Ian Simon

@iansimon

Gum-chewing nerd & Magentan.

San Francisco
Vrijeme pridruživanja: veljača 2008.

Tweetovi

Blokirali ste korisnika/cu @iansimon

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @iansimon

  1. Prikvačeni tweet
    16. ruj 2019.

    We ( + + + + ) have now released a Colab notebook where you can play around with some Transformer models for generating piano performances: And the models are trained on transcribed YouTube recordings!

    Poništi
  2. 2. velj

    My Uber driver was listening to two music streams at the same time, one from his phone and one from the radio.

    Poništi
  3. proslijedio/la je Tweet
    15. sij

    Differentiable Digital Signal Processing (DDSP)! Fusing classic interpretable DSP with neural networks. ⌨️ Blog: 🎵 Examples: ⏯ Colab: 💻 Code: 📝 Paper: 1/

    Prikaži ovu nit
    Poništi
  4. 24. pro 2019.

    The western Zodiac models high frequencies, the Chinese zodiac models mid frequencies, and Boomer/GenX/etc models low frequencies.

    Poništi
  5. proslijedio/la je Tweet
    12. pro 2019.

    Autoencoders for entire performances of expressive piano! did an awesome job this summer with her Magenta internship, and it was a blast being her research mentor.

    Prikaži ovu nit
    Poništi
  6. proslijedio/la je Tweet
    2. pro 2019.

    We built a really cool machine learning drummer that accompanies your piano in real-time!! This is DrumBot! It’s a web app, so it can come along to your next gig. 🎉 🥁 📝 Blog post: 🎧 Live action:

    Prikaži ovu nit
    Poništi
  7. 13. lis 2019.

    Two competing trends: 1) Only computer programmers need to do X (e.g. access files directly). 2) Everyone needs to be a computer programmer.

    Poništi
  8. 7. lis 2019.

    I'm convinced that absolutely no one who works on the iOS Photos app has two or more children.

    Poništi
  9. 30. ruj 2019.

    I wonder what the fixed points are of training the models on their own samples.

    Poništi
  10. 27. ruj 2019.

    I assume someone has tried training a modern neural network architecture to factor products of two large prime numbers, you know, just to see if it works. Right?

    Poništi
  11. 12. ruj 2019.

    Most video game puzzles:

    Poništi
  12. 21. kol 2019.

    Remember how in the early days of cell phones the Internet would be working fine and then you’d get just barely within range of a Wi-Fi network and everything would grind to a halt?

    Poništi
  13. proslijedio/la je Tweet

    New Episode with on Algorithmic Composition and the Music Transformer project from Google’s Magenta labs. Representation of music for ML, self reference and memory, and AI support for composers. Also the .

    Prikaži ovu nit
    Poništi
  14. proslijedio/la je Tweet
    17. srp 2019.

    Remember the ? We’re excited to release paper on Behind-the-Scenes design, , scaling it up, and dataset of 21.6M melodies from around the world! 📜 w/ Leon Hong

    Prikaži ovu nit
    Poništi
  15. 13. srp 2019.

    An app like Yo, but consisting entirely of read receipts.

    Poništi
  16. 8. srp 2019.

    Watching my kids try to use various user interfaces, it occurs to me that a lot more actions should be idempotent.

    Poništi
  17. proslijedio/la je Tweet
    8. svi 2019.

    Tune in tonight at 8:30 PDT to see do *something* with Piano Genie! I honestly have no idea what they have in store for the performance but this video gives some hints. Stream link: 

    Prikaži ovu nit
    Poništi
  18. proslijedio/la je Tweet
    2. svi 2019.

    Last summer, intern recorded MIDI of professional drummers to train MusicVAEs (GrooVAEs🥁). This awesome work is core Magenta Studio and was accepted to 🎉 Today we release the dataset, code and colab on our blog: 👀

    Prikaži ovu nit
    Poništi
  19. 10. tra 2019.

    I do not understand the difference between scrolling down in my twitter feed and scrolling to the top of my twitter feed.

    Poništi
  20. proslijedio/la je Tweet
    20. ožu 2019.

    Check out our new blog post on Coconet 🥥, the behind the Bach Doodle that’s live now! It’s a flexible infilling model that generates counterpoint through rewriting. , , , , ,

    Prikaži ovu nit
    Poništi

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.

    Možda bi vam se svidjelo i ovo:

    ·