Tech News

Google Cloud Next conference in San Francisco in July 2018 saw some exciting new developments in the field of serverless technology. The company is giving development teams the ability to build apps without worrying about managing servers with their new serverless technology. Bringing the best of both worlds: Serverless and containers, Google announced that Cloud Functions is now generally available and ready for production use. Here is a list of the all-new features that developers can watch out for- #1 Write Cloud Functions using  Node 8, Python 3.7 With support for async/await and a new function signature, you can now write Cloud Functions using Node 8. Dealing with multiple asynchronous operations is now easier thanks to Cloud Functions that provide data and context. You can use the await keyword to await the results of asynchronous operations. Python 3.7 can also be used to write Cloud Functions.  Similar to Node, you get data and context for background functions, and request for HTTP. Python HTTP functions are based on the popular Flask microframework. Flask allows you to get set up really fast. The requests are based on flask.Request and the responses just need to be compatible with flask.make_response. As with Node, you get data (dict) with Python background functions and context (google.cloud.functions.Context). To signal completion, you just need to return from your function or raise an exception and Stackdriver error handling will kick in. And, similarly to Node (package.json), Cloud Functions will automatically do the installation of all of your Python dependencies (requirements.txt) and build in the cloud. You can have a look at the code differences between Node 6 and Node 8 behavior and at a Flask request on the Google Cloud website. #2 Cloud Functions is now out  for Firebase Cloud Functions for Firebase is also generally available. It has full support for Node 8, including ECMAScript 2017 and async/await. The additional granular controls include support  for runtime configuration options, including region, memory, and timeout. Thus allowing you to refine the behavior of your applications. You can find more details from the Firebase documentation. Flexibility for the application stack now stands improved. Firebase events (Analytics, Firestore, Realtime Database, Authentication) are directly available in the Cloud Functions Console on GCP. You can now trigger your functions in response to the Firebase events directly from your GCP project. #3 Run headless Chrome by accessing system libraries Google Cloud functions have also broadened the scope of libraries available by rebasing the underlying Cloud Functions operating system onto Ubuntu 18.04 LTS. Access to system libraries such as ffmpeg and libcairo2 is now available- in addition to imagemagick- as well as everything required to run headless Chrome. For example, you can now process videos and take web page screenshots in Chrome from within Cloud Functions. #4 Set environment variables You can now pass configuration to your functions by specifying key-value pairs that are bound to a function. The catch being, these pairs don’t have to exist in your source code. Environment variables are set at the deploy time using the --set-env-vars argument. These are then injected into the environment during execution time. You can find more details on the Google cloud webpage. #5 Cloud SQL direct connect Now connect Cloud Functions to Cloud SQL instances through a fully managed secure direct connection.  Explore more from the official documentation. What to expect next in Google Cloud Functions? Apart from these, Google also promises a range of features to be released in the future. These include: 1. Scaling controls This will be used to limit the number of instances on a per-function basis thus limiting traffic. Sudden traffic surge scenarios will , therefore,come under control when Cloud Functions rapidly scales up and overloads a database or general prioritization based on the importance of various parts of your system. 2. Serverless scheduling You’ll be able to schedule Cloud Functions down to one-minute intervals invoked via HTTP(S) or Pub/Sub. This allows you to execute Cloud Functions on a repeating schedule. Tasks like daily report generation or regularly processing dead letter queues will now pick up speed! 3. Compute Engine VM Access Now connect to Compute Engine VMs running on a private network using --connected-vpc option. This provides a direct connection to compute resources on an internal IP address range. 4. IAM Security Control The new Cloud Functions Invoker IAM role allows you to add IAM security to this URL. You can control who can invoke the function using the same security controls as used in Cloud Platform 5. Serverless containers With serverless containers, Google provides the same infrastructure that powers Cloud Functions. Users will now be able to simply provide a Docker image as input. This will allow them to deploy arbitrary runtimes and arbitrary system libraries on arbitrary Linux distributions This will be done while still retaining the same serverless characteristics as Cloud Functions. You can find detailed information about the updated services on Google Cloud’s Official page. Google Cloud Next: Fei-Fei Li reveals new AI tools for developers Google Cloud Launches Blockchain Toolkit to help developers build apps easily Zeit releases Serverless Docker in beta

