Episode 22 - Artificial

welcome to Show 22 . Thank you for taking the time to listen . If this is the first time you've heard the podcast , I hope you enjoy what you hear . Let me know if you're learning the piano like I am , or another musical instrument . Let me know . If you're learning the piano like I am , or another musical instrument , let me know how you're finding it .

You can contact me at david at pianofinalyshow and if you're a returning listener . Thanks very much for coming back . Here's a bit of follow-up from the last episode . I'm still thinking about the performance request . As I said , it's not for another nine months , so I don't have to rush into anything .

I've let the students know that I'm considering it and they've given me a list of five possible Adele songs they want me to look at . Only one is needed for the performance . I've now got the PVG sheet music for their preferred number Someone Like you . It doesn't look too difficult , but it's well beyond where I'm up to at the moment .

But then there are nine months for improvement to happen . I'll have a talk with my piano teacher at the next lesson .

I think the YouTube algorithm is getting to know me better . This week's YouTube review is of a channel that was in my YouTube suggestions . The Music Professor channel is a collaboration between Professor Matthew King and Ian Coulter . Professor King is a composer of orchestral music , opera and chamber music . Ian is also a composer , working in musicals and television .

He also directs and produces the channel . Poser working in musicals and television . He also directs and produces the channel On camera . You'll see Professor King , who has worked on BBC radio , so his way of presenting material is clear and straightforward .

In the videos I have watched he is always at a piano or a keyboard so that he can demonstrate musically the points he is making . I haven't seen Ian on camera in any of the videos , but he often contributes from behind the camera , adding to Matthew's points .

The point of the channel , according to the notes on YouTube , is to ignite a passion for music in viewers with a mix of stimulating musical content and a tiny spot of humour . They seem to manage that quite well Together .

The two have some interesting and perhaps controversial opinions on some topics , but the videos are interesting and I've learned at least one new thing in each episode . I've watched their most recent video asked one of the large language model bots to compose a piece in the style of Mozart .

Matthew then looked at it with the eye of a composition professor and was able to show just how lacking this technology is for music composition at the present time . I should perhaps lacking this technology is for music composition at the present time .

I should perhaps mention that there is a third presenter on the channel , matthew's dog Loki , who is often somewhere in frame during his presentations . If you want some interesting , reasonably high-level discussion on classical music and its use in the modern world , I would recommend checking out the Music Professor YouTube channel ,

artificial . During the past week , I've had a fair bit of interaction with what is being marketed as artificial intelligence . I say marketed as AI because actual AI doesn't exist . What is being marketed as AI is really just machine learning , and so for the rest of the essay I'm going to refer to it as that . So how has it come up over the past week ?

Firstly , I have two groups of students competing in the upcoming Ethics Olympiad competition , and two of the eight cases in the competition are based around machine learning algorithms and their interactions with people , so we've been having in-depth discussions about it in one actual and one theoretical scenario .

Then Apple released the first versions of its operating systems with Apple Intelligence to the public and a much more powerful next version to developers . Finally , I've been teaching students how to properly use large language models such as ChatGPT to help with their research and not just copy and paste a possibly completely made-up response .

This got me thinking about machine learning and music . Is it a good or a bad thing ? Like the scenarios in the Ethics Olympiad cases , the answer depends on many other factors , so it's not simple . Is it a good or a bad thing ? Yes , as I told my students about ChatGPT , the service itself is neither good nor bad . It all comes down to how it's used .

The same is true of machine learning in music . Let's look at the bad to begin with . Machine learning is a complex mathematical process in which information is encoded into small chunks called training data . The training data are then used to build a model that can take a new chunk and produce output based on the output of the training data .

It's not making something new . It's simply fitting some new chunk into an existing model made from the other chunks . The bad comes with the creation of the model in the first place . The chunks of data that go into building the model have to come from somewhere , and largely that's going to be material available on the internet .

Of course , some person probably created that material , but the modeling process just takes the material and incorporates it without attribution or payment to the original creator . Is this any different to one composer listening to the works of another composer , taking inspiration from them and then using that inspiration to create their own original work ?

Yes and no , while in both cases prior work by one person is used to inform a second work . In both cases , prior work by one person is used to inform a second work . In the case of the human composer , it's usually one artist taking inspiration from many .

