Tech News

Every year Accenture releases their annual forecast of the technology trends unfolding in the next three years. This year their Accenture Technology Vision 2018  highlights five emerging trends shaping the way technology is increasing businesses’ impact across society. The report covers people on all rungs of the corporate ladder, from strategy to operations, moving closer to the center of people’s lives. Accenture’s Technology Vision comprises a three-year set of technology trends. Accenture Evolution Chart Here are the 5 major trends identified in the 2018 report. Trend #1 CITIZEN AI: Raising AI to benefit business and society As artificial intelligence grows in its capabilities, businesses must move to “raise” their AIs to act as responsible, productive members of society. For businesses, this means deploying AI is no longer just about training it to perform a given task. It’s about “raising” it to act as a responsible representative of the business, and a contributing member of society. Per a survey conducted by Accenture, “72 percent of executives report that their organizations seek to gain customer trust and confidence by being transparent in their AI-based decisions and actions. This will be a crucial step in the integration of AI into society. We call it Citizen AI.” Trend #2 EXTENDED REALITY: The End of Distance Extended Reality XR refers to the spectrum of experiences that blurs the line between the real world and the simulated world. The technology immerses the user through visual, audio, and potentially olfactory and haptic cues. The two major types of XR are virtual reality and augmented reality. The report ascertains that Virtual and augmented reality technologies are removing the distance to people, information, and experiences, transforming the ways people live and work. Per their Survey, “27 percent of executives state it is very important for their organizations to be a pioneer in XR solutions.” Trend #3 DATA VERACITY: The Importance of Trust By transforming themselves to run on data, businesses have created a new kind of vulnerability: inaccurate, manipulated, and biased data that leads to corrupted business insights with a major impact on society. Businesses can address this new vulnerability by building confidence in three key data-focused tenets: Provenance, or verifying the history of data from its origin throughout its life cycle Context, or considering the circumstances around its use Integrity, or securing and maintaining data. Strong cybersecurity and data science capabilities are prerequisites for building a strong data intelligence practice to ensure data veracity. Trend #4 FRICTIONLESS BUSINESS: Built to partner at scale Businesses depend on technology-based partnerships for growth, but their own legacy systems aren’t designed to support partnerships at scale. To fully become the connected Intelligent Enterprise, companies must first re-architect themselves. Organizations must actively reshape their business from top to bottom to meet the challenges of creating and managing relationships at scale. They must adopt microservices architectures, and use blockchain and smart contracts to build a strong foundation for technology-based partnerships. Accenture Survey shows that “ 60% of executives report that blockchain and smart contracts will be critical to their organizations over the next three years.” Trend #5 INTERNET OF THINKING: Creating Intelligent Distributed Systems Businesses are making big bets on intelligent environments via robotics, AI and immersive experiences. But to bring these intelligent environments to life, they must extend their infrastructures into the dynamic, real-world environments. For this, companies must extend compute beyond the cloud, toward the edge of networks. They must explore custom hardware solutions and hardware accelerators that let systems reduce latency and compute limitations. Cloud processing remains appealing for high-value learning, predictions, AI-model generation, and storage in situations that are not time critical. But for real-time action, processing must happen at the edge of networks, where the event is occurring. Read the complete  Accenture Technology Vision 2018 report to learn about the technology trends in detail along with the research methodology and the survey demographics. 18 striking AI Trends to watch in 2018 – Part 1 18 striking AI Trends to watch in 2018 – Part 2 Skill Up is back: 5 things we want to find out

The Electron team announced the first stable release Electron 3.0.0, yesterday. Electron 3.0.0 comes with upgrades, fixes, and other changes. Electron is a framework which helps create native cross-platform applications with web technologies such as JavaScript, HTML, and CSS. Let’s see what’s new in Electron 3.0.0. New Features and changes There are two new features added in Electron 3.0.0, namely, TextField and Button APIs. However, these APIs are currently experimental and have been therefore turned off by default. You can enable these APIs with the enable_view_api build flag. The fs.readSync feature now works with massive files.  There’s a new netLog API added for dynamic logging control. Breaking API changes With Electron 3.0.0, updates have been made to menu item order control. The did-get-response-details and did-get-redirect-request events have been removed. JSON is no longer used to send the result of ipcRenderer.sendSync. The api::Window has been renamed to api::BrowserWindow. The app-command media-play_pause has been renamed to media-play-pause. macOS A new workspace notifications support has been added. There’s also a mouse forward functionality now on macOS. Windows A DIP has been added to/from screen coordinate conversions. Bug Fixes The issue with fs.statSyncNoException throwing exceptions has been fixed. The occurrence of a crash when loading site with jquery has been fixed. Also, the crash in net::ClientSocketHandle destructor has been fixed. Now, with Electron 3.0.0, the focus change will be notified right away rather than on next tick. macOS The context menu click callback has been fixed. The tray.setContextMenu crash has been resolved. Also, tray title crash has been fixed. The mac protocol that used to get set to none after calling app.removeAsDefaultProtocolClient, is resolved. Linux BrowserWindow.focus() for offscreen windows has been fixed. For more information on Electron 3.0.0, check out the official release notes. How to build a cross-platform desktop application with Node.js and Electron How to build a desktop app using Electron Web Scraping with Electron