Google, yesterday, released an archive of political ads purchased on its platforms. The new library of political ads reveals how much money is spent on these ads across different states and congressional districts, along with a list of top advertisers. Political ads feature federal candidates or currently elected federal officeholders. Google has been modifying its transparency report by adding different sections over the years due to European privacy laws, encryption adoption on websites i.e. HTTPS, among other evolving policy and user expectations. Read also: EU slaps Google with $5 billion fine for the Android antitrust case The latest archive is another newly added section in the company's regular transparency report This report shares data revealing “how the policies and actions of governments and corporations affect privacy, security, and access to information. This is Google’s efforts to make things more transparent when it comes to online political advertisements. Now, for any advertiser purchasing election ads on Google in the U.S., they have to “provide a government-issued ID and other key information that confirms they are a U.S. citizen or lawful permanent resident, as required by law. We also required that election ads incorporate a clear “paid for by” disclosure”, says Google. The new election ad library is searchable, downloadable and provides information about the ads with the highest views, the latest election ads running on our platform, and specific advertisers’ campaigns. The data from the Ad Library is publicly available on Google Cloud’s BigQuery. This data is particularly helpful for researchers, political watchdog groups and private citizens as they can leverage this data to develop charts, graphs, tables or other visualizations of political ads on Google Ads services. Apart from Google, Facebook and Twitter are other tech giants, who launched ad archives in recent months. Twitter ad archives are a part of the company’s increased transparency efforts. “We clearly label and show disclaimer information for federal political campaigning ads,” says Twitter. Facebook has been under a lot of controversy regarding advertisements, especially after an outcry over Russians’ alleged purchase of political ads during the 2016 elections. Also, A.G., Bob Ferguson, last month, proved Facebook guilty of providing discriminatory advertisements on its platform. Facebook, now has its own political ad archive that shows information about who paid for these ads along with other details. Google seems to be following Twitter and Facebook’s footsteps when it comes to political and issue-based advertising on its platform. Whether this comes at a right time, with the election season coming up soon, is another matter to be debated.   The new database is updated every week and anyone can see the newly uploaded ads and the advertisers uploading these ads. Google mentioned in their blog that despite the Ad Library providing many new insights, it’s still “working with experts in the U.S. and around the world to explore tools that capture a wider range of political ads—including ads about political issues (beyond just candidate ads), state and local election ads, and political ads in other countries”. Google’s aim with this is to protect these campaigns from digital attacks. “We hope this provides unprecedented, data-driven insights into election ads on our platform,” says Google. For more information regarding Google’s new political ad archive, check out the official Google blog post. Facebook must stop discriminatory advertising in the US, declares Washington AG, Ferguson Google’s new facial recognition patent uses your social network to identify you! Google is missing out $50 million because of Fortnite’s decision to bypass Play Store

At the 26th Annual DEFCON Conference in Vegas last week, attendees were reminded of US election infrastructure being susceptible to ulterior motives, by an alarming video posted on Twitter. https://twitter.com/RachelTobac/status/1029449569266884608 Rachel Tobac, CEO of SocialProof Security demonstrated on her Twitter status about the voting machines hacked in under two minutes. SocialProof Security provides assessments for social engineering based security. Social engineering involves tricking people into giving up information that lets hackers bypass physical and computer security systems. It’s most commonly done with a simple phone call, talking to a tech support agent into resetting a password or getting information about a company’s network by asking an unwary staffer few leading questions. Tobac explained that accessing the voting machine’s admin function is synonymous toopening the hood of a car with a release button, unplugging the card reader, picking a lock and turning on a machine with a ballpoint pen. The model of voting machine used was the Premier AccuVote TS or TSX which is used in more than 18 states for elections. Jack Braun, organizer of the Voting Village commented to the Wall Street Journal, “This is not the cyber mature industry.” While the National Association of Secretaries of State, one of the biggest providers