WHATS IMPORTANT
📍 ICLR 2021 Conference
The landmark Deep Learning conference ICLR is taking place from May 3rd to May 7th. For a second time the gathering is held completely virtual. Researchers from around the world are gathering for poster sessions, keynotes and workshops. Out of 2997 papers, 860 were accepted for the conference. The official Outstanding Paper Awards have been awarded for 8 papers of "exceptional quality" leading up to the conference. You can find a full list here.
Meanwhile, Topbots made its own selection of ICLR conference papers, with "breakthrough potential". These include the Visual Transformer (ViT), DETR, DeBERTa and Performers papers we previously introduced in this Newsletter. Read more about each breakthrough paper on topbots.com.
📍 More to Come
ICLR only marks the beginning of the annual Conference Summer. With the global health situation only improving slowly, more conferences have moved to virtual meetings. Find an updated version of the AI Conference Calendar below. Find out more about the most important AI conferences here.
THINGS WE FOUND WORTH SHARING
🚀 Contribute – The crowdsourcing project Common Voice aims to create a free database for speech recognition in 50+ languages. Since 2017 Mozilla asks volunteers to record sample sentences and review recordings of others. They recently saw a USD 1.5 million investment by Nvidia and are continuously looking for more participants. Contribute here (switch the language in the top right).
📚 A personal recommendation from AMAI CEO Jürgen Stumpp:
"The book "Real World AI: A Practical Guide for Responsible Machine Learning" provides executives with a quick but thorough overview of the steps necessary for successful AI projects. Moreover, it can help soon-to-be university graduates to prepare for work in the field."
In their book, Alyssa Rochwerger, Director of Product at Blue Shield and Appen CTO Wilson Pang, share their practical experiences and present an approach that claims a 3x higher success rate for AI projects compared to industry average. Learn more about "Real World AI: A Practical Guide for Responsible Machine Learning" in this authors interview or in an open Webinar today at 20:00 (CET).
📄 Paper – DINO: In the paper "Emerging Properties in Self-Supervised Vision Transformers" (arXiv), researchers from Inria, Facebook AI and Sorbonne University introduced DINO. Short for "self-distillation with no labels", this method can segment images in a self-supervised manner, meaning without labels required upfront.
Read more about DINO and the accompanying PAWS on venturebeat.com and the Facebook AI Blog ai.facebook.com. One-man paper discussion group, Yannic Kilcher, walks his viewers through the paper in detailed 39 minutes on youtube.com.
📈 Markets – Microsoft acquired Nuance in April, a Massachusetts company focussing on AI-driven speech recognition. At USD 19.7 billion this merger marks the second highest acquisition in Microsoft's history (LinkedIn was acquired for 26.2 billion in 2016). Nuance Communications is behind the speech recognition capabilities of Apple's voice assistant, Siri. More on axios.com.
🧑⚖️ Regulation – The European Commission released the Artificial Intelligence Act, a 108-page proposal for regulation of AI. More on the proposal and the legislative hurdles ahead, on politico.eu.
Outside the EU, Chinese regulators have begun to enforce data localization on local companies. From now on data can only be stored in China for certain applications such as facial recognition (biometricupdate.com).
🤗 Tongue in Cheek – After XKCD posted 12 types of scientific papers, Natasha Jaques and Max Kleiman-Weiner, two PhDs from MIT, put together 12 Types of Machine Learning Papers.
👁 Miscellaneous – 15 Graphs You Need to See to Understand AI in 2021: In the last issue we shared the Stanford 2021 AI Index Report. From that Eliza Strickland takes 15 visualizations to highlight the most important developments. Scroll through on ieee.org.
🎓 Education – One of the best educational resources for AI is Andrew Ng's Deep Learning specialization on Coursera. The course just received the 2021 update. All programming exercises are now in Tensorflow 2 and the syllabus includes Transformer Networks. Find the updated course here coursera.org.
EVENTS
📅 May 6th (online, 20:00 CET) – As mentioned above, the authors of "Real World AI" host a Webinar today to discuss their approach boasting a 3x higher AI project success rate when compared to industry averages. – Register for free here.
📅 May 12 (online, 16:00 CET) – How AI turns existing Business Models Upside Down – "Join online to interactively hear some astonishing inputs about disruptive AI business models from the Harvard Post Doc Johannes" from Merantix Labs. Organized by KI-Garage fro the institute of Entrepreneurship and Innovation Research from Stuttgart University and our partner the German Digital Hub Initiative. – Register here.
