←
Daniel Arteaga (@dnlrtg.bsky.social)243 posts
Daniel Arteaga (@dnlrtg.bsky.social) avatar

Daniel Arteaga (@dnlrtg.bsky.social)

@dnlrtg

From now in Bluesky: bsky.app/profile/dnlrtg.bsky…

◷ Joined December 2018▧ 21 Photos and videos
Daniel Arteaga (@dnlrtg.bsky.social) avatar
Daniel Arteaga (@dnlrtg.bsky.social)@dnlrtg·15 Jun 2024

I almost fully agree with @geoffreyhinton's interview, though I think AI's chance of surpassing human intelligence in the next 20 years is much lower than 50%. However, regardless of whether the probability is 50% or 0.1%, the risk exists and needs to be studied and prevented.

“There is a 50-50 chance AI will get more intelligent than humans in the next 20 years. We’ve never had to deal with things more intelligent than us. And we should be very uncertain about what it will look like.” ~ Geoffrey Hinton

14:41
Daniel Arteaga (@dnlrtg.bsky.social) avatar
Daniel Arteaga (@dnlrtg.bsky.social)@dnlrtg·16 May 2024

Here are some usecases: - Ambisonics decoding to arbitrary irregular layouts - Transcoding multichannel to Ambisonics - Decoding multichannel to arbitrary irregular layout. - Decoding object-based formats to arbitrary irregular layouts -...

Daniel Arteaga (@dnlrtg.bsky.social) avatar
Daniel Arteaga (@dnlrtg.bsky.social)@dnlrtg·16 May 2024

We relased the open-source implementation of USAT on GitHub. Check out the code to optimize audio conversion for various formats and speaker layouts based on the optimization of a cost function with psycho-acoustic grounds. Check it out: github.com/DolbyLaboratories…

Daniel Arteaga (@dnlrtg.bsky.social) avatar
Daniel Arteaga (@dnlrtg.bsky.social)@dnlrtg·16 May 2024

Check out our latest paper on the Universal Spatial Audio Transcoder (USAT). We tackle the conversion & decoding of spatial audio formats, based on psycho-acoustic principles. USAT method adapts to any input/output format & speaker layout. Paper: arxiv.org/abs/2405.04471

Daniel Arteaga (@dnlrtg.bsky.social) avatar
⟳ Daniel Arteaga (@dnlrtg.bsky.social) retweeted
Daniel Arteaga (@dnlrtg.bsky.social)@dnlrtg·26 Feb 2024

Happy to share results of the Dolby Barcelona Scientific Paper Award 2023! This year, we were impressed by the quality of submissions, so we recognized not only the main winner but also present a runner-up prize. Both winning papers delve into the realm of AI in sound.

Post attachmentPost attachment
Daniel Arteaga (@dnlrtg.bsky.social) avatar
Daniel Arteaga (@dnlrtg.bsky.social)@dnlrtg·26 Feb 2024

Pablo's paper investigates using transformer models for music representation learning in supervised settings. Pablo and his coauthors show that it is possible to obtain better performance than with convolutional models trained in the same data. Code -> github.com/palonso/MAEST/

Daniel Arteaga (@dnlrtg.bsky.social) avatar
Daniel Arteaga (@dnlrtg.bsky.social)@dnlrtg·26 Feb 2024

Eloi's paper introduces a diffusion approach to audio inverse problems, particularly in piano music restoration. What sets this method apart is its ability to work without prior knowledge of degradation. Code -> github.com/eloimoliner/CQTdi… Preprint -> arxiv.org/abs/2210.15228