Tech News - Data

With more than 3 years from the time of previous version release OpenCV 3.0, the team happily announced the alpha release of the most awaited OpenCV 4.0. The new version is set to encompass exclusive features such as 3D dense reconstruction algorithm, newest improvements and bug fixes to recent maintenance release of OpenCV 3.4. Key Features: OpenCV is a C++11 library now Default features include lambda functions, convenient iteration, and initialization of cv::Mat Added new chessboard detector Added exclusive HPX parallel backend and basic FP16 support Standard std::string and std::shared_ptr replaced hand-crafted cv::String and cv::Ptr parallel_for can now use the pool of std::threads as backend Major improvements and bug fixes: ONNX parser added to existing OpenCV DNN module Support to various classification networks - AlexNet, Inception v2, Resnet, VGG Partial support to YOLO v2 object detection network Faster object detection using Intel Inference Engine, part of Intel OpenVINO Added stability improvements in the OpenCL backend Fast QR code detector with support to add QR code decoder soon SSE4-, AVX2- and NEON-optimized kernels expanded Legacy C API from OpenCV 1.x partially excluded This alpha release appears to be a massive version with 85 patches including 28 merge requests. This release is assumed to be quite stable although few changes in OpenCV API and implementation are expected before 4.0 final release. For more information on the detailed list of features and improvements, please read official documentation. Image filtering techniques in OpenCV 3 ways to deploy a QT and OpenCV application OpenCV and Android: Making Your Apps See

On Tuesday, Stanford University researchers introduced two recent datasets collected by the Stanford NLP Group to further advance the field of machine reading. These two new datasets CoQA (Conversational Question Answering), and HotpotQA work towards incorporating more “reading” and “reasoning” in the task of question answering and move beyond questions that can be answered by simple pattern matching. CoQA aims to solve the problem by introducing a context-rich interface of a natural dialog about a paragraph of text. The second one, HotpotQA goes beyond the scope of one paragraph and presents the challenge of reasoning over multiple documents to arrive at the answer. Lately, solving the task of machine reading or question answering is becoming an important section  towards a powerful and knowledgeable AI system. Recently, large-scale question answering datasets like the Stanford Question Answering Dataset (SQuAD) and TriviaQA have progressed a lot in this direction. These datasets have enabled good results in allowing researchers to train deep learning models What is CoQA? Most of the question answering systems are limited to answering questions independently. But usually while having a conversation there happens to be a few interconnected questions. Also, it is more common to seek information by engaging in conversations involving a series of interconnected questions and answers. CoQA is a Conversational Question Answering dataset developed by the researchers at Stanford University to address this limitation and working in the direction of conversational AI systems. Features of CoQA dataset The researchers didn’t restrict the answers to be a contiguous span in the passage. As a lot of questions can’t be answered by a single span in the passage, which will limit the naturalness of the conversations. For example, for a question like How many times a word has been repeated?, the answer can be simply three despite text in the passage not spelling this out directly. Most of the QA datasets mainly focus on a single domain, which makes it difficult to test the generalization ability of existing models. The CoQA dataset is collected from seven different domains including, children’s stories, literature, middle and high school English exams, news, Wikipedia, Reddit, and science. The CoQA challenge launched in August 2018, has received a great deal of attention and has become one of the most competitive benchmarks. Post the release of Google’s BERT models, last November, a lot of progress has been made, which has lifted the performance of all the current systems. Microsoft Research Asia’ state-of-the-art ensemble system “BERT+MMFT+ADA” achieved 87.5% in-domain F1 accuracy and 85.3% out-of-domain F1 accuracy. These numbers are now approaching human performance. HotpotQA: Machine Reading over Multiple Documents We often find ourselves in need of reading multiple documents to find out about the facts about the world. For instance, one might wonder, in which state was Yahoo! founded? Or, does Stanford have more computer science researchers or Carnegie Mellon University? Or simply, How long do I need to run to burn the calories of a Big Mac? The web does contain the answers to many of these questions, but the content is not always in a readily available form, or even available at one place. To successfully answer these questions, there is a need for a QA system that finds the relevant supporting facts and to compare them in a meaningful way to yield the final answer. HotpotQA is a large-scale question answering (QA) dataset that contains about 113,000 question-answer pairs. These questions require QA systems to sift through large quantities of text documents for generating an answer. While collecting the data for HotpotQA, the researchers have annotators to specify the supporting sentences they used for arriving at the final answer. To conclude, CoQA considers those questions that would arise in a natural dialog given a shared context, with challenging questions that require reasoning beyond one dialog turn. While, HotpotQA focuses on multi-document reasoning, and challenges the research community for developing new methods to acquire supporting information. To know more about this news, check out the post by Stanford. Stanford experiment results on how deactivating Facebook affects social welfare measures Thank Stanford researchers for Puffer, a free and open source live TV streaming service that uses AI to improve video-streaming algorithms Stanford researchers introduce DeepSolar, a deep learning framework that mapped every solar panel in the US