of election supplies in the US, issued a statement discrediting the hackers: “Our main concern with the approach taken by DEFCON is that it uses a pseudo environment which in no way replicates state election systems, networks, or physical security,” it said. “Providing conference attendees with unlimited physical access to voting machines,” NASS said, “does not replicate accurate physical and cyber protections established by state and local governments before and on Election Day.” This is the second year in a row where DEFCON have hacked election systems with the Voting Village. Other experiments included an 11 year girl old hacking a replica of Florida secretary of state website and changing the results in 10 minutes. There were suggestions to use blockchain based voting systems to maintain the integrity of elections. Regardless of its implementation this is an area of concern and should be addressed to alleviate tampering of future elections. 7 Black Hat USA 2018 conference cybersecurity training highlights: Hardware attacks, IO campaigns, Threat Hunting, Fuzzing, and more DCLeaks and Guccifer 2.0: How hackers used social engineering to manipulate the 2016 U.S. elections Twitter allegedly deleted 70 million fake accounts in an attempt to curb fake news

Microsoft has added new features in Microsoft Edge Beta, which is the testing version of Microsoft Edge. Microsoft Edge is a browser for all your devices, giving you a continuous browsing experience across all the devices. Microsoft is constantly adding new features to Edge giving readers an organized, secure, and fast user experience. Recently they rolled out features like, intelligent visual search, support for paste and go option in the search bar, along with some performance improvements in Microsoft Edge for iOS. What are the new updates in Microsoft Edge Beta? To continue in the “best browser race”, they have added more features in Microsoft Edge Beta on iOS: You can now access both personal and work/school accounts simultaneously. This is enabled by a toggle to keep the respective browsing histories separate. A new option is added to receive breaking news alerts A new developer options menu Now iPad users can use the Command key on their attached keyboard to see shortcuts A What's new and tips link is added in settings Icons for the Favorites, Reading List, Books, and History in the Hub is now moved to the bottom You can now reorganize the Favorites folder Users can view the book annotations except for PDF free-form inking Performance improvements and bug fixes! You can get these features on your iOS device if you have signed up for Apple’s TestFlight, a service that makes it easy to invite users to test your apps and collect valuable feedback before you release them on the App Store. You can find the regular version of Microsoft Edge and follow the “What’s New” section to know if these features are rolled out! Microsoft Edge introduces Web Authentication for passwordless web security Microsoft Azure’s new governance DApp: An enterprise blockchain without mining Microsoft launches Quantum Katas, a programming project to learn Q#, its Quantum programming language

The Mozilla team announced Rust 2018 Edition Preview 2 today, the final release cycle before Rust 2018 goes into beta. The new release explores features such as the cargo fix command, NLL, along with other changes and improvements. Rust is a systems programming language by Mozilla. It is a "safe, concurrent, practical language", which supports functional and imperative-procedural paradigms. Rust provides better memory safety while still maintaining the performance. Let’s have a look at major features in Rust’s 2018 edition preview 2. The cargo fix command now comes as a built-in feature in Rust. This command is used during migration and addition of this new feature in Rust now further streamlines the migration process. Speaking of migration, extensive efforts have gone into improving and polishing the lints which help you migrate smoothly. Apart from that, the module system changes are now broken into several smaller features that help with independent tracking issues. There is no need of mod.rs anymore for parent modules Also, the extern crate is not needed anymore for including dependencies. Support has been provided for crate as a visibility modifier.   Another new addition in the Rust 2018 edition preview 2 is that NLL or Non-lexical lifetimes has been enabled by default, in migration mode. NLL improves the Rust compiler's ability to reason about lifetimes. It removes most of the remaining cases where people commonly experience the borrow checker rejecting valid programs. As per the Rust team, “If your code is accepted by NLL, then we accept it -- if your code is rejected by both NLL and the old borrow checker, then we reject it-- If your code is rejected by NLL but accepted by the old borrow checker, then we emit the new NLL errors as warnings”. In-band lifetimes have been split up in the latest release. Both rustfmt and the RLS have reached 1.0 “release candidate” status. For more information, check out the official release notes. Multithreading in Rust using Crates [Tutorial] Rust and Web Assembly announce ‘wasm-bindgen 0.2.16’ and the first release of ‘wasm-bindgen-futures’ Warp: Rust’s new web framework for implementing WAI (Web Application Interface)  