Version 5 of Apple’s Swift language, used for iOS and MacOS application development, will release early next year. The main focus of this release will be an ABI (application binary interface) stability in the standard Swift library, in addition to Standard Library Improvements, Foundation Improvements and Syntactic Additions. ABI (application binary interface) features in Swift 5.0 ABI defines how to call a function, how data is represented in memory, where metadata is and how to access it. The current version of Swift is not ABI stable, so every binary (App), bundles its own version of the Swift Dynamic Library. For instance, if App1 is using Swift 3.0, it bundles Swift 3.0 Dynamic Library (containing the 3.0 ABI) inside. And if App2 is using Swift 3.2, it bundles Swift 3.2 and it’s 3.2 ABI. Here Swift doesn’t live on the iOS Operating System, it lives within each App. The ABI in Swift 5.0 will enable future compiler versions to produce binaries that conform to the stable ABI. A stable ABI tends to persist for the rest of the platform’s lifetime due to ever-increasing mutual dependencies. If Swift becomes ABI Stable, its ABI will be compatible with every version of Swift. For example, if App1 is using Swift 5.0, but App2 is using Swift 5.3, both will be consuming the Swift ABI embedded in the Operating System. The ABI feature was originally intended for Swift 4 release. Carryover goals from Swift 4 that are required for implementing the ABI in Swift 5 include: Generics features for the standard library. This includes conditional conformances for generic types and recursive protocol types and lifting restrictions on associated types in protocols API resilience, which will allow public APIs for a library to evolve A memory ownership model Besides ABI stability, expect these improvements in Swift 5: #1 String ergonomics Processing of the string type is expected to get better as users will have the ability to create raw strings, distinguish between enums that are fixed and enums that might change in the future,  and check whether one number is a multiple of another by using isMultiple(of:) #2 Groundwork for a new concurrency model Swift 5 will focus on designing language capabilities for building and using asynchronous APIs and dealing with problems created by callback-heavy code. #3 Targeted improvements to the Foundation API The Cocoa SDK- which was originally designed for Objective-C, can work seamlessly with Swift. To try out Swift 5.0 ahead of its release early next year, download the latest Swift trunk development snapshot, activate it inside your current Xcode version,  and then head over to Xcode Playground for examples you can edit. Read in depth about the new features to be implemented in Swift 5.0 at HackingwithSwift. Swift 4.2 releases with language, library and package manager updates! Apple bans Facebook’s VPN app from the App Store for violating its data collection rules iPhone XS and Apple Watch details leaked hours before product launch  

Be it a tweet about taking the company private or smoking weed on a radio show, Elon Musk has been in news for all the wrong reasons recently and he is in news again but this time for what he is best admired as a modern day visionary. As per reports the Tesla and SpaceX founder is working on a 'superhuman' product that will connect your brain to a computer. We all know Musk along with eight others founded a company called Neuralink two years ago. The company has been developing implantable brain–computer interfaces, better known as BCIs. While in the short-term the company’s aim is to use the technology to treat brain diseases, Musk’s eventual goal is human enhancement, which he believes will make us more intelligent and powerful than even AI.  According to hints he gave a week ago, Neuralink may soon be close to announcing a product unlike anything we have seen: A brain computer interface. Appearing on the Joe Rogan Experience podcast last week, Musk stated that he’ll soon be announcing a new product – Neuralink – which will connect your brain to a computer, thus making you superhuman. When asked about Neuralink, Musk said "I think we’ll have something interesting to announce in a few months that’s better than anyone thinks is possible. Best case scenario, we effectively merge with AI. It will enable anyone who wants to have superhuman cognition. Anyone who wants. How much smarter are you with a phone or computer or without? You’re vastly smarter, actually. You can answer any question pretty much instantly. You can remember flawlessly. Your phone can remember videos [and] pictures perfectly. Your phone is already an extension of you. You’re already a cyborg. Most people don’t realise you’re already a cyborg. It’s just that the data rate, it’s slow, very slow. It’s like a tiny straw of information flow between your biological self and your digital self. We need to make that tiny straw like a giant river, a huge, high-bandwidth interface." If we visualize what Musk said, it feels like a scene straight from a Hollywood movie. However, many of the creations, from a decade ago, that were thought to belong solely in the world of science-fiction, have become a  reality now. Musk argues that through our over-dependence on smartphones, we have already taken the first step towards our cyborg future. Neuralink is an attempt to just accelerate the process by leaps and bounds. That's not all, Elon Musk was also quoted saying on CNBC. "If your biological self dies, you can upload into a new unit. Literally, with our Neuralink technology". Read the full news on CNBC. About Author Sandesh Deshpande is currently working as a System Administrator for Packt Publishing. He is highly interested in Artificial Intelligence and Machine Learning. Tesla is building its own AI hardware for self-driving cars Elon Musk’s tiny submarine is a lesson in how not to solve problems in tech. DeepMind, Elon Musk, and others pledge not to build lethal AI

The API is the building block of much modern software. With Kong 1.0, launching today at Kong Summit, Kong believes it has cemented its position as the go-to platform for developing APIs on modern infrastructures, like cloud-native, microservices, and serverless. The release of the first stable version of Kong marks an important milestone for the company as it looks to develop what it calls a 'service control platform.' This is essentially a tool that will allow developers, DevOps engineers, and architects to manage their infrastructure at every point - however they choose to build it. It should, in theory off a fully integrated solution that let's you handle APIs, manage security permissions, and even leverage the latest in cutting edge artificial intelligence for analytics and automation. CEO Augusto Marietti said that "API management is rapidly evolving with the industry, and technology must evolve with it. We built Kong from the ground up to meet these needs -- Kong is the only API platform designed to manage and broker the demands that in-flight data increasingly place on modern software architectures." How widely used is Kong? According to the press release, Kong has been downloaded 45 million times, making it the most widely used open source API platform. The team stress that reaching Kong 1.0 has taken three years of intensive development work, done alongside customers from a wide range of organizations, including Yahoo! Japan and Healthcare.gov. Kanaderu Fukuda, senior manager of the Computing Platform Department at Yahoo! Japan, said: "as Yahoo! Japan shifts to microservices, we needed more than just an API gateway – we needed a high-performance platform to manage all APIs across a modern architecture... With Kong as a single point for proxying and routing traffic across all of our API endpoints, we eliminated redundant code writing for authentication and authorization, saving hundreds of hours. Kong positions us well to take advantage of future innovations, and we're excited to expand our use of Kong for service mesh deployments next." New features in Kong 1.0 Kong 1.0, according to the release materials "combines sub-millisecond low latency, linear scalability and unparalleled flexibility." Put simply, it's fast but also easy to adapt and manipulate according to your needs. Everything a DevOps engineer or solutions architect would want. Although it isn't mentioned specifically, Kong is a tool that exemplifies the work of SREs - site reliability engineers. It's a tool that's designed to manage the relationship between various services, and to ensure they not only interact with each other in the way they should, but that they do so with minimum downtime. The Kong team appear to have a huge amount of confidence in the launch of the platform - the extent to which they can grow their customer base depends a lot on how the marketplace evolves, and how much the demand for forward-thinking software architecture grows over the next couple of years. Read next: How Gremlin is making chaos engineering accessible [Interview] Is the ‘commons clause’ a threat to open source?