Nicholas Bergson-Shilcock, the co-founder of Recurse Center, announced yesterday, that the company has nearly achieved its 2012 goal of making RC 50% women. Recurse Center is a self-directed, community-driven educational retreat for programmers in New York City. After seven years, 48% of new hires at RC this year are women, trans, or non-binary. “We believe nearly every aspect of RC gets better when RC becomes more diverse...my cofounders and I have experienced RC across 60 batches: some with significant gender, racial, age, and other forms of diversity, and others with very little diversity. We believe firmly that the former are a better experience for everyone”, states Shilcock. RC mainly focused on three things as a part of its strategy to achieve its goal. These three things include: getting a strong and diverse pool of applicants, minimizing bias and evaluating everyone on the same admissions criteria, and building an environment where different people can easily thrive. As a part of RC’s strategy: It first focussed on getting as strong and diverse a range of applicants as possible. To achieve this, RC funded women and people from other traditionally underrepresented groups for the program. For instance, RC  partnered with Etsy in April 2012 to fund living expense grants for women who can’t afford to attend RC. RC further expanded its grants program in 2014 to include Black, Latina/o, Native American, and Pacific Islander people. By 2015, RC began funding grants itself and has managed to disburse over $1.5 million in grants so far. Apart from that, RC also offered merit-based fellowships of up to $10,000   to women, trans, and non-binary people that work on open source projects, research, and art. RC launched Joy of Computing last year which is a site that features technical work by members of the RC community. Secondly, RC focused on minimizing bias and evaluate everyone on the same admissions criteria. To achieve this, RC uses pseudonyms and hides the demographic information. For instance, RC updated its admissions review software in 2014 to replace people’s names with pseudonyms ( “Keyboarding Animal” or “Temperature Jeans” instead of “José Smith” or “Kimberly Lin”). Shilcock recommends that companies should document and clearly explain their admission criteria and process. Firms should also be very specific about what precisely comes under their criteria. RC also recommends training its interviewers and recording the interviews for quality control and training. Additionally, RC offers ongoing support and has a process in place for giving interviewers feedback Lastly, RC focused on building and nurturing an environment where different kind of people can thrive. To foster a healthy work environment, RC has explicit social rules in place that contribute towards making RC a friendly, and productive place to program and grow. RC also has a code of conduct in place which is a system for reporting violations and a response protocol. Moreover, it also focuses on welcoming people and making them a part of its community. RC is further working towards itself as a firm more accessible to the programming community. For instance, apart from attending RC for six or 12-week batches, people can now also attend for a week-long program and become alumni and lifelong member of the community. RC has also modified and updated some of its policies to make RC more family-friendly. There is now a lactation and wellness room at RC, which will allow parents to bring their children along with them. Shilcock states that earlier in 2012, only 5% of the Recursers were women, trans, or non-binary, but that figure has changed to 34% now. Also, of the nearly 150 people who have already joined RC’s batch for this year, 48% identify as women, trans, or non-binary. However, Shilcock states that although the numbers are quite promising, they can fluctuate. “We know it will take continuous investment and work to have any chance of consistently achieving a gender-balanced environment at RC”, writes Shilcock. For more information, check out the official Recurse Center announcement. Women win all open board director seats in Open Source Initiative 2019 board elections Google’s pay equity analysis finds men, not women, are underpaid; critics call out design flaws in the analysis Apollo 11 source code: A small step for a woman, and a huge leap for software engineering

