SoCal
Joined April 2013
- Tweets26
- Following76
- Followers69
- Likes59
Kehang Han retweeted
Introducing Magenta RealTime 2 (MRT2): the live music model you can play as an instrument.
MRT2 offers MIDI and prompt controls, and runs natively on a MacBook with <200ms latency.
Open weights. Open source inference engine. Suite of apps and plugins.
Hear what it can do and try it out for yourself below 🧵
Kehang Han retweeted
We built a web app that lets you fly a spaceship through a 3D constellation of music - powered by our Lyria RealTime model. 🎶
Space DJ is an interactive visualization where every star represents a different music genre. As you explore, your path is translated into prompts for the API, creating a continuously evolving soundtrack. ↓
Kehang Han retweeted
Made a walkthrough vid for Magenta RealTime “Audio Injection”! The notebook takes ~10m to spin up, but totally worth it for the surreal experience 🎤💻🎧⁉️
Excited to announce our latest Live Music model -- Magenta RealTime. You can stream it, play an instrument along with it and even customize it.
Colabs: github.com/magenta/magenta-r…
Tech report: arxiv.org/abs/2508.04651
Kehang Han retweeted
NeurIPS acknowledges that the cultural generalization made by the keynote speaker today reinforces implicit biases by making generalisations about Chinese scholars. This is not what NeurIPS stands for. NeurIPS is dedicated to being a safe space for all of us. We want to address the comment made during the invited talk this afternoon, as it is something that NeurIPS does not condone and it doesn't align with our code of conduct. We are addressing this issue with the speaker directly.
NeurIPS is dedicated to being a diverse and inclusive place where everyone is treated equally.
Kehang Han retweeted
Mitigating racial bias from LLMs is a lot easier than removing it from humans!
Can’t believe this happened at the best AI conference @NeurIPSConf
We have ethical reviews for authors, but missed it for invited speakers? 😡
Kehang Han retweeted
If you are at NeurIPS, come chat with us at West Ballroom A-D #7307
Paper (updated): arxiv.org/abs/2406.04329
Code: github.com/google-deepmind/m…
5mins video intro: neurips.cc/virtual/2024/post…
We have released code for our paper "Simplified and Generalized Masked Diffusion for Discrete Data" — SOTA discrete diffusion results — beating prior diffusion language models & exceeding AR likelihood on pixel-level image modeling. Try it out: github.com/google-deepmind/m…
Kehang Han retweeted
Discrete diffusion models made simple & competitive on both language and pixel-level image modeling! arxiv.org/abs/2406.04329
✅New variational objective (integrate cross-entropy!)
✅Beating prior diffusion language models & matching best AR on pixel-level image modeling
...(1/n)
Kehang Han retweeted
Today, we announced 𝗥𝗧-𝟮: a first of its kind vision-language-action model to control robots. 🤖
It learns from both web and robotics data and translates this knowledge into generalised instructions.
Find out more: dpmd.ai/introducing-rt2
Kehang Han retweeted
The phenomenal teams from Google Research’s Brain and @DeepMind have made many of the seminal research advances that underpin modern AI, from Deep RL to Transformers. Now we’re joining forces as a single unit, Google DeepMind, which I’m thrilled to lead! dpmd.ai/announcing-google-de…
Kehang Han retweeted
Today we’re releasing a new collection of tasks, templates and methods for instruction tuning of #ML models. Training on this collection can enable language models to reason more competently over arbitrary, unseen tasks. Learn all about it at: goo.gle/3XWlqjc
ALT Comparing public instruction tuning collections on held-in, chain-of-thought, and held-out evaluation suites, such as BigBench Hard and MMLU. All models except OPT-IML-Max (175B) are trained by us, using T5-XL with 3B parameters. Green text indicates improvement over the next best comparable T5-XL (3B) model.
Kehang Han retweeted
How good of a BERT can one get in ONE DAY on ONE GPU?
With all the recent studies about scaling compute up, this paper takes a refreshing turn and does a deep dive into scaling down compute.
It's well written, stock full of insights. Here is my summary and my opinions.
🧶 1/N
Applications are open for Google’s PhD Fellowships in Latin America. Selected Fellows receive $15,000 of funding for up to three years, and mentorship from a Google researcher to support their career development. Learn more and apply by January 11, 2023 at research.google/outreach/phd…
Kehang Han retweeted
Practical deep learning systems need to be more reliable, know what they don't know, and make good decisions under uncertainty: I'm very excited about our new work introducing Plex, a general framework for more reliable DL, applicable to vision, language and many other domains!
In our work "Plex", we propose a framework for reliability in AI. We also introduce new models (ViT-Plex & T5-Plex) for reliable decision-making across a broad array of scenarios.
Blog: ai.googleblog.com/2022/07/to…
Paper: goo.gle/plex-paper
Code: goo.gle/plex-code
Kehang Han retweeted
Today we describe a framework for reliable deep learning, which includes tasks and datasets to stress-test model reliability. We also introduce Plex, a set of pre-trained large model extensions that can be applied to architectures to improve reliability. goo.gle/3O6eJ8O
Kehang Han retweeted
We’re back with more researcher recommended resources available to use #AtHomeWithAI. Today is the turn of research engineer @KerenGu! (1/5)
I came into Prof. Winston's 6.034 class in 2015 almost "by accident". But later chose to stay and got so much more than AI - his way of teaching, his style of humor, his wisdom of life. So it became one of my best experiences at MIT. Thank you Prof. Winston. RIP.
We're saddened to report that Professor Patrick Winston, who directed MIT's AI lab for 25 years, died today.
Since being a PhD student of Minsky in 1965 he taught thousands of students and did pioneering work on AI that can tell and understand stories: bit.ly/PHWobit