Facebook data scientists had released a paper, Constrained Bayesian Optimization with Noisy Experiments in 2017 where they describe using Bayesian optimization to design rounds of A/B tests based on prior test results. An A/B test is a randomized experiment, used to determine which variant of A and B is more "effective". They are used for improving a product. Facebook has a large array of backend systems serving billions of people every day. They have a large number of internal parameters that must be tuned carefully using live, randomized experiments, also known as A/B tests. Individual experiments may take a week or longer, so there is a challenge to optimize a set of parameters with the least number of experiments. Bayesian optimization Bayesian optimization is a technique used to solve optimization problems where the objective function (the online metric of interest) does not have an analytic expression. It can only be evaluated through some time consuming operations like a randomized experiment. Bayesian optimization allows joint tuning of more parameters with fewer experiments compared to a grid search or manual tuning. It also helps in finding better values. The Gaussian process (GP) is a Bayesian model that works well for Bayesian optimization. GP provides good uncertainty estimates of how an online metric varies with the parameters of interest. It is illustrated as follows: Source: Facebook research blog The work in the paper was motivated by several challenges in using Bayesian optimization for tuning online systems. The challenges are noise, constraints, and batch experimentation. In the paper, the authors describe a Bayesian approach for handling observation noise in which they include the posterior uncertainty induced by noise in EI’s expectation. In the paper, they describe a Bayesian approach for handling observation noise. A posterior uncertainty is induced by noise in EI’s expectation. Instead of computing the expectation of I(x) under the posterior of f(x), it is computed under the joint posterior of f(x) and f(x*). This expectation no longer has a closed form like El but can easily draw samples of values at past observations f(x_1), …, f(x_n) from the GP posterior. The conditional distribution f(x) | f(x_1), …, f(x_n) has closed form. The results The approach described in the paper is used to optimize various systems at Facebook. Two such optimizations are described in the paper. The first is to optimize six parameters from one of Facebook’s ranking systems. The second one was to optimize seven numeric compiler flags for the HipHop Virtual Machine (HHVM). The web servers powering Facebook use the HHVM to serve requests. The end goal of this optimization was to reduce CPU usage on the web servers, with a constraint of keeping the peak memory usage less. This following figure shows the CPU usage of each configuration tested. There is a 100 total, it also shows the probability that each point satisfied the memory constraint: Source: Facebook research blog The first 30 iterations were randomly generated configurations depicted as a green line. After this, the Bayesian optimization was used to identify parameter configurations to be evaluated. It was observed that Bayesian optimization was able to find better configurations that are more likely to satisfy the constraints. The findings are that Bayesian optimization is an effective and robust tool for optimizing via noisy experiments. For full details, visit the Facebook research blog. You can also take a look at the research paper. NIPS 2017 Special: A deep dive into Deep Bayesian and Bayesian Deep Learning with Yee Whye Teh Facebook’s Glow, a machine learning compiler, to be supported by Intel, Qualcomm and others “Deep meta reinforcement learning will be the future of AI where we will be so close to achieving artificial general intelligence (AGI)”, Sudharsan Ravichandiran

Mojolicious, a next generation web framework for the Perl programming language announced its upgrade to the latest 8.0 version. Mojolicious 8.0 was announced at the Mojoconf in Norway held from 6th to 7th September 2018. This release is codenamed as ‘Supervillain’ and is by far the major release in Mojolicious. Mojolicious allows users to easily grow single file prototypes into well-structured MVC web applications. It is a powerful web development toolkit, that one can use for all kinds of applications, independently of the web framework. Many companies such as Alibaba Group, IBM, Logitech, Mozilla, and others rely on Mojolicious to develop new code bases. Even companies like Bugzilla are getting themselves ported to Mojolicious. The Mojolicious community has decided to make a few organizational changes, to support the continuous growth. This includes: All new development will be consolidated in a single GitHub organization. Mojolicious’ official IRC channel named say hi! that has almost 200 regulars will be moving to Freenode (#mojo on irc.freenode.net). This will make it easier for people not yet part of the Perl community to get involved. Some highlights of the Mojolicious 8.0 Promises/A+ Mojolicious 8.0 includes Promises/A+, a new module and pattern for working with event loops. A promise represents the eventual result of an asynchronous operation. Roles and subprocess The version 8.0 now includes roles, a new way to extend Mojo classes. Also, the subprocesses can now mix event loops and computationally expensive tasks. Placeholder types and Mojo::File With the placeholder types, one can avoid repetitive routes. Whereas the Mojo::File, is the brand new module for dealing with file systems. Cpanel::JSON::XS and Mojo::PG With the Cpanel::JSON::XS, users can process JSON at a much faster rate now. The Mojo::PG includes many new SQL::Abstract extensions for Postgres features. To know more about Mojolicious 8.0 in detail, visit its GitHub page. Warp: Rust’s new web framework for implementing WAI (Web Application Interface) What’s new in Vapor 3, the popular Swift based web framework Beating jQuery: Making a Web Framework Worth its Weight in Code

The Swift team released Swift version 4.2 yesterday. Swift 4.2 comes with updates to the Swift language including generics improvement, and standard library updates. Swift 4.2 also comprises changes to the package manager. Swift is a general-purpose and multi-paradigm programming language developed by Apple Inc. for iOS, macOS, watchOS, tvOS, and Linux. Swift is built for uses that range from systems programming to mobile and desktop apps, to scaling up to cloud services. Let’s have a look at the key features in Swift 4.2. Language updates Swift 4.2 is a major language release and comprises of language changes such as Generics improvements, and Standard library updates. Generics Improvement Better support has been added for Generics which ultimately makes more of your code reusable. Standard Library updates The standard library in this latest release consists of a number of new features such as improvements to the Hashable protocol and a new unified set of randomization functions and protocols. Other updates Swift 4.2 has enabled the binary compatibility for future releases of Swift. Support has been added for batch mode compilation that results in faster build times. There’s a change in calling convention for retain/release cycle which helps reduce the code size and improve runtime performance. Package Manager Updates Swift 4.2 explores three new features namely, batch mode support, improved scheme generation logic, and automatic Xcode project generation. With Swift 4.2, the swift targets will be now compiled using the Swift compiler’s batch mode The scheme generation logic has been improved and it generates the following schemes: One scheme comprising all regular and test targets of the root package. One scheme per executable target consisting of the test targets whose dependencies intersect with the dependencies of the executable target. Swift 4.2 offers automatic Xcode project generation. The generate-xcodeproj here has a new –watch option which enables it to watch the file system and regenerate the Xcode project automatically, if needed. For more information on this release, check out the official release notes. Your First Swift Program What’s new in Vapor 3, the popular Swift based web framework Swift’s Core Libraries

IBM is defending itself from yet another round of accusations on its firing practices. Just months after a ProPublica report said the technology giant had dismissed more than 20,000 workers older than 40 in the last five years, Yesterday, Bloomberg reports that IBM is facing another age discrimination lawsuit. The suit that follows heavily on the ProPublica report, alleges that the company systematically fired tens of thousands of older workers in recent years as part of an effort to "make the face of IBM younger." Shannon Liss-Riordan, a lawyer known for battling tech giants over mistreatment of their employees, filed a class-action lawsuit in federal court in Manhattan on behalf of three former IBM employees in their 50s and 60s. They claimed that the tech giant fired them earlier this year based on their age. The three employees declare in the suit that "Over the last several years, IBM has been in the process of systematically laying off older employees in order to build a younger workforce," The complaint also states that the company "discriminates against older workers” and "does not consider them for open positions." Ed Barbini (a spokesman for IBM) said in an emailed statement to multiple news outlets: “Changes in our workforce are about skills, not age,” “since 2010 there is no difference in the age of our U.S. workforce, but the skills profile of our employees has changed dramatically. That is why we have been and will continue investing heavily in employee skills and retraining — to make all of us successful in this new era of technology.’’ In the last decade, IBM has reportedly fired thousands of people in the U.S., Canada, and other high-wage jurisdictions -to cut costs and retool its workforce. If the judge allows a class action lawsuit to proceed, it could lead to a drawn-out and costly court battle. IBM would potentially have to pay hundreds of millions of dollars to its former employees. To know more about this lawsuit, head over to Seattletimes. The Intercept says IBM developed NYPD surveillance tools that let cops pick targets based on skin color IBM Files Patent for “Managing a Database Management System using a Blockchain Database” Say hello to IBM RXN, a free AI Tool in IBM Cloud for predicting chemical reactions

The free and open internet has come under attack in recent weeks. Just last week, the European Parliament voted on the EU Copyright Directive that strengthens publisher rights online, with activists claiming the result to be a victory for everyone who opposes internet freedom. That's why Day Against DRM 2018 couldn't come at a better time. Now in its 12th year, Day Against DRM is a grass roots campaign that aims to raise awareness of DRM and encourage organization and action against it. Organized by defectivebydesign.org, a team of passionate activists committed to removing DRM, the day is supported by a wide range of other influential organizations (as well as Packt): The Electronic Frontier Foundation (EFF) LeanPub Open Rights Group Creative Commons  Free Software Foundation What is DRM? If you're reading this and wondering what DRM is, exactly, DRM stands for Digital Rights Management - to opponents, it's also know as Digital Restrictions Management. Essentially, it's about imposing restrictions on content online, controlling the ways it can be copied and shared. For software developers, DRM is particularly important - taken to its logical conclusion, DRM would inhibit open source software to such an extent to render it unrecognizable. Here at Packt, we take the principles of a DRM free internet seriously. We know that being able to use content in different ways - to have full control over it - is essential if you're a software developer. That's why we're throwing our support behind Defective by Design's campaign. It's also why all of our digital products are DRM free. You can learn more about the Day Against DRM here. 