“Security, customizability, flexibility and cost are a few of the benefits of open-source software for developers. They’ll get all these and more from NVIDIA’s Material Definition Language software development kit.” —NVIDIA Blog. The material definition language (MDL) is a programming language used to define and render physical materials. This includes the creation of a wide range of physical materials such as woods, fabrics, translucent plastics and more. It is a set of tools integrating the precise look and feel of real-world materials into rendering applications. It gives users the freedom to share these materials between applications that support them. Now users can move a library of these materials between applications without worrying about them losing their appearance. This will allow developers to focus on building their applications. Allegorithmic had already built an entire MDL authoring tool, but with NVIDIA making the MDL SDK open source, developers get a deeper unrestricted access to the entire spec. The Blog states that Unreal Studio 4.20 now offers native support for MDL. “Being able to use a single material definition, like NVIDIA’s MDL, across multiple applications and render engines is a huge benefit to the end-user,” said Ken Pimentel, senior product manager of the Enterprise team at Epic Games. “Now that we’ve added MDL support to Unreal Studio, our enterprise customers can see their material representations converted to real time in Unreal Engine without baking every parameter. This means their creative intent can be carried to new forms of expression.” The MDL SDK API is C++ based and used for integration and customization tasks. It can be loaded dynamically and linked to visualization applications. The API also allows applications to load MDL modules, and analyze and understand the structure of a material. Therefore it can build a UI for editing materials then rendering the results. Some of the features in MDL SDK include: Can be used on GPU as well as CPU Database view on the imported MDL package space MDL editing C++ component-based API, and plugin architecture for extensibility MDL SDK supports Windows (only 64-bit), Linux, and macOS. To know more, visit the NVIDIA website and to get started here is the GitHub repository. Nvidia unveils a new Turing architecture: “The world’s first ray tracing GPU” Nvidia and AI researchers create AI agent Noise2Noise that can denoise images Nvidia GPUs offer Kubernetes for accelerated deployments of Artificial Intelligence workloads

Rigetti computers are committed to building the world’s most powerful computers and they believe the true value of quantum will be unlocked by practical applications. Rigetti CEO Chad Rigetti, posted recently on Medium about their plans to deploy 128 qubit chip quantum computing system, challenging Google, IBM, and Intel for leadership in this emerging technology. They have planned to deploy this system in the next 12 months and shared their investment in resources at the application layer to encourage experimentation on quantum computers. Over the past year, Rigetti has built 8-qubit and 19-qubit superconducting quantum processors, which are accessible to users over the cloud through their open source software platform Forest. These chips have been useful in helping researchers around the globe to carry out and test programs on their quantum-classical hybrid computers. However, to drive practical use of quantum computing today, Rigetti must be able to scale and improve the performance of the chips and connect them to the electronics on which they run . To achieve this, the next phase of quantum computing will require more power at the hardware level to drive better results. Rigetti is in a unique position to solve this problem and build systems that scale. Chad Rigetti adds, “Our 128-qubit chip is developed on a new form factor that lends itself to rapid scaling. Because our in-house design, fab, software, and applications teams work closely together, we’re able to iterate and deploy new systems quickly. Our custom control electronics are designed specifically for hybrid quantum-classical computers, and we have begun integrating a 3D signaling architecture that will allow for truly scalable quantum chips. Over the next year, we’ll put these pieces together to bring more power to researchers and developers.” While they are focussed on building the 128 qubit chip, the Rigetti team is also looking at ways to enhance the application layer by pursuing quantum advantage in three areas; i.e. quantum simulation, optimization and machine learning. The team believes quantum advantage will be achieved by creating a solution that is faster, cheaper and of a better quality. They have posed an open question as to which industry will build the first commercially useful application to add tremendous value to researchers and businesses around the world. Read the full coverage on the Rigetti Medium post. Quantum Computing is poised to take a quantum leap with industries and governments on its side Q# 101: Getting to know the basics of Microsoft’s new quantum computing language PyCon US 2018 Highlights: Quantum computing, blockchains and serverless rule!

