Piano For AI

👀

Our R&D team is speaking at PyData Global NYC 2024! 🎉 Meet us on stage on November 7th, 2024, from 16:00 to 16:45 (US/Eastern) represented by Tomek Roszczynialski and Wojciech Matejuk. In our talk, we will explore different strategies for teaching billion-parameter models to play the piano, applying everything we know about LLMs to the challenge of expressive and emotional music. We will demonstrate how methods inspired by masked language modeling, image inpainting, and next-token prediction can be adapted to preprocess MIDI data, capturing the harmonic, dynamic, and temporal information essential to music. We can't wait to see you there! Come and say Hi! 👋 Read more here: https://lnkd.in/dGtN3Rvi

Did you know that you can watch our latest presentation at PyData London 2024 online? 👀 Check out Wojciech Matejuk speech on "Modeling Emotional Nuance of Musical Performance." It's a really great introduction to what we do daily. 🎹 ➡️ https://lnkd.in/dZvyQHkk

It was awesome to have my conferential debut on PyData London 🇬🇧! I got to know so many amazing people and many dimensions and domains of data science. Attending conferences ultimately broadens the horizons of what problems lie ahead and showcases the innovative work others are doing! And they are a fantastic place to yap about your nerdy interests … I cannot express how grateful I am for the opportunity to attend some very interesting and absorbing talks and to speak about what I am passionate about. It was a pleasure to share how the beginning of the journey of the Piano For AI project looks like and to spark interest in algorithmic music composition 🎶. Huge thanks to Tomek and Aleksandra who helped me A LOT in preparing my presentation. And above all - thank you to all the organizers, volunteers, speakers, and attendees of PyData. I hope to be back soon! #PyData

We're working on a custom Streamlit component for playing and visualizing keyboard music - here's a dashboard where you can test it with data from the Maestro Dataset: https://lnkd.in/eHn3iuxk This component works seamlessly with MIDI files, DataFrames, and NumPy arrays. We're sharing this to gauge interest and gather feedback - let us know if this looks interesting! 🎹🎹🎹 And if you have a ticket for PyData Global in London this week, this component will be featured prominently in Wojciech Matejuk's talk: https://lnkd.in/ePMnW87Q

We're going to PyData London this week! 🇬🇧 Wojciech Matejuk will give a talk about his experience trying to embed emotional understanding into large transformer models. We believe that it's crucial to teach our future AI overlords compassion, and what better place to start than with a little Chopin? 🎶🎶🎶 Come join us! 👋

“The doors to the mathematical nature of music can be opened with Machine Learning.” While it's uncertain if music truly has a mathematical nature or any doors to be opened, our exploration of this concept amusingly led us to a different kind of 'doors.' During our investigation into the artistic process, we had the pleasure of recording a glorious piano cover of 'Riders on the Storm' by The Doors. The performance features Wiktoria Miłkowska, who marked the beginning of her music-oriented data-science internship with us through this recording. The full version of the song has been released on her YouTube channel. Go watch it! https://lnkd.in/dKfH4qip