place in this world. That someone I've played it for connected with something they hadn't before, and that someone will again.

Since that day, I've relied more and more on recording and analyzing my own practice sessions and less and less on listening to other pianists. Before starting work on the Goldberg Variations, I forced myself to stop listening to recordings of them entirely for months, lest Gould find his way back into my head. But those great recordings are as seductive as Ondine, and it's difficult to imagine coming of age as a pianist in the YouTube era. How do you have the discipline to listen to heed your teacher's advice on a given Beethoven sonata when you have at your fingertips the opinions of Schiff, Goode, Kempff, Pollini, Kovacevich, Frank, Arrau, Badura-Skoda, Brendel, and Schnabel, to name but a few – let alone the young pianists who have played them in winning campaigns in the Cliburn competition, or the Tchaikovsky, or the Queen Elisabeth, or the Leeds, or the Rubinstein.

Some musicians bemoan the ubquity of recordings as a major factor in a perceived lack of originality and variety in music-making today. I think there's probably some truth to that. When recordings are used as SparkNotes to fast-track an analysis, certain potential interpretative paths are necessarily bypassed. Further, when certain recordings are as revered as the ones mentioned above, it takes a substantial amount of courage to go against the grain and say, "I disagree with Brendel here." Works like the Beethoven sonatas comprise the pianist's bible, and to dare to disagree means to risk the admonishment of a teacher saying "that's not what Beethoven means by this marking." The result is a student who studies the sacred text, and in the words of Madeleine Thein's character Zhuli, "plays Beethoven as if he had never been alive."

The irony, of course, is that those great recordings have no shortage of disagreements with each other, and any pianist who's played a Beethoven sonata for more than a couple different teachers has found themself with directly contradictory advice. In tech, we talk of standing on the shoulders of giants to reach greater and greater heights. In music, our giants loom so large that we can never get to their shoulders in the first place. It's simply not the case that a pianist can take Richter as a starting point and then always be playing at least as well as him.

The music student today simply cannot avoid the great figures of the musical past, and must instead have a strong enough sense of self to stand tall when she is inevitably confronted with and compared to her heroes. This, YouTube cannot teach.