Amidst the discussions going on around social media websites regulating their content or facing legal actions, Twitter CEO Jack Dorsey announced plans to rethink the core of how Twitter works. In an interview with the Washington Post, Dorsey said,  that he is experimenting with features that would promote alternative viewpoints in Twitter’s timeline to address misinformation and reduce echo chambers. “The most important thing that we can do is we look at the incentives that we’re building into our product,” Dorsey said. “Because they do express a point of view of what we want people to do — and I don't think they are correct anymore.” https://twitter.com/jack/status/1029846451524960261 Dorsey’s move is a clear indication of the fact that Silicon Valley leaders are getting serious about improving safety, security, and privacy across their services. In recent months, Twitter has made several moves to combat fake news and other data related scandals. Earlier this month, Apple, Facebook, and Spotify took action against Alex Jones. Initially, Twitter allowed Jones to continue using its service. But on Tuesday, Twitter imposed a seven-day “timeout” on Jones after he encouraged his followers to get their “battle rifles” ready against critics in the “mainstream media” and on the left. Last month, the social media giant allegedly deleted 70 million fake accounts in an attempt to curb fake news. It has been constantly suspending fake accounts which are inauthentic, spammy or created via malicious automated bots. Another solution Twitter is exploring is to surround false tweets with factual context. Dorsey said, that more context about a tweet, including tweets that call it out as obviously fake could help people make judgments for themselves. It is planning to label automated accounts; Legislators and federal lawmakers have already proposed putting such requirements into law. The social media website is also auditing existing accounts for signs of automated sign-up and improving the overall sign-up process. What is left to see now is whether Twitter can actually effectively implement these claims. Or Dorsey’s statements will go down the drain. You can read Dorsey’s entire interview on the Washington Post. How to stay safe while using Social Media Facebook plans to use Bloomsbury AI to fight fake news YouTube has a $25 million plan to counter fake news and misinformation

Oracle has released GraphPipe, an open source tool to simplify and standardize deployment of Machine Learning (ML) models easier. Development of ML models is difficult, but deploying the model for the customers to use is equally difficult. There are constant improvements in the development model but, people often don’t think about deployment. This is were GraphPipe comes into the picture! What are the key challenges GraphPipe aims to solve? No standard way to serve APIs: The lack of standard for model serving APIs limits you to work with whatever the framework gives you. Generally, business application will have an auto generated application just to talk to your deployed model. The deployment situation becomes more difficult when you are using multiple frameworks. You’ll have to write custom code to create ensembles of models from multiple frameworks. Building model server is complicated: Out-of-the-box solutions for deployment are very few because deployment gets less attention than training. Existing solution not efficient enough: Many of the currently used solutions don't focus on performance, so for certain use cases they fall short. Here’s how the current situation looks like: Source: GraphPipe’s User Guide How GraphPipe solves these problems? GraphPipe uses flatbuffers as the message format for a predict request. Flatbuffers are like google protocol buffers, with an added benefit of avoiding a memory copy during the deserialization step. A request message provided by the flatbuffer definition includes: Input tensors Input names Output names The request message is then accepted by the GraphPipe remote model and returns one tensor per requested output name, along with metadata about the types and shapes of the inputs and outputs it supports. Here’s how the deployment situation will look like with the use of GraphPipe: Source: GraphPipe’s User Guide What are the features it comes with? Provides a minimalist machine learning transport specification based on flatbuffers, which is an efficient cross platform serialization library for C++, C#, C, Go, Java, JavaScript, Lobster, Lua, TypeScript, PHP, and Python. Comes with simplified implementations of clients and servers that make deploying and querying machine learning models from any framework considerably effortless. It's efficient servers can serve models built in TensorFlow, PyTorch, mxnet, CNTK, or Caffe2. Provides efficient client implementations in Go, Python, and Java. Includes guidelines for serving models consistently according to the flatbuffer definitions. You can read plenty of documentation and examples at https://oracle.github.io/graphpipe. The GraphPipe flatbuffer spec can be found on Oracle's GitHub along with servers that implement the spec for Python and Go. Oracle reveals issues in Object Serialization. Plans to drop it from core Java. What Google, RedHat, Oracle, and others announced at KubeCon + CloudNativeCon 2018 Why Oracle is losing the Database Race

