Artificiality Bites đź’Š Issue #51
Hello Human! This is a new issue from my weekly newsletter, holding a tiny compilation made of interesting articles from last week, projects, tutorials and tools; all related to Data, Artificial Intelligence and adjacent topics. Buen provecho!
đź“ť Interesting publications from last week
- Generally capable agents emerge from open-ended play
13'
Deepmind detailed their first steps to train an agent capable of playing many different games without needing human interaction data. - The Chatbot Problem
9'
"As we teach computers to use natural language, we are bumping into the inescapable biases of human communication". - What Have Language Models Learned?
11'
By asking language models to fill in the blank, we can probe their understanding of the world.
- Hundreds of AI tools have been built to catch covid. None of them helped
10'
Some of these tools have been used in hospitals, despite not being properly tested before. This pandemic could help make medical AI better in the future. - Thrill-K: A Blueprint for The Next Generation of Machine Intelligence
15'
Rethinking knowledge layering and construction for long-term viable AI.
đź’ˇ Projects
- Tour of the Sacred Library
16'
A brief story accompanied by images synthesized using a combination of machine learning models over the provided text, VQGAN for generation, and CLIP for directing the image to match the text.
🔧 Tutorials
- GPT-J-6B Inference Demo
A notebook that demonstrates how to run the GPT-J-6B model.
📦 Repositories
- deepmind/wikigraphs
This package provides tools to download the WikiGraphs dataset, collected by pairing each Wikipedia article from WikiText-103 with a knowledge graph. - deepmind/perceiver
Perceiver is a general architecture that works on many kinds of data, including images, video, audio, 3D point clouds, language and symbolic inputs, multimodal combinations, etc. Read more at DeepMind blog. - borisdayma/dalle-mini
A mini version of Dall-E, which generates images from any prompt.
- patrick-kidger/equinox
Equinox brings more power to your model building in JAX, by using a PyTorch-like class based API without sacrificing JAX-like functional programming. - fastai/fastdownload
Make datasets or other archives available for your users while ensuring they are downloading the latest version.
🚀 Extra bits
đź‘‹ Hasta la vista!