|September 12 · Issue #95 · View online |
Hey and welcome to a new week in deep learning!
Happy reading and hacking!
If you like receiving this newsletter and would like to support our work, you can do so by sharing this issue with friends and colleagues who might find it interesting. Thanks!
| Beneath AI hype is likelihood it threatens the poor says former Microsoft exec |
This article takes a look at Kai-Fu Lee’s
words about AI’s influence on low level jobs, which may be replaced. These fuelled quite a harsh discussion, but given the possible implications there probably can’t be enough of those.
| Huawei’s GPU Turbo: Valid Technology with Overzealous Marketing |
Huawei made great promises about their new ‘AI Turbo’ and are finally coming out with details. It’s based on TensorFlow and uses individually trained neural networks to optimise GPU performance for certain titles.
| Skydio Autonomy Platform brings R1 self-driving drone smarts to devs |
While autonomous driving still isn’t as great as we’d like it to be, the folks at Skydio have built an autonomous drone platform and now open up that platform to developers.
| Why Google's Inclusive Images Challenge Is So Important |
Google introduced the new ’Inclusive Images Challenge
’ trying to counter the general tendency to train models on images from Western Europe and the United States. This article explains why this is such an important task.
| Anatomy of an AI System |
This essay is hard to explain, as it’s actually a collection of 21 essays, which become an anatomical map of an AI system when combined. Definitely worth a read, but really hard to sum up in this short snippets.
| Open Machine Learning Course |
A very promising new machine learning course, aiming to balance theory and practice, created by OpenDataScience and spiced up with combined Kachle competitions.
| Train Donkey Car in Unity Simulator with Reinforcement Learning |
Really entertaining explanation of reinforcement learning using Unity simulator. The goal is teaching a so called Donkey Car, an open source RC car platform, to drive on a road.
| Dataset Search |
Handy new tool from Google, indexing thousands of datasets from different repositories. Might allow you to quickly find the right dataset to get started.
| The What-If Tool: Code-Free Probing of Machine Learning Models |
An interesting new Tensorboard extension which allows you to poke around in your machine model without touching your code (and introducing new bugs).
| Embracing Machine Learning as a Mobile Developer |
Helpful introduction to machine learning for mobile developers giving hints on which libraries to use and which courses or books might be a helpful start.
| Multimodal Unsupervised Image-to-Image Translation |
| Inclusive Images Challenge |
This challenge, mentioned in the article above, tries to counter the general tendency to train models on datasets collected in certain regions of the world by geographically splitting train and test sets. This will hopefully lead to more general models, which do not inherit a local bias from their training data.
| Deep image reconstruction from human brain activity |
Fascinating paper that takes the analysis of human functional magnetic resonance imaging patterns to the next level by tweaking images to mimic human brain activity using deep neural networks.
| Full-body High-resolution Anime Generation with Progressive Structure-conditional Generative Adversarial Networks |
And last but not least, some fun with really impressive anime generation and pose sequence based animation.