Jon Saad-Falcon, Joe Barrow, Alexa Siu, Ani Nenkova, Ryan A. Rossi and Franck Dernoncourt propose an approach called PDFTriage that enables models to retrieve the context based on either structure or content.

Ars Technica - Combining Stable Diffusion and ControlNet, users have been able to create mesmerising Ai-artwork, such as QR codes earlier this year. This tiime, they take steganography a step further

LlamaIndex propose a tutorial to build the “LLM synthesis” component of a RAG pipeline from scratch.

Radamés Ajna proposes a tutorial to generate sentence embeddings visualize them after dimension reduction in the browser

Ivan Vendrov writes a short blog explaining why all popular consumer software seem to trend towards minimal user agency, infinitely scrolling feeds, and garbage content.

Zhengqi Li, Richard Tucker, Noah Snavely, and Aleksander Holynski present an approach to modeling an image-space prior on scene motion in 'Generative Image Dynamics'. The model is learned from real video sequences depicting natural dynamics and can predict a spectral volume that spans an entire video. It has various applicati...

Astronomy.com - Peer back through the history of the universe, all the way to its first microsecond, with the help of something that’s all around us : the atom.

Tianle Cai, Yuhong Li, Zhengyang Geng, Hongwu Peng and Tri Dao (together.ai) unveil Medusa, a simpler, more user-friendly framework for accelerating LLM generation.

Ziwen Zhuang, Zipeng Fu, Jianren Wang, Christopher Atkeson, Soeren Schwertfeger, Chelsea Finn and Hang Zhao propose a system for learning a single end-to-end vision-based parkour policy of diverse parkour skills using a simple reward without any reference motion data.

Quantamagazine - Physicists are now harnessing nonlocality to probe the spread of quantum information and control it.

Vgel walks you through the steps to make a transformer by hand

Big Think - Ethan Siegel explores the "speed" of gravity, a concept that did not even exist in the Newtonian approach

Chengrun Yang, Xuezhi Wang, Yifeng Lu, Hanxiao Liu, Quoc V. Le, Denny Zhou and Xinyun Chen propose Optimization by PROmpting (OPRO), a simple and effective approach to leverage large language models (LLMs) as optimizers, where the optimization task is described in natural language.

Clever article from Tim Hwang on how, sometimes, it makes a lot of sense to fund research to continue to explore what is deemed a dead end

Foundation Mozilla, as part of the "Privacy Not Included" initiative, reviewed a number of car brands for their privacy and security standards. And they were not impressed!

ChinaTalk - The ChinaTalk team puts Baidu's LLM chatbot to test

Wired - ChatGPT was just the start. OpenAi's ultimate goal is to change everything.

Chirp is a text-to-music model developed by Suno.

Jeremy Howard and Jonathan Whitaker from fast.ai conducted a series of experiments to validate and better understand why model appear to rapidly memorize examples from datasets after seeing them just once

Hrishi Olickel proposes a guide to prompt engineering

Jan Fijor writes about the failed implementation of central bank digital currencies (CBDC) in Nigeria.

Nature - JohnMark Taylor and Nikolaus Kriegeskorte introduce TorchLens, a new open-source Python package for extracting and characterizing hidden-layer activations in PyTorch models.

Futurism - a quadriplegic patient recovers movement and sensation in his arms thanks to AI-chips implanted in his brain

Gizmodo - An ancestral human species faced a startling population bottleneck and teetered on the brink of extinction around 800,000 years ago

Tom's Hardware - New restrictions imposed by the Biden administration on exports of advanced compute GPUs