Intel's’ chips have been struck with yet another significant flaw called ‘Foreshadow’. This flaw, alternatively called as L1 Terminal Fault or L1TF, targets Intel’s Security Guard Extensions (SGX) within its Core chips. The US government’s body for computer security testified that an attacker could take advantage of this vulnerability in Intel’s chips to obtain sensitive information. This security flaw affects processors released right from 2015. Thankfully,  Intel has released a patch to combat the problem. Check the full list of affected hardware on Intel's website. While Intel confirmed that they are not aware of reports that any of these methods have been used in real-world exploits, the tech giant is now under scrutiny. This was bound to happen as Intel strikes a  hattrick following two similar attacks - Spectre and Meltdown - that were discovered earlier this year in January. Intel confirms that future processors would be built in such a way as to not be affected by Foreshadow. How does Foreshadow affect your data? The flaw was first brought to Intel’s notice by researchers from KU Leuven University in Belgium and others from the universities of Adelaide and Michigan. Foreshadow can exploit various flaws in a computing technique known as speculative execution. It can specifically target a lock box within Intel’s processors. This would let a hacker leak any data desired. To give you a gist, a  processor can run more efficiently by guessing the next operation to be performed. A correct prediction will save resources, while work based on an incorrect prediction gets scrapped. However, the system leaves behind clues like how long it will take the processor to fulfill a certain request. This can be used by an attacker to find weaknesses, ultimately gaining the ability to manipulate what path the speculation takes. Thus, hacking into the data at opportune moments that leaks out of a process's data storage cache. Speculative execution is important to guard against, because an attacker could use them to access data and system privileges meant to be off-limits. The most intriguing part of the story, as stated by hardware security researcher and Foreshadow contributor Jo Van Bulck is,  “Spectre is focused on one speculation mechanism, Meltdown is another, and Foreshadow is another”.   "This is not an attack on a particular user, it’s an attack on infrastructure."                          YUVAL YAROM, UNIVERSITY OF ADELAIDE   After the discovery of Spectre and Meltdown, the researchers found it only too fitting to look for speculative execution flaws in the SGX enclave. To give you an overview, Security Guard Extensions, or SGX, were originally designed to protect code from disclosure or modification. SGX is included in 7th-generation Core chips and above, as well as the corresponding Xeon generation. It remains protected even when the BIOS, VMM, operating system, and drivers are compromised. Meaning that an attacker with full execution control over the platform can be kept away. SGX, allows programs to establish secure enclaves on Intel processors. These are regions of a chip that are restricted to run code that the computer's operating system can't access or change. The creates a safe space for sensitive data,. Even if the main computer is compromised by malware, the sensitive data remains safe. That apparently isn’t totally the case. Wired furthers stress on the fact that the Foreshadow bug could break down the walls between virtual machines, a real concern for cloud companies whose services share space with other theoretically isolated processes. Watch this youtube video for more clarity on how foreshadow works. https://www.youtube.com/watch?v=ynB1inl4G3c&feature=youtu.be The Quick Fix to Foreshadow Prior to details of the flaw being made public, Intel had created its fix and coordinated its response with the researchers on Tuesday. The fix disables some of chips features that were vulnerable to the attack. Along with software mitigations, the bug will also be patched at the hardware level with Cascade Lake, an upcoming Xeon chip, as well as future Intel processors expected to launch later this year. This mitigation limits the extent to which the same processor can be used simultaneously for multiple tasks, and hence companies running cloud computing platforms could see a significant hit to their collective computing power. On Tuesday, cloud services companies - Amazon, Google and Microsoft - said they had put in place a fix for the