Big Squid, Inc. releases Kraken platform, Baidu’s Machine Reading Comprehension Challenge, Apache Kylin new version supports SparkSQL, and more in today’s top stories around machine learning, deep learning, and data science news. Top Data science News Stories of the Day Anaconda version 5.1.1 released! AWS releases image recognition AI for Asia-Pacific developers Other Data Science News at a Glance 1. “Released the Kraken”, announces Big Squid, Inc.. The Kraken platform, through self-service machine learning, aims to bring powerful analytical insights to customers, which will further extend the ability of organizations to gain insight into future trends affecting their business and to make decisions on those forecasts with greater certainty, maximizing their data and business intelligence investment.  Read more on PRWeb. 2. Baidu Research launches a Machine Reading Comprehension Challenge to advance the state of the art in Natural Language Processing (NLP). Contestants will have access to the world's largest Chinese MRC dataset and a chance to win 100k RMB. Read more on Baidu’s Challenge Page. 3. Progress Launches AI-Driven Chatbot, Progress NativeChat. It is an artificial intelligence-driven platform for creating and deploying chatbots. NativeChat is based on patent pending CognitiveFlow technology that can be trained with goals, examples and data from existing backend systems, similar to the process used for training new customer service agents. Read more on Digital Journal. 4. Apache Kylin has been updated with a new version that supports SparkSQL in building intermediate flat Hive tables. Kylin is an open source distributed analytics engine designed to provide a SQL interface and multi-dimensional analysis (OLAP) on Apache. Read more on I Programmer blog. 5. Google Open Sources its Exoplanet-Hunting AI, Kepler. Google is saving everyone the time of training a neural network on Kepler data by releasing its code freely. One can get the TensorFlow code on GitHub. Read more on Extreme Tech.

The PostGIS development team released version 2.5.0 of PostGIS, a spatial database extender for PostgreSQL, yesterday. PostGIS 2.5.0 explores new features, breaking changes, improvements, and fixes.  PostGIS is an open source software program which provides support for geographic objects to the PostgreSQL object-relational database. PostGIS comprises simple features for SQL specification from the Open Geospatial Consortium (OGC). New Features ST_OrientedEnvelope (Returns a minimum rotated rectangle enclosing a geometry) and ST_QuantizeCoordinates  (Sets least significant bits of coordinates to zero) have been added in PostGIS 2.5.0. Apart from that, other new features such as ST_FilterByM (Filters vertex points based on their m-value) and ST_ChaikinSmoothing (Returns a "smoothed" version of the given geometry with the help of Chaikin algorithm) have also been added in PostGIS 2.5.0. Breaking Changes The Version number has been removed from address_standardize lib file in PostGIS 2.5.0. The raster support functions can be loaded only in the same schema with core PostGIS functions. The dummy pgis_abs type has been removed from aggregate/collect routines. Support has been removed for drop support GEOS < 3.5 and PostgreSQL < 9.4. Improvements and bug Fixes There’s been performance improvement for sorting POINT geometries in PostGIS 2.5.0. An external raster band index has been added to ST_BandMetaData. Also, there’s an added Raster Tips section in Documentation for information about Raster behavior (e.g. Out-DB performance, maximum open files). The use of GEOS in topology implementation has been reduced in PostGIS 2.5.0. A bug that created MVTs with incorrect property values under parallel plans has been fixed. Geometry in PostGIS 2.5.0 has been simplified using map grid cell size before generating MVT.  BTree sort order has now been defined on collections of EMPTY and same-prefix geometries in PostGIS 2.5.0. Hashable geometry feature enables direct use in CTE signatures in PostGIS 2.5.0. PostGIS 2.5.0 will not be accepting EMPTY points as topology nodes. ST_GeometricMedian now provides support for point weights in PostGIS 2.5.0. Duplicated code in lwgeom_geos has been removed. For more information on other updates and fixes in PostGIS 2.5.0, check out the official release notes. Writing PostGIS functions in Python language [Tutorial] Adding PostGIS layers using QGIS [Tutorial] PostGIS extension: pgRouting for calculating driving distance [Tutorial]

DeepMind introduces NarrativeQA, a data repository setup for understanding complex narratives. Reading comprehension (RC)—in contrast to information retrieval—requires integrating information and reasoning about events, entities, and their relations across a full document. The question answering technique is traditionally used to assess the abilities of RC, both in AI agents and in children who are learning to read. However, DeepMind surveyed that the existing RC datasets such as MCTest, Children’s Book Test(CBT), CNN/Daily Mail, NewsQA, SearchQA, and so on and found out certain limitations which include, presence of small datasets, unnatural data, requirement of a single sentence of information to answer the questions, and so on. Hence, these RC datasets are unable to test an important integrative aspect of machine’s Reading Comprehension. In order to encourage deeper comprehension of language, DeepMind presents a brand new dataset and a set of tasks, known as the NarrativeQA. This dataset includes fictional stories, which are 1,567 complete stories from books and movie scripts, with human written questions and answers based solely on human-generated abstract summaries. The dataset is divided into three parts: non-overlapping training validation and testing There are 46,765 pairs of answers to questions written by humans and includes mostly the more complicated variety of questions such as "when / where / who / why". This dataset permits the training of neural network-based models over word embeddings and provide decent lexical coverage and diversity.Thus, this dataset would test and reward agents that approach human level of competency. Having given a quantitative and qualitative analysis of the difficulty of the more complex tasks, DeepMind suggests research directions that may help bridge the gap between existing models and human performance. DeepMind also hopes that this dataset will serve not only as a challenge for the machine reading community, but also as a driver for the development of a new class of neural models which will take a significant step beyond the level of complexity which existing datasets and tasks permit. To have a detailed understanding on the working of NarrativeQA dataset, you can have a look at the research paper here.