In the case of machine learning , this moves to an industrial scale , with millions of items being mined indiscriminately and then made available to others who've never even heard of the originator . In this case , machine learning is copying creativity . There is nothing new being added .

If the process continues long enough , it will be machine learning algorithms just absorbing other machine learning algorithms , and like taking a photograph of a photograph of a photograph , eventually all that will be left behind is a bland grey blob .

So how can machine learning be a good thing for creativity when it is used as a tool to allow someone to express their ideas in a way they would not otherwise be able to . I'm not a composer , but I think it's something I might eventually want to try my hand at .

If I do , then I would have the problem that I might want to include other instruments in my compositions but I don't play them . For a long time , sample libraries have been available that work with digital audio workstations . They can imitate other instruments , but they don't really sound like a person is playing the instrument .

Some of the newer software now uses machine learning to make the sampled instruments play much more like a real player . That's not to say that the final product should include anything other than real players , but if I want to try something out , a machine language player is going to be a lot more flexible and cheaper than using a real orchestra .

Similarly , if I was just getting started making my own tunes , having a tool like the drummer that is in GarageBand or Logic Pro would mean that I could experiment without also needing to understand a lot about drumming .

If used correctly , a machine language model can open a world of possibilities for new creators , and that can only be a good thing , provided that the models have been trained ethically and the artists whose work has been used are properly acknowledged and compensated .

I was first introduced to George Winston's music by the Alamo Piano Gallery's YouTube channel . Ted the presenter there spoke about working with George Winston and played some examples of his piano compositions . I thought they sounded interesting , so I went looking for material . Of course , as a beginner , the works are far too complex for me to consider now .

But then I found the music book that I'm reviewing today George Winston Easy Piano Sheet Music Collection . Now , easy is a relative term and the pieces are still beyond my current level . Well , they are as whole pieces In little segments . Some parts are manageable . There are 15 pieces in the collection , ranging from the early 70s to the start of this century .

The book starts with a short essay about the pieces written by George , and each piece has a few words about its origin or inspiration . George's essay includes permission to interpret the arrangements freely , as he says . That's what he does himself .

Although the book doesn't come with links to recordings of the pieces , all of them are on sheetmusicdirectcom and can be played through their rather low-quality audio engine . It's much easier to find full recordings of the original complex pieces on all the streaming services .

If you like George Winston's piano music and you're not up to the level you need to play the full versions , all of which are available in other sheet music collections then this easy collection is a great book to have . Eventually I'll get around to learning some of the pieces , just not quite yet .

Well , that's it for this week . If you'd like to contact me , email is the best way . You'll find me at david at pianofinelyshow and the website at wwwpianofinelyshow . In both cases , piano , finally , is all one word . Let me know where you are on your piano journey . This show is also on Facebook and Instagram and it's available audio only on YouTube .

Where you are on your piano journey ? This show is also on Facebook and Instagram and is available audio only on YouTube . You can subscribe via any popular iOS or Android podcast application or from directories such as Apple Podcasts , spotify or Player FM . So until the next episode , I hope your piano stays in tune and you enjoy your time at the keys .

This week I've been working on a Canon Not Bach's Crab Canon which you hear a snippet of throughout the podcast , although I do plan to eventually record a version of me playing it on the piano keyboard rather than on the computer keyboard I use for the current version . No , this canon is a much simpler one by Fritz Spindler .

It's from one of the early Ansgar piano exam books . The piece is 16 bars long and I originally broke it up into four bar segments to learn the notes . Because of the nature of a canon , this means that at the end of each fourth bar you are halfway through one of the voice's themes . So I also started with four bar groups , starting at bars 3 , 7 and 11 .

In the end it made it all come together much better when it came to playing the whole piece . I'm not completely happy with it yet , but it's coming along nicely . The recording was made using the Kawaii NV-10 as the keyboard and Piano Tech 8 running on the Mac Mini .

Piano Tech is set up with a Bosendorfer 280 VC in player mode with a custom velocity curve to match the NV10 . And he's Pete and he's Peach and he's

and he's Pete and he's Peach and he's and he , oh dear .