problem. Intel is working with these cloud providers—where uptime and performance is key—to “detect L1TF-based exploits during system operation, applying mitigation only when necessary,” Leslie Culbertson, executive vice president and general manager of Product Assurance and Security at Intel, wrote. Individual computer users are advised, as ever, to download and install any software updates available. The research team confirmed that is was unlikely that individuals would see any performance impact. As long as you’re system is patched up, you should be okay. Check out PCWorld’s guide on how to protect your PC against Meltdown and Spectre. You can also head over to the Red Hat Blog for more knowledge on Foreshadow. NetSpectre attack exploits data from CPU memory Intel’s Spectre variant 4 patch impacts CPU performance 7 Black Hat USA 2018 conference cybersecurity training highlights

The SQLite community has released a sneak peek to what users can expect in the upcoming version, SQLite 3.25.0, which could be released next month. The SQLite 3.25.0 draft which the community published on its official website yesterday includes a list of some upcoming features and bug fixes. The primary update being support for windows functions and improvements in the query optimizer. Expectations from SQLite 3.25.0 Support for windows functions will be added This release will bring in an added window function support. Prior to this, SQLite developers used the PostgreSQL window function documentation as their primary reference for how window functions ought to behave. The community has carried out several test cases against PostgreSQL to ensure that window functions operate the same way in both SQLite and PostgreSQL. Improvements in the Query Optimizer Unnecessary loads of columns in an aggregate query are avoided. These columns are neither within an aggregate function nor a part of the GROUP BY clause. The IN-early-out optimization: When doing a look-up on a multi-column index, an IN operator is used on a column other than the left-most column. If no rows match against the first IN value, one should check the existence of rows that match the columns to the right before continuing with the next IN value. Transitive property can be used to propagate constant values within the WHERE clause. For example, convert "a=99 AND b=a" into "a=99 AND b=99". Separate mutex on every inode In the SQLite 3.25.0, one can use a separate mutex on every inode in the Unix VFS, rather than a single mutex shared among them all. This results in better concurrency in multi-threaded environments. Improvised PRAGMA integrity_check command The PRAGMA integrity_check command will be enhanced for improved detection of problems on the page freelist. The integrity_check pragma looks for out-of-order records, missing pages, malformed records, missing index entries, and UNIQUE, CHECK, and NOT NULL constraint errors. .dump command infinity output This version will showcase the infinity output as 1e999 in the ".dump" command of the command-line shell. Bug fixes in the upcoming version SQLite 3.25.0 A fix for ticket 79cad5e4b2e219dd197242e9e On an UPSERT when the order of constraint checks is rearranged, ensure that the affinity transformations on the inserted content occur before any of the constraint checks. Fix for ticket 7be932dfa60a8a6b3b26bcf76 Avoid using a prepared statement for ".stats on" command of the CLI after it has been closed by the ".eqp full" logicc.. To know more about SQLite Release 3.25.0 visit its release log draft. How to use SQLite with Ionic to store data? Introduction to SQL and SQLite NHibernate 3.0: Testing Using NHibernate Profiler and SQLite

With the launch of Unity 2018.2 release last month, Unity is finally making the switch to WebAssembly as their output format for the Unity WebGL build target. WebAssembly support was first teased in Unity 5.6 as an experimental feature. Unity 2018.1 marked the removal of the experimental label. And finally in 2018.2, Web Assembly replaces asm.js as the default linker target. Source: Unity Blog WebAssembly replaced asm.js because it is faster, smaller and more memory-efficient, which are all pain points of the Unity WebGL export. A WebAssembly file is a binary file (which is a more compact way to deliver code), as opposed to asm.js, which is text. In addition, code modules that have already been compiled can be stored into an IndexedDB cache, resulting in a really fast startup when reloading the same content. In WebAssembly, the code size for an empty project is ~12% smaller or ~18% if 3D physics is included. Source: Unity Blog WebAssembly also has its own instruction set. In Development builds, it adds more precise error-detection in arithmetic operations. In non-development builds, this kind of detection of arithmetic errors is masked, so the user experience is not affected. Asm.js added a restriction on the size of the Unity Heap; its size had