We had an opportunity to chat with Vladimir Rybakov (Head of #DataScience at WaveAccess) about: - solving problems under pressure, - taking responsibility, - why asking for help is an important skill and more! bit.ly/2yr9vTK

The most modern way to implement a GAN in Keras is actually simply this.

Introducing Jukebox, a neural net that generates music, including rudimentary singing, as raw audio in a variety of genres and artist styles. We're releasing a tool for everyone to explore the generated samples, as well as the model and code: openai.com/blog/jukebox/

This brilliant ⁦OpenAI⁩ work and the video of ⁦Andrej Karpathy⁩ I shared recently are very exciting AI frontiers. The story repeats itself: Big net, curated data, and common sense are the ingredients. Congrats ⁦Ilya Sutskever⁩ et al. arxiv.org/abs/2005.14165

Today we present SpineNet, a novel alternative to standard scale-decreased backbone models for visual recognition tasks, which uses reordered network blocks with cross-scale connections to better preserve spatial information. Learn more below: goo.gle/3dNq9fa

GPT-3 can't quite pass a coding phone screen, but it's getting closer.

Introducing a new approach to RL that uses complex duality to convert problems with a large number of constraints to equivalent, more computationally friendly forms, enabling mathematically principled algorithms that are useful in practice. Learn more ↓ goo.gle/2AEKW7h

Too l̶a̶z̶y̶ busy too keep up with reading the latest #DeepLearning papers? I just discovered this great channel that explains important papers in real depth 🎓 but is still great fun and easy to follow 🤡 youtube.com/channel/UCZHmQ…

What makes production ML hard? - Cleaning, labeling, and augmenting data - Troubleshooting training and ensuring reproducibility - Deploying models and monitoring their real-world impact To help, we're excited to announce our online production ML course: course.fullstackdeeplearning.com

Creating a system that can execute written/spoken instructions on a graphical interface requires a model that grounds the instructions to executable UI actions. Check out foundational work that does so through use of a multi-part language grounding model: goo.gle/3fo8qwQ

ACL 2020 #acl2020nlp ends this week! If you didn't manage to attend all #KnowledgeGraph related talks - I comprised a review of KG-related papers focusing on question answering, KG embeddings, graph-to-text NLG, some ConvAI and OpenIE 😊 #NLProc medium.com/@mgalkin/knowl…

AutoML-Zero: new research from Google AI researchers Esteban Real, Chen Liang, David R. So, & Quoc Le that that can rediscover fundamental ML techniques by searching a space of different ways of combining basic mathematical operations. Arxiv: arxiv.org/abs/2003.03384

1/ Crazy exp: take Resnet embedding of Imagenet as dataset A. Train linear predictor on A; get accuracy p. Now make fake dataset B = a mixture of Gaussians w/ same class mean & covariance as A. Train linear predictor on B => get *SAME ACCURACY* p. WTF proceedings.icml.cc/static/paper_f…

Madam (multiplicative adam) needs little to no learning rate tuning and brings the numerical representation of the synapse closer to neuroscience. w/ Jiawei Zhao, Markus Meister, Ming-Yu Liu, Prof. Anima Anandkumar & Yisong Yue paper: arxiv.org/abs/2006.14560 code: github.com/jxbz/madam

Suffering from post-#acl2020nlp withdrawal? There are also lots of Percy Liang, Tatsunori Hashimoto, and other Stanford people’s papers at ICM 2020 this coming week! ai.stanford.edu/blog/icml-2020/

I had a few samples from my latest GAN project printed on canvas, love how they turned out 😋👏 See more at vimeo.com/neuralsynesthe…

The Neural Rendering engines of the future are going to be insane 🤯👏👌

Open-source, high-quality texture synthesis and style-transfer using PyTorch! Amazing work by Alex J. Champandard 🌱 👏

We've used reinforcement learning from human feedback to train language models for summarization. The resulting models produce better summaries than 10x larger models trained only with supervised learning: openai.com/blog/learning-…

Java still try-harding to catch-up to python in the DS/ML field. Though it is good to have a scikit-like lib on Java just in case. blogs.oracle.com/javamagazine/p… PS Christoph Henkelmann - @[email protected], your thoughts? :)