A summary of top updates in the field of Artificial Intelligence (AI) for week 40 of 2020. For more regular updates and related discussions, join us in the Artificial Intelligence Community. As always, thanks for your support for sharing it with your friends and colleagues on social media.

Industry updates for week 40 of 2020

This section is for recent industry updates related to Artificial Intelligence.

  • Google’s AI Residency program: Through Google’s ‘My Path to Google’ blog series, Awarenesses Dieng, an AI resident in the Ghana office shares her life experience towards AI and Machine Learning. With this, Google has started to take applications for the year 2021.
  • Google announces AI platform Prediction: The Google cloud released a press statement that its AI Platform Prediction is available for usage. The platform has Kubernetes backend (GKE) and is ready to use domain for ML models. To get more details, visit here.
  • Facebook’s new NLP collaboration, RAG: The Retrieval Augmented Generation (RAG) is an open-source NLP model by Hugging Face. Which now happens to be Facebook’s collaborators. The model skips the re-training step and draws information to keep updated, followed by results by the seq2seq generator. Here is the entire news feed.
  • Palm-based authentication at Amazon: The large e-commerce giant, Amazon has developed a new biometric scheme for its customers to go cashless and cardless. The palm of the customer and his card would be scanned while entering the store and the remaining payment would be made via his palm thereafter. To know more, visit here.
  • The Batch, GAN Special Issue: This is a special issue based on GAN after deep learning.ai organized a virtual event on 30th September 2020. It also includes key insights by Ian Goodfellow on GANs and deep fake detection. Read more about GANs over here.
  • LinkedIn Open-Sources GDMix: The Generalised Deep Mixed Model is available publicly after its predecessor Photon ML. The model is efficient in terms of search personalization and recommendation systems. To know more about GDMix, read this.
  • Facebook’s wav2vec v2.0 open-sourced: The self-supervised framework that takes raw audio as input is available publicly and is expected to work efficiently on 10 minutes of speech. To read more, visit here
  • Twitter’s voice feature for iOS with transcriptions: Twitter is planning to add an extension to the voice tweet feature for iOS users through transcriptions to enhance accessibility. Here is the entire news feed.
  • Top list of research in Neural Architecture Search(NAS): The automation of architecture engineering is done using NAS. Here are the top research articles based on NAS.
  • Top AI blockchain projects: Here is a post on a list of well known AI Blockchain projects currently popular in the industry. 
  • Elon Musk on AI, our friendly destroyer: This is a podcast where Elon Musk is in conversation with the New York Times writer on Mission to Mars, Brain chips, and of course sustainable energy. 
  • State of AI Report 2020: This is an overview of new achievements, developments, and advancements in AI for the year 2020. 
  • Fire tracking and management via AI: The California firefighters are using AI to investigate and predict wildfire spreading to manage and reduce the damage. To know more, read this article.
  • Google’s MENGER: A Massive Large scale distributed Reinforcement Learning Menger reduces training time by 8.6 times the state-of-the-art implementations. To know more, read this article.

Articles for week 40 of 2020

This section is for recent Blog Posts/Tutorials related to Artificial Intelligence.

  • Neural Network upgrade: A key tip in building Machine Learning applications is trying to upgrade the type of neural network architecture like deep networks which could improve the performance of the overall system. Here is how you do it.
  • “less-than-one-shot learning”: The LO-shot learning developed by AI researchers at the University of Waterloo is an extended version to One-shot learning. To know more, read here.
  • Google’s enhanced Pho to editor
    : Google Photos is launching a new feature in its application that uses machine learning to beautify your memories. This technique learns one’s choices over the period to give better recommendations. Read more here.
  • Real-time sign language detection: This is the best solution so far for detecting an active speaker willing to converse in sign language at the cost of good quality detection and lightweight model development. This is achieved through Google’s research article “Real-time sign language detection using Human Pose estimation”. To read more, visit here.
  • Audio-visual speech enhancement in Youtube Stories: Google has been able to successfully apply the ‘Looking to Listen’ feature to Youtube stories. This feature captures visual and audio cues and isolates them from other noise and manages to synchronize them as well. To know more, visit here.
  • New MDP models for highways: The Markov Decision Process (MDP) is proposed for automated driving. The research takes into account the road shape and ignores the vehicle speed to reduce computational complexities. Here are the article details.
  • Images from Captions: The researchers from Allen Institute of AI have developed a visual language model, X-LXMERT that produces images relating to the input captions. Here is the entire news feed.
  • Bias and Variance estimation: A well written yet simple article on estimating bias and variance using python is explained here
  • Tensorflow Recommenders: If you are new to TensorFlow recommenders and wish to get a quick start, this one’s for you. Includes a quick guide and installation steps.
  • TinyML: an introductory tutorial on tiny ML is discussed along with few implementations and applications. To know more, read from this feed. 

Softwares

This section is for recent Tools/Applications related to Artificial Intelligence.

  • KerasGA: KerasGA is used for training Keras models and is a part of pygad specifically genetic algorithm. The Github code is available here.

Videos

This section is for recent Lectures/Conferences/Webinars related to Artificial Intelligence.

  • World’s First Virtual & Augmented Reality Laptop: The zSpace laptop series use AR and VR for encouraging experimental learning. Here is a demo of it in action.
  • Convert textbook to an audiobook: A new episode from ‘5 line series’ as to how to convert a hard book text to an audiobook in just 5 lines of code. Here is the working demo video.
  • Optimizer training and self-training: The paper is explained in this video for training optimizers with several thousand tasks obtaining generalization in unseen data.

Research

This section is for recent research published related to Artificial Intelligence.

  • Utterance level Dialogue Understanding: This study is based on highlighting the involvement of context in a conversation and its effect on utter-level dialogues. The research article is accessible here while the code is available here.
  • The mapping between Hopfield Networks and Restricted Boltzman’s Machine: Article download.
  • Transformers for Image Recognition: This article points out that CNN is not necessary for image classification, mere use of transfers with a pre-trained model on a large model will yield excellent results. Here is the full article.

Miscellaneous

This section is for updates for other various related topics to Artificial Intelligence.

That’s all for the week xx of 2020. In case you think we have missed out something and that is significant to add, please free to reach out and submit to us via our Contact form or via Social media channels: Twitter and Facebook.