This week in AI & Machine Learning: AI research to help predict COVID-19 resource needs, the White House Launches the National Artificial Intelligence Initiative Office, Amazon opens Alexa AI Tech, Implementing a Transformer with PyTorch, On-Device ML Study Jam, and more.
Artificial Intelligence News:
New AI research to help predict COVID-19 resource needs from a series of X-rays
Facebook AI has collaborated with New York University (NYU) Langone Health’s Predictive Analytics Unit and Department of Radiology to help predict how much care and resources a patient may need for COVID-19 treatment.
It’s challenging for doctors to predict the course of #COVID19 in a patient. In partnership with @nyulangone we’re open-sourcing AI models to help hospitals predict whether a patient’s condition will deteriorate and to help plan for resource allocation. https://t.co/vf3IvRt2bh pic.twitter.com/0cOymNaPsp— Facebook AI (@facebookai) January 15, 2021
The White House Launches the National Artificial Intelligence Initiative Office
We have been hearing increasing talk about how the U.S government should approach and support artificial intelligence (A.I.) development. This week, the White House Office of Science and Technology Policy (OSTP) has officially created the National Artificial Intelligence Initiative Office to oversee nationwide AI strategy, research collaboration, and policymaking.
Amazon opens Alexa AI Tech for the First Time:
Amazon opens its Alexa AI technology to third-party companies for the first time, allowing customized Alexa assistants to be embedded in virtually any platform, starting with automobile manufacturers. This is a huge move from Amazon and may spur an AI voice assistant boom across many industries.
Developer Tools & Education:
OpenCV Video Augmented Reality
In the latest pyimagesearch post, learn how to perform real-time augmented reality in video streams using OpenCV. This continues the computer vision + augmented reality series that has been going on for the past several weeks.
Implementing a Transformer with PyTorch and PyTorchLightning
Check out this interesting thread and Google Colab about using transformers and how to implement them in pytorch.
Recognizing Pose Similarity in Images and Videos
Google AI shows off its new algorithm, “View-Invariant Probabilistic Embedding for Human Pose” (Pr-VIPE) to recognize pose similarity with computer vision in images and videos.
Pr-VIPE is a new approach to pose perception that uses probabilistic embeddings that are view-invariant to avoid the ambiguity arising from the 2D projection of 3D poses. The model is simple and compact, and can be trained in ~1 day on CPUs. Learn more at https://t.co/OEQx5VAPAk pic.twitter.com/TIoE7cs8X6— Google AI (@GoogleAI) January 14, 2021
Interesting Podcasts & Interviews:
Trends in Graph Machine Learning with Michael Bronstein | TWiML
In the latest episode of AI rewind TWiML’s guest, Michael Bronstein will give his thoughts and predictions on the graph machine learning field.
Scaling Up Machine Learning | Super Data Science
Join Erica Greene on the super data science podcast to learn how to scale up machine learning, avoid model drift, set priorities for large ML teams, and more.
A Future of Work for the Invisible Workers in A.I. with Saiph Savage | TWiML
Saiph gave a talk at NeurIPS 2020 called “A Future of Work for the Invisible Workers in A.I.”. This conversation dives deeper into the roles and potential issues around people doing much of the labeling for machine learning.
Notable Research Papers:
- Provable Generalization of SGD-trained Neural Networks of Any Width in the Presence of Adversarial Label Noise