to be specified at build-time and could never change. WebAssembly enables the Unity Heap size to grow at runtime, which lets Unity content memory-usage exceed the initial heap size. Unity is now working on multi-threading support, which will initially be released as an experimental feature and will be limited to internal native threads (no C# threads yet). Debugging hasn’t got any better. While browsers have begun to provide WebAssembly debugging in their devtools suites, these debuggers do not yet scale well to Unity3D sizes of content. What’s next to come Unity is still working on new features and optimizations to improve startup times and performance: Asynchronous instantiation Structured cloning, which allows compiled WebAssembly to be cached in the browser Baseline and tiered compilation, to speed-up instantiation Streaming instantiation to compile Assembly code while downloading it Multi-Threading You can read the full details on the Unity Blog. Unity 2018.2: Unity release for this year second time in a row! GitHub for Unity 1.0 is here with Git LFS and file locking support What you should know about Unity 2018 Interface

The last couple of months have seen TensorFlow releases coming thick and fast. Clearly, the Google team are working hard to ship new updates for a framework that seems to be defining deep learning as we know it. But TensorFlow 2.0 remains on the horizon - and that, really, is the release we've all been waiting for. Amid speculation and debate, we now have the first inkling of what we can expect thanks to a post by Google Brain engineer Martin Wicke. In a somewhat unassuming post on Google Groups, Wicke said that work was underway on TensorFlow 2.0, with a preview version expected later this year. The big changes that the team are working towards include: Making TensorFlow easier to learn and use by putting eager execution (TensorFlow's programming environment) at the center of the new release Support for more platforms and languages Removing deprecated APIs How you can support the TensorFlow 2.0 design process Wicke writes that TensorFlow 2.0 still needs to go through a public review process. To do this, the project will be running a number of public design reviews that run through the proposed changes in detail and give users the opportunity to give feedback and communicate their views. What TensorFlow 2.0 means for the TensorFlow project Once TensorFlow 2.0 is released migration will be essential - Wicke explains that "We do not anticipate any further feature development on TensorFlow 1.x once a final version of TensorFlow 2.0 is released" and that the project "will continue to issue security patches for the last TensorFlow 1.x release for one year after TensorFlow 2.0’s release date." The end of tf.contrib? TensorFlow 2.0 will bring an end (of sorts) to tf.contrib, the repository where code contributed to TensorFlow sits, waiting to be merged. "TensorFlow’s contrib module has grown beyond what can be maintained and supported in a single repository." Wicke writes. "Larger projects are better maintained separately, while we will incubate smaller extensions along with the main TensorFlow code." However, Wicke promises that TensorFlow will help the owners of contributed code to migrate appropriately. Some modules could be integrated into the core project, others moved into another, separate repository, and others simply removed entirely. If you have any questions about TensorFlow 2.0 you can get in touch with the team directly by emailing discuss@tensorflow.org.  TensorFlow has also set up a mailing list for anyone interested in regular updates - simply subscribe to developers@tensorflow.org. Read next Why Twitter (finally!) migrated to Tensorflow Python, Tensorflow, Excel and more – Data professionals reveal their top tools Can a production ready Pytorch 1.0 give TensorFlow a tough time?

Zeit, the organization behind the cloud deployment software Now, yesterday launched Serverless Docker in beta. The concept was first discussed by the Zeit team at Zeit Day 2018 back in April, but it's now available to use and promises to radically speed up deployments for engineers. In a post published on the Zeit website yesterday, the team listed some of the key features of this new capability, including: An impressive 10x-20x improvement in cold boot performance (in practice this means cold boots can happen in less than a second A new slot configuration property that defines resource allocation in terms of CPU and Memory, allowing you to fit an application within the set of constraints that are most appropriate for it Support for HTTP/2.0 and WebSocket connections to deployments, which means you no longer need to rewrite applications as functions. The key point to remember with this release, according to Zeit, is that  "Serverless can be a very general computing model. One that does not require new protocols, new APIs and can support every programