Jorge Bustos

Machine Learning || Audio & Image

ARQUIMEA Research Center. Prev

[Jorge Bustos @bustos_jor]

@nasrinmmm Startups have a very specific market niche and are mainly based on existing Open Source(OS) technologies with a lot of potential that still need some refinement. And this is what we are experiencing today. Lots of new OS DL models that can be used for a wide variety of apps.

[Jorge Bustos @bustos_jor]

@dadabots That sounds sooo good. Count with me!

[Suhail @Suhail]

There is so much AI research momentum, I cannot remotely keep up. The pace is unreal.

There are gems and prizes in GitHub repos everywhere—mostly unnoticed.

[Jorge Bustos @bustos_jor]

@tunguz Deep Learning in a nutshell

[Jorge Bustos @bustos_jor]

Research Friday - One TTS Alignment To Rule Them All (https://t.co/YKEibn13OV) Easy but powerful idea. Once you read it, it seems obvious that you can use CTC for alignment learning. Especially, when it's a paradigm for STT models... How haven't we used it before for TTS? Smart!

[Jorge Bustos @bustos_jor]

Research Friday - CLAP : Learning Audio Concepts From Natural Language Supervision (https://t.co/oUcC3iWX8t) I have already read about CLIP but I am curious about how they approach it for audio.

[Jorge Bustos @bustos_jor]

@cyrta @csteinmetz1 I have never been on that situation. It sounds like there is no trivial solution. Maybe instead of doing it after the talk try to make it live with headsets. Although it does not sound
comfortable at all.

[Jorge Bustos @bustos_jor]

Someone has to say it. It sucks to make subjective evaluations of audio models.

[Jorge Bustos @bustos_jor]

@csteinmetz1 Jajajaja indeed. I had a much better time evaluating album artwork during my thesis. Imagine evaluating DALL-E... what a dream

[Jorge Bustos @bustos_jor]

@tunguz Insane. But what about git?

[Jorge Bustos @bustos_jor]

Does emphasis help to improve the audio quality of neural vocoders?

[Jorge Bustos @bustos_jor]

Nice adversarial attack

[Jorge Bustos @bustos_jor]

Reading this post reminds me how difficult is to keep up to date the sota... Many interesting works... Little time... And an incredible world out there. Overwhelming and beautiful at the same time!

[Jakub Tomczak @jmtomczak]

If you are looking for your niche in AI, here are five topics to avoid 😉
https://t.co/NIFuslBR1c
Very nice to see deep generative models (diffusion models) to be listed out.

[Jorge Bustos @bustos_jor]

- Civilization -

[Jorge Bustos @bustos_jor]

Musimathics by Dr. Gareth Loy. Highly recommended for advanced and beginner audio engineers and researchers.

https://t.co/Sm46nKYIBr

[Jorge Bustos @bustos_jor]

The closing screen of our questionnaire for MOS evaluation. Thanks and sorry for this @SchmidhuberAI

[Progress Bar 2026 @ProgressBar202_]

2021 is 90% complete.

[Jorge Bustos @bustos_jor]

I declare myself a fan of @avdnoord and @OriolVinyalsML

[Jorge Bustos @bustos_jor]

Still dont fully get the math behind this paper. But the idea is just brilliant. https://t.co/0QyKgFr1Vt