Artificiality Bites 💊 Issue #43
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. Eet smakelijk!
📝 Interesting publications this week
- Chinese AI lab challenges Google, OpenAI with a model of 1.75 trillion parameters
8'
Wudao is a multi-modal model trained to tackle both text and image, having 150 billion more parameters than Google's Switch Transformers, and 10 times bigger than OpenAI's GPT-3. - Session-based Recommender Systems
40'
This report explores a simple NLP-based approach to recommend a next item to a user, solely based on a user’s interactions in an ongoing session.
- Extending Contrastive Learning to the Supervised Setting
8'
Google proposes a novel loss function called SupCon, which enables contrastive learning to be applied in fully supervised learning. - Extrapolating to Unnatural Language Processing with GPT-3’s In-context Learning
24'
In this publication, Stanford AI Lab explores some surprising and not-so-surprising facets of in-context learning. - PyTorch builds the future of AI and machine learning at Facebook
21'
Facebook announced that that they’re migrating all their AI systems to PyTorch.
🔧 Tutorials
- Drug Discovery Using Machine Learning and Data Analysis
1h43'
📹
Learn how to use Python and machine learning to build a bioinformatics project for drug discovery. - JAX for the Impatient
21'
A notebook covering the basics of JAX. - Contrastive Representation Learning
51'
The main idea of contrastive learning is to learn representations such that similar samples stay close to each other. This article is a thorough publication on this topic.
📦 Repositories
- neuralmagic/sparseml
SparseML is a toolkit that includes APIs, CLIs, scripts and libraries that apply state-of-the-art sparsification algorithms such as pruning and quantization to any neural network. - giswqs/leafmap
A Python package for geospatial analysis and interactive mapping with minimal coding in a Jupyter environment. - facebookresearch/PyTouch
PyTouch is a machine learning library for tactile touch sensing. - jtpio/jupyterlite
JupyterLite is a JupyterLab distribution that runs entirely in the browser built from the ground-up using JupyterLab components and extensions.
- fchollet/deep-learning-with-python-notebooks
Jupyter notebooks for the code samples of the book “Deep Learning with Python” by François Chollet. - google-research/byt5
ByT5 is a tokenizer-free extension of the mT5 model. Instead of using a subword vocabulary like most other pretrained language models (BERT, XLM-R, T5, GPT-3), the ByT5 model operates directly on UTF-8 bytes, removing the need for any text preprocessing. - csjliang/LPTN
Official implementation of the paper ‘High-Resolution Photorealistic Image Translation in Real-Time” (CVPR 2021). - JoelForamitti/agentpy
AgentPy is an open-source framework for the development and analysis of agent-based models in Python. - nocodb/nocodb
Turns any MySQL, PostgreSQL, SQL Server, SQLite & MariaDB into a smart-spreadsheet, like Airtable.
🎓 Courses / Events
- Mathematical Tools for Data Science (NYU)
This course taught by Carlos Fernandez-Granda at New York University this year provides an introduction to mathematical tools for data science drawn from linear algebra, Fourier analysis, probability theory, and convex optimization. - Deep Learning Specialization (Coursera)
A new specialization from DeepLearning.ai. Develop practical skills to deploy data science projects effectively and overcome machine learning challenges using Amazon SageMaker.
🚀 Extra bits
- Nvidia is making a 3D map of the universe with the world’s most powerful AI supercomputer
3'
- EU faces €5–10 billion investment gap on AI and blockchain
3'
- From Motor Control to Team Play in Simulated Humanoid Football
7'
- AI drone may have ‘hunted down’ and killed soldiers in Libya with no human input
7'
- Google’s Voice Playbook
25'
👋 See you next week!