Earlier this month, AKQA, a global innovation agency, introduced a new outdoor sport called Speedgate, which is created by an AI system built by them. This AI system was trained on more than 400 sports including Rugby, Soccer, and football to form the rules and regulations for Speedgate. In Speedgate, each team has six players consisting of forwards and defenders. The teams playing the game have to score goals by kicking through two consecutive gates. When a player kicks the ball through an X gate, the center gate will unlock the goal gate. After the center gate is unlocked, the team in possession can score by kicking the ball through the end gate in any direction. Here’s a video showing how this game actually works: https://www.youtube.com/watch?v=Uj4CQiuX8GM&feature=youtu.be Developers at AKQA trained a recurrent neural network and a deep convolutional generative adversarial network on over 400 sports. It uses NVIDIA Tesla GPUs for training the neural networks as well as for inferencing. Additionally, the neural network was also trained on 10,400 logos to generate the official Speedgate logo. The model was able to generate over 1,000 different sport concepts. Though many of them were interesting, the team wanted the AI system to come up with a game that was in addition to being fun and easy to understand was also active and accessible. And, Speedgate checked all the boxes for them. Kathryn Webb, AI Practice Lead at AKQA, said, “GPU technology helped us to condense training and generation phases down to a fraction of what they would’ve been. We would not have been able to achieve so many unique ML contributions in the project without that speed. It gave us more time to test, learn and adapt, and ultimately helped to produce the best final result.” Read more in detail, visit AKQA’s official website. OpenAI Five beats pro Dota 2 players; wins 2-1 against the gamers Google announces Stadia, a cloud-based game streaming service, at GDC 2019 Microsoft announces Game stack with Xbox Live integration to Android and iOS  

Yesterday, Senator Josh Hawley proposed a bill to voice against the different techniques tech giants use to exploit users’ attention for keeping them addicted to their apps. The Social Media Addiction Reduction Technology (SMART) Act would “ban certain features that are designed to be addictive, would require choice parity for consent, and would give users the power to monitor their time spent on social media”,  Sen. Hawley’s official post states. “Big tech has embraced a business model of addiction. Too much of the ‘innovation’ in this space is designed not to create better products, but to capture more attention by using psychological tricks that make it difficult to look away. This legislation will put an end to that and encourage true innovation by tech companies,” Senator Hawley said. “Deceptive design played an enormous part in last week’s FTC settlement with Facebook, and Hawley’s bill would make it unlawful for tech companies to use dark patterns to manipulate users into opting into services”, The Verge reports. The bill would ban user in-app achievements such as “Snapstreak” on Snapchat that gets the user addicted and difficult to leave the social media platform. [box type="shadow" align="" class="" width=""]Snapchat explains Snapstreaks as “The number next to the 🔥 tells you how many days you've been on a Snapstreak. For example, if you have an 8 next to the 🔥 it means you both have Snapped (not chatted) back and forth with this friend for 8 days.[/box] The bill, if passed, would require social media organizations to, within six months,  implement a feature allowing users to set a time limit on how long they can access the platform each day. With the default time limit being 30 minutes, "if the user elects to increase or remove the time limit, [it] resets the time limit to 30 minutes a day on the first day of every month," the bill text says. The bill also demands including a pop-up every 30-minute that would notify users of the total time spent. Apple and Google have included these monitoring systems with Screen Time and Digital Wellbeing. Instagram and Facebook also let you keep tabs on how much time you spend on them each day. Josh Golin, Executive Director of Campaign for a Commercial-Free Childhood, said, “Social media companies deploy a host of tactics designed to manipulate users in ways that undermines their wellbeing. We commend Senator Hawley for introducing legislation that would prohibit some of the most exploitative tactics, including those frequently deployed on children and teens.” Sen. Hawley talked about natural stopping points, like the end of a page, naturally prompt users to choose whether to continue reading. However, tech giants eliminate these mental opportunities by using structures like infinite scroll for newsfeeds and autoplay for videos. Within three months, if the bill is passed, the companies would be banned from offering features that automatically load and display content "other than music or video content that the user has prompted to play" without that person opting in. If users have reached the end of a block of a tweet, they will have to "specifically request (such as by pushing a button or clicking an icon, but not by simply continuing to scroll) that additional content is loaded and displayed." In a hearing on putting legislative limits on the persuasiveness of technology, late last month, Tristan Harris, a former Google design ethicist, explained how platforms create products to increase the amount of time users spend on a site. “If I take the bottom out of this glass and I keep refilling the water or the wine, you won’t know when to stop drinking. That’s what happens with infinitely scrolling feeds”, Harris explained the committee. According to Bloomberg, Google and Facebook declined to comment. NetChoice, a trade group that counts both companies as members, said, “The goal of this bill is to make being online a less-enjoyable experience.” Many users and app developers are not in favour of the bill and have exclaimed why this tech is was implemented. A user on HackerNews writes, “I'm the dev that built Netflix's autoplay of the next episode. We built it first on the web player because it is easy to A/B test new features there. We called it "post-play" at the time…...So yes, Netflix wants you to spend more hours watching Netflix and the product team is scientifically engineering the product to make it more addictive. But...the product team at Doritos does the same thing.” https://twitter.com/ptbrennan11/status/1156221816983248896 A user on Reddit comments, “I design user interactions for a living, and infinite scroll is used ALL the time, and often in different levels/areas of a page. Just like we can design things to be fast, easy, addicting, etc, we can design things to be slow and require more consideration….. This seems like a thoughtless proposal that needs to be a step up from where it is: …..” https://twitter.com/petersuderman/status/1156296953069744128 https://twitter.com/JDVance1/status/1156285549638012930 A lot of users may not accept the conditions of this bill as they feel it would be too restrictive and the pop-ups after certain time interval may break their continuity online, and a lot other factors. However, some feel, social media companies can at least provide choices for users to keep autoplay settings on. A user David Kwan commented on The Verge article, “With the intent of establishing UX/UI design policies, instead of a ”ban,” the US gov could establish design guidelines (like the UK gov) that can help mitigate online addiction and other design matters that affect how people interact with digital media. For example, instead of allowing platforms like YouTube to set autoplay settings “on” by default, the default setting should be set to “off” instead.” https://twitter.com/reckless/status/1156206435887439874 In May, Sen. Hawley introduced a bill to ban loot boxes in video games that said such microtransactions exist to exploit children. In June, he introduced a bill that declares top internet companies to undergo external audits to evaluate whether their content moderation systems are free of political bias. To know more about Sen. Hawley’s SMART Act in detail, read the proposed bill. Along with platforms like Facebook, now websites using embedded ‘Like’ buttons are jointly responsible for what happens to the collected user data, rules EU court Microsoft mulls replacing C and C++ code with Rust calling it a “modern safer system programming language” with great memory safety features Microsoft adds Telemetry files in a “security-only update” without prior notice to users  

Artificial Intelligence has been climbing on the success of playing difficult classic board games like Chess and Go to complex multiplayer online games like DOTA 2 and StarCraft. Last month, Google DeepMind’s AI AlphaStar defeated StarCraft II pros. Unity also launched an ‘Obstacle Tower Challenge’ to test AI game players. In a similar move, yesterday Allen Institute for Artificial Intelligence released Iconary an AI Pictionary game which allows you to collaborate with an Artificial Intelligence software. It’s not a man vs machine but more of a man and machine collaborative game. Per the researchers behind this game, “Iconary is a breakthrough AI game in that it is the first Common Sense AI game involving language, vision, interpretation and reasoning.” Gameplay Iconary offers players a limited set of icons along with a phrase describing a situation. Players need to use the icon set to compose a scene that represents the phrase and the AllenAI will try to guess it correctly. It can also update its compositions based on its human partner's guesses to help successfully guide them towards the correct phrase. The AI plays both on the drawing side and guessing side. For guessing, the Artificial Intelligence arranges icons and the players have to guess the phrase. There are over 75,000 phrases supported in Iconary, with more being added regularly. However, the astonishing thing is that there are uncountable ways of representing them. This is challenging for an AI system, according to researcher Ani Kembhavi, “because it tests a wide range of common sense skills. The algorithms must first identify the visual elements in the picture, figure out how they relate to one another, and then translate that scene into simple language that humans can understand. This is why Pictionary could teach computers information that other AI benchmarks like Go and StarCraft can’t” The main goal of Iconary is to help AI systems come to an understanding of what humans are asking of it. This will help in overcoming multiple roadblocks in simple tasks by having humans and AI understand complex phrases. The researchers write, “AllenAI has never before encountered the unique phrases in Iconary, yet our preliminary games have shown that our AI system is able to both successfully depict and understand phrases with a human partner with an often surprising deftness and nuance.” You may give Iconary a try at iconary.allenai.org. Introducing SCRIPT-8, an 8-bit JavaScript-based fantasy computer to make retro-looking games Deepmind’s AlphaZero shows unprecedented growth in AI, masters 3 different games Electronic Arts (EA) announces Project Atlas, a futuristic cloud-based AI powered game development platform

After the recent release to TensorFlow 1.10.0 release candidate family, rc-0, the new release candidate rc-1 is out and available. Key highlights of this new version include major features and improvements to model training and evaluation, along with lots of bug fixes to the existing ecosystem. What’s new in TensorFlow 1.10.0 RC1? Modular changes The tf.lite runtime module now supports complex64 type Bigtable is a high-performance storage system which can help you store and serve training data. This new version will support the initial bigtable integration for tf.data With improved local run behavior in tf.estimator.train_and_evaluate function, there is no need to reload checkpoints for evaluation Now you can restrict the way workers and PS interact by setting device_filters in RunConfig class. Thus speeding up the training process and ensuring clean shutdowns in specific situations. However, if you want the workers and PS to communicate in order to complete the jobs, you will have to set customized session_options in RunConfig class. Feature additions and improvements Now you can find Distributions and Bijectors in TensorFlow Probability, which was initially found at tf.contrib.distributions. By the end of 2018 tf.contrib.distributions will be removed. New endpoints are added for existing TensorFlow symbols. Going forward these new endpoints are expected to be the preferred endpoints and may replace some of the existing endpoints in the future. You can find the new symbols added to the following modules: tf.debugging, tf.dtypes, tf.image, tf.io, tf.linalg, tf.manip, tf.math, tf.quantization, tf.strings. Breaking changes done to the ecosystem All the new prebuilt libraries are built against NCCL 2.2. They no longer include NCCL in the binary install. If you want to bring the complete usage of TensorFlow with multiple GPUs and NCCL you will need to upgrade it to NCCL 2.2. You can find the updated installation guide on Installing TensorFlow on Ubuntu and Install TensorFlow from Sources. From TensorFlow 1.11 release onwards, Windows builds will use Bazel. Hence this change will drop the official support for cmake. To get full details on the features list and bug fixes done in this release candidate, you can check out Tensorflow’s official release page on Github. Read more Why Twitter (finally!) migrated to Tensorflow How TFLearn makes building TensorFlow models easier Distributed TensorFlow: Working with multiple GPUs and servers

The MongoDB Go driver team has announced the second alpha release of the official Go driver. The official MongoDB Driver API for Go has certain changes as a part of the second alpha release. This release mainly contains improvements to the user experience and bug fixes. A point to be noted here is that this is an alpha software, so it is not recommended for production use. The MongoDB Go Driver team said, “following semantic versioning, the v0 version of the public API should not be considered stable and could change.” Changes since the prior release include: New Features: Examples for sample shell commands created Marshal, Unmarshal, and UnmarshalDocument functions added to BSON library Stringer for objectid.ObjectID implemented Improvements: New BSON library is tested against BSON corpus ReplaceOptions replaced from UpdateOptions CRUD tests resynced to update insertMany test format to a map Namespace type added for options in mongo package. DecodeBytes method added to the Cursor A method added to bson.Value to get the offset into the underlying []byte mongo.Cursor is made its own interface Document.ElementAt usability improved bson.ArrayConstructor renamed to bson.ValueConstructor FromHex function added to the objectid package Bugs: Lookup should properly traverse Arrays Documentation for the bson package wrt the builder.Builder type needs to be clarified Ensure methods of *Document handle the case where *Document is nil Update bson.ErrTooSmall bson.Reader.Lookup should return ErrElementNotFound if no element is found The official documentation is available on GoDoc. Questions and inquiries can be asked on the mongo-go-driver Google Group.

The EU parliament voted last week to develop what is being described as the largest biometric database on earth. Once created, the database will connect the systems used by various border control, migration and law enforcement agencies into a truly gigantic searchable database for both EU and Non EU citizens. The new database will be called the Common Identity Repository (CIR) and will unify records of over 350 million people. What’s the purpose of the Common Identity Repository? The CIR will streamline a number of operations, bringing together information that is highly distributed - and even siloed - into one place. It will mean that officials will only need to search a single database rather than multiple ones. But accessibility is only one element - it also brings together layers of biometric information such as fingerprints, faces and personal data, like passport numbers. According to Politico Europe, the new system “will grant officials access to a person’s verified identity with a single fingerprint scan.” The multifaceted nature of the system can be explained by the way it was approved by the European Parliament. It went through on two separate votes: one for merging systems used for things related to visas and borders were approved 511 to 123 (with nine abstentions), and the other for streamlining systems users for law enforcement, judicial, migration, and asylum matters was approved 510 to 130 (also with nine abstentions). On this EU officials stated last week that, "The systems covered by the new rules would include the Schengen Information System, Eurodac, the Visa Information System (VIS) and three new systems: the European Criminal Records System for Third Country Nationals (ECRIS-TCN), the Entry/Exit System (EES) and the European Travel Information and Authorisation System (ETIAS)" Criticism of the Common Identity Repository The plan has come in for serious criticism from those who argue that there are serious privacy rights at stake. The civil liberties advocacy group Statewatch had asserted last year that it would lead to the “creation of a Big Brother centralised EU state database and have called CIR as the point of no return.” The European Parliament says “the system will make EU information systems used in security, border and migration management interoperable enabling data exchange between the systems.” It is also argued by the critics that once up and running, CIR will be one of the biggest people-tracking databases in the world, right behind the systems used by the Chinese government and India's Aadhar system. https://twitter.com/fs0c131y/status/1120374735693598720 Microsoft and Cisco propose ideas for a Biometric privacy law after the state of Illinois passed one Biometric Information Privacy Act: It is now illegal for Amazon, Facebook or Apple to collect your biometric data without consent in Illinois SafeMessage: An AI-based biometric authentication solution for messaging platforms

Google announced last week that it has improved the handwriting recognition feature in Gboard, Google’s popular keyboard for mobile devices, as it is quite fast and makes 20%-40% fewer mistakes than before. It was last year when Google added support for handwriting recognition in Gboard for Android that supported more than 100 languages. Also, advancements in Machine Learning allowed Google to come out with new model architectures and training methodologies. Google made changes to its initial approach that relied on hand-designed heuristics to build a single machine learning model. This machine learning model operates on the whole input and reduces error rates significantly as compared to the old version. Google also published a paper titled “Fast Multi-language LSTM-based Online Handwriting Recognition” explaining its research regarding online handwriting recognition. Google team states that since Gboard is used on a range of devices and screen resolutions, their first measure involves normalizing the touch-point coordinates. Then, the team converts the sequence of points into a sequence of cubic Bézier curves, which are then further used as inputs to a recurrent neural network (RNN). This RNN is trained to accurately identify the character being written. Bézier curves provide a consistent representation of the input across devices consisting of different sampling rates and accuracies. Another benefit is that the sequence of Bézier curves is way more compact than the underlying sequence of input points. This makes it easier for the model to pick up temporal dependencies along the input. Now, although the sequence of curves represents the input, there is still a need for the researchers to translate the sequence of input curves into the actual written characters. Hence, a multi-layer RNN is used in order to process the sequence of curves and produce an output decoding matrix. Researchers settled on using a bidirectional version of Quasi-recurrent neural networks (QRNN). QRNNs alternate between convolutional and recurrent layers, and offers good predictive performance. Additionally, in order to "decode" the curves, RNN produces a matrix, where each column corresponds to one input curve, and each row corresponds to a letter in the alphabet The QRNN-based recognizer converts the curves’ sequence into character sequence probabilities of the same length. Also, to offer the best user-experience, accurate recognition models are not enough. This is why researchers have converted their recognition models (trained in TensorFlow) to TensorFlow Lite models. “We will continue to push the envelope beyond improving the Latin-script language recognizers. The Handwriting Team is already hard at work launching new models for all our supported handwriting languages in Gboard”, states the Google team. For more information, check out the official Google AI blog. Google Cloud security launches three new services for better threat detection and protection in enterprises Google releases a fix for the zero day vulnerability in its Chrome browser while it was under active attack Google open-sources GPipe, a pipeline parallelism Library to scale up Deep Neural Network training

Before and during surgery anesthesiologists need to keep track of the anesthesia administered and the patient’s vitals. Imbalance in the level of anesthesia can cause low oxygen levels in the blood known as hypoxemia. Currently, there is no system to predict when this could happen during surgery and the patient is at the mercy of an anesthesiologist’s experience and discretion. The machine learning system called ‘Prescience’ A team of researchers from the University of Washington have come up with a system to predict if a patient is at the risk of hypoxemia. This is done using patient data like age and body mass index. Data from 50,000 surgeries was collected to train the machine learning model. The team wanted the model to solve two different kinds of problems. To look at pre-surgery patient information and predict whether a patient would have hypoxemia after anesthesia is administered. To predict the occurrence of hypoxemia at any point during the surgery by using real-time data. While predicting, for the first problem, the BMI was a crucial factor while for the second, the oxygen levels. Then, Lee and Lundberg worked on a new approach to train Prescience in a way that it would generate understandable explanations behind its predictions. Testing the model Now it was time to test Prescience. Lee and Lundberg created a web interface. It ran the anesthesiologists through cases from surgeries in the dataset that were not used to train Prescience. For the real-time test, the researchers specifically chose cases that would be hard to predict. For example, when a patient’s blood oxygen level is stable for 10 minutes and then drops. It was noted that Prescience improved the ability of doctors to correctly predict a patient’s hypoxemia risk by 16 percent before a surgery. In real-time, during a surgery it was able to predict the risk by 12 percent. With the help of Prescience, the anesthesiologists were able to correctly distinguish between the two scenarios nearly 80 percent of the time both before and during surgery. Prescience is not ready to be used in real operations yet. Lee and Lundberg plan to continue working with anesthesiologists to improve Prescience. In addition to hypoxemia, the team hopes to predict low blood pressure and recommend appropriate treatment plans with Prescience in the future. This method ‘opens the AI black box’ Although they could have successfully built a model that could predict hypoxemia, the researchers also wanted to answer the question “Why?”. A change from the traditional black box AI models engineers and researchers are used to. Lee, author of the paper said: “Modern machine-learning methods often just spit out a prediction result. They don’t explain to you what patient features contributed to that prediction. Our new method opens this black box and actually enables us to understand why two different patients might develop hypoxemia. That’s the power.” Who are the team members? The research team consisted of four people, two from medicine and two from computer science. Bala Nair, research associate professor of anesthesiology and pain medicine at the UW School of Medicine, Su-In Lee, an associate professor in the UW’s Paul G. Allen School of Computer Science & Engineering, Monica Vavilala, professor of anesthesiology and pain medicine at the UW School of Medicine and Scott Lundberg, a doctoral student in the Allen School. The system is not meant to replace doctors. You can read the research paper at Nature science journal and the University of Washington website. Swarm AI that enables swarms of radiologists, outperforms specialists or AI alone in predicting Pneumonia How to predict viral content using random forest regression in Python [Tutorial] SAP creates AI ethics guidelines and forms an advisory panel

IBM unveiled a new data science and machine learning platform for accelerating AI adoption. Termed as IBM Cloud Private for Data, it is an integrated data science, data engineering and app building platform which can ingest and analyze massive amounts of data –one million events per second. According to the official press release, “The platform is built to enable users to build and exploit event-driven applications capable of analyzing the torrents of data from things like IoT sensors, online commerce, mobile devices, and more.” The core features of the IBM Cloud Private for Data is an in-memory database, a real-time processing engine and the ability to ingest and analyze massive amounts of data. The platform can ingest up to 1 million rows per second and 250 billion events per day for both transactional and analytical processing. Another feature of the platform is an ML-powered intelligent data catalog that automates the process of creating meta-tags. It also has a real-time ingestion engine based on Apache Spark and the Apache Parquet column-oriented data store. The IBM Cloud Private Data solution also includes key capabilities from IBM’s Data Science Experience, Information Analyzer, Information Governance Catalogue, Data Stage, Db2 and Db2 Warehouse. This new solution provides a data infrastructure layer for AI behind the firewall. Talking about the architecture, IBM Cloud Private for Data is an application layer deployed on the Kubernetes open-source container software. It forms an integrated environment for data science and application development using microservices. In the future, the Cloud Private for Data will run on all cloud and will be available in industry-specific solutions, for financial services, healthcare, manufacturing, and more. In addition, IBM has also announced the formation of the Data Science Elite Team – an elite team of consultants who will help solve real-world data science problems of clients, with no charge. “These two data science efforts,” according to Rob Thomas, General Manager, IBM Analytics, “will bring the AI destination closer and give access to powerful machine learning and data science technologies that can turn data into game-changing insight.” Have a look at the IBM Cloud Private for Data official announcement for further details.