Unity has released a benchmarking report after two years since the last Unity Benchmark report comparing the performance and load times of WebAssembly with asm.js. They have compared the performance of Unity WebGL in four major web browsers: Firefox 61, Chrome 70, Safari 11.1.2 and Edge 17.  Last month, Unity officially announced that it is finally making the switch to WebAssembly as their output format for the Unity WebGL build target. Note: All images and graphs are taken from the Unity Blog. For running the tests, the team rebuilt the Benchmark project with Unity 2018.2.5f1 using the following Unity WebGL Player Settings: Here are the findings from the report. Criteria 1: Total amount of time taken to get to the main screen for both WebAssembly and asm.js. Findings: Firefox is comparatively fast to load on both Windows and macOS Chrome and Edge load faster when using WebAssembly All browsers, except Safari, load faster with WebAssembly compared to asm.js. Criteria 2: In-Depth Load Times for WebAssembly-only. The team compared four factors: WebAssembly compilation and instantiation. Unity engine initialization and first scene load. Time it takes to render first frame. Time it takes to load and have a stable frame-rate. Findings: Firefox is the fastest overall on both Windows and Mac Edge compiles Wasm quickly (even faster than Firefox) but is slower in Unity engine initialization. Criteria 3: Performance and Load times for Real-World Projects Real-world projects result in larger builds which impact the end-user’s experience. Here is an overview of total scores using WebAssembly and asm.js Findings: All browsers perform better when using WebAssembly On Windows, all browsers perform similarly On macOS, Firefox outperforms all other browsers. Safari is the browser that benefits the most by WebAssembly since it doesn’t support asm.js optimizations. Conclusion The report findings conclude that modern browsers load faster and perform better thanks to WebAssembly. It also provides more consistent user experience as compared to asm.js. Read more about the findings on the Unity Blog. Unity releases ML-Agents toolkit v0.5 with Gym interface, a new suite of learning environments. Key Takeaways from the Unity Game Studio Report 2018. Unity switches to WebAssembly as the output format for the Unity WebGL build target.  

Microsoft released the version 2.6 of their popular deep learning framework, CNTK or Microsoft Cognitive Toolkit, last week. CNTK v2.6 explores features such as an added .NET support, efficient group convolution, improved sequential convolution, more operators, and ONNX feature update among others. Added .Net Support The Cntk.Core.Managed library has now been converted to .Net Standard. It now supports .Net Core and .Net Framework applications on both Windows as well as Linux. .Net developers will now be able to restore CNTK Nuget packages. To restore the CNTK Nuget packages, use the new .Net SDK style project file with package management format set to PackageReference. Efficient group convolution With CNTK v2.6, the group convolution has been updated. The updated implementation uses cuDNN7 and MKL2017 APIs directly instead of having to create a sub-graph for group convolution (slicing and splicing). This leads to improved experience in terms of both performance and model size. Improved Sequential Convolution Sequential Convolution implementation has also been updated in CNTK v2.6. The new implementation creates a separate sequential convolution layer. This layer offers support for broader cases, such as, where stride > 1 for the sequence axis. So, if sequential convolution is performed over a batch of one-channel black-and-white images then these images will have the same fixed height of 640 with the width of variable lengths. The width is then represented by the sequential axis. More Operators More support has been added in CNTK v2.6 for operators such as depth_to_space and space_to_depth, Tan and Atan, ELU, and Convolution. depth_to_space and space_to_depth There are breaking changes in the depth_to_space and space_to_depth operators. These two operators are updated to match the ONNX specification. Tan and Atan Support has been added for trigonometric ops Tan and Atan. ELU Support added for alpha attribute in ELU op. Convolution The auto padding algorithms of Convolution have been updated to produce symmetric padding at best effort on CPU, without influencing the final convolution output values. This leads to an increase in the range of cases which could be covered by MKL API and also improves the performance, E.g. ResNet50. Updated ONNX CNTK's ONNX import/export has been updated to support ONNX 1.2. A major update has been added on how the batch and sequence axes are handled in export and import.  CNTK's ONNX BatchNormalization op export/import has been updated to the latest spec. A model domain has been added to the ONNX model export. Support has also been added for exporting alpha attribute in ELU ONNX op. Change in Default arguments order There is a major updated to the arguments property in CNTK python API. The default behavior has been updated so now it returns the arguments in python order instead of in C++ order. Because of this, it will return arguments in the same order as they are fed into ops. Bug Fixes Improved input validation added for group convolution. Validation added for padding channel axis in convolution. Proper initialization added for ONNX TypeStrToProtoMap. The Min/Max import implementation has been updated to handle variadic inputs. There are even more updates that come with CNTK v2.6. For more information on those, check out the CNTK official release notes. The Deep Learning Framework Showdown: TensorFlow vs CNTK Deep Learning with Microsoft CNTK ONNX 1.3 is here with experimental function concept

The Cloudflare team has introduced Cloudflare’s IPFS Gateway which will make accessing content from the InterPlanetary File System (IPFS) easy and quick without having to install and run any special software on a user’s computer. The gateway which supports new distributed web technologies is hosted at cloudflare-ipfs.com. The team asserts that this will lead to highly-reliable and security-enhanced web applications. A brief gist of IPFS When a user accesses a website from the browser, it tracks down the centralized repository for the website’s content. It then sends a request from the user’s computer to that origin server, and that server sends the content back to the user's computer. However, this centralization mechanism makes it impossible to keep content online if the origin servers rolls back the data. If the origin server faces a downtime or the site owner decides to take down the data, the content stands unavailable. On the other hand, IPFS is a distributed file system that allows users to share files that will be distributed to other computers- throughout the networked file system. This means that a user’s content is stored on all the nodes of the network and data can be safely backed up. Key Differences between IPFS and the traditional Web #1 Free caching and serving of content IPFS provides free caching and serving of content. Anyone can sign up their computer to be a node in the system and start serving data. On the flipside, the traditional web relies on big hosting providers to store content and serve it to the rest of the web. Setting up a website with these providers costs money. #2 Content addressed data Rather than location-addressed data, IPFS focuses on content addressed data. In the traditional web, when a user navigates to a website, it fetches data stored at the websites IP address. The server sends back the relevant information from that IP. With IPFS, every single block of data stored in the system is addressed by a cryptographic hash of its contents. When a user requests for a piece of data in IPFS, they request it by its hash .i.e  content that has a hash value of, for example, QmXnnyufdzAWL5CqZ2RnSNgPbvCc1ALT73s6epPrRnZ1Xy Why is Cloudflare’s IPFS Gateway Important? The IPFS increases the resilience of the network. The content with a hash of-QmXnnyufdzAWL5CqZ2RnSNgPbvCc1ALT73s6epPrRnZ1Xy could be stored on dozens of nodes. So, if one of the nodes that was storing the content goes down, the network will just look for the content on another node. In addition to resilience, there is an automatic level of security introduced in the system. If the data requested by the user was tampered with during transit, the hash value the user gets will be different than the hash that he/she had asked for. This means that the system has a built-in way of knowing whether or not content has been tampered with. Users can access any of the billions of files stored on IPFS from their browser. Using Cloudflare’s gateway, they can also build a website hosted entirely on IPFS available to users at a custom domain name. Any website connected to IPFS gateway will be provided with a free SSL certificate. IPFS is embracing a new, decentralized vision of the web. Users will be able to create static web sites- containing information that cannot be censored by governments, companies, or other organizations- that are served entirely over IPFS. To know more about this announcement, head over to Cloudflare’s official Blog. 7 reasons to choose GraphQL APIs over REST for building your APIs Laravel 5.7 released with support for email verification, improved console testing Javalin 2.0 RC3 released with major updates!

Gatsby.js or more commonly known as Gatsby is a React-based website and app generator. It is powered by GraphQL and is used as a static site generator. But it’s not all static, it can be viewed more like a modern front end framework. It is used for creating blogs, apps ecommerce sites, and documentation. Yesterday the second major release, Gatsby 2.0 was released. Gatsby 2.0 comes 18 months after the first major release and has the hard work of the Gatsby core team and nearly 315 contributors. This second major release is focused on performance and developer experience. The highlights are 75% reduced build times, JavaScript client runtime by shrunk by 31%. Gatsby’s core dependencies are also upgraded to their latest versions: webpack 4, Babel 7, React 16.5. Gatsby 2.0 has faster site building The focus is heavily on improving build speeds for v2 and there are significant speed improvements across various parts of the build pipeline. The improvements include: Three to four times improved server side rendering performance due to React 16 Less memory usage for server rendering pages Many speedups to JavaScript and CSS bundling with webpack 4 A pull called “hulksmash” made many small fixes to refactor slow algorithms All available cores are used for rendering server pages JavaScript client runtime reduced by 31% The core JavaScript shipped with every Gatsby site is shrunk by 31%. Less use of JavaScript means faster websites. The core JavaScript size in Gatsby 1.0 was 78.5kb and in Gatsby 2.0 it is 53.9kb, both are sizes of GZIP files. The reduced sizes are largely due to the hard work done by the libraries. The code size is decreased by 30% in React 16. That is 34.8kb from the previous 49.8kb in React 15. The routers are switched from react-router to @reach/router bringing a 25% smaller bundle of 6kb from 8kb. For a complete list of changes, visit the Gatsby Blog. To know more visit their documentation and GitHub. low.js, a Node.js port for embedded systems Browser based Visualization made easy with the new P5.js Deno, an attempt to fix Node.js flaws, is rewritten in Rust

Earlier this month, some students of Robotics Institute Carnegie Mellon University published a paper proposing a learning approach called competitive reinforcement learning using visual transfer. This method, with the help of asynchronous advantage actor critic (A3C) architecture, generalizes a target game using an agent trained on a source game in Atari. What is A3C architecture? The A3C architecture is an asynchronous variant of the actor-critic model, in which the actor takes in the current environment state and determines the best action to take from there. It consists of four convolutional layers, an LSTM layer, and two fully connected layers to predict actions and value functions of the states. In this architecture, multiple worker agents are trained in parallel, each with their own copy of the model and environment. Advantage here refers to a metric that is set to judge how good the agents' actions were. What is the aim of competitive reinforcement learning? The learning approach introduced in this paper aims to use a reinforcement agent to generalize between two related but vastly different Atari games like Pong-v0 and Breakout-v0. This is done by learning visual mappers: given a frame from the source game, we should be able to generate the analogous frame in the target game. In both these games, a paddle is controlled to hit a ball to obtain a certain objective. Using this method the six actions of Pong-v0 {No Operation, Fire, Right, Left, Right Fire, Left Fire} are mapped to the four actions of Breakout-v0 as {Fire, Fire, Right, Left, Right, Left} respectively. The rewards are mapped directly from source game to target game without any scaling. The source and target environment they experimented on was OpenAI gym. They found underlying similarities between the source and the target game to represent common knowledge using Unsupervised Image-to-image Translation (UNIT) Generative adversarial networks (GANs). The target game competes with its visual representation obtained after using the UNIT GAN as a visual mapper between the source and target game. How competitive reinforcement learning works? The following diagram depicts how knowledge is transferred from source game to target game by competitively and simultaneously fine-tuning the model using two different visual representations of the target game: Source: arXiv First stage: The baseline A3C network is trained for source game (Pong-v0) in the first stage of the training process. The knowledge learned is then transferred from this model to learning to play target game (Breakout-v0). The efficiency of transfer learning method in terms of training time and data efficiency across parallel actor-learners is measured. Second stage: In this stage of the training process, two representations of the target game are used amongst the workers in parallel. The first representation of transfer process uses the target game frames taken directly from the environment. The second representation of transfer process uses the frames learned from the visual mapper (visual analogies between games). The ratio of number of workers that train directly on frames queried from the target game and frames mapped from the source game is a hyperparameter that is determined through experimentation. Results They concluded that it is possible to generate a visual mapper for semantically similar games with the use of UNIT GANs. Learning from two different representations of the same game and using them simultaneously for transfer learning stabilizes the learning curve. Although the workers using representations of the target game obtained from the visual mappers did not perform well in a stand alone setting, they showed improvements when used for the competitive learning. To read more about this learning approach and its efficiency, check out this research paper published by Akshita Mittel, Purna Sowmya Munukutla, and Himanshi Yadav: Visual Transfer between Atari Games using Competitive Reinforcement Learning. “Deep meta reinforcement learning will be the future of AI where we will be so close to achieving artificial general intelligence (AGI)”, Sudharsan Ravichandiran This self-driving car can drive in its imagination using deep reinforcement learning Dopamine: A Tensorflow-based framework for flexible and reproducible Reinforcement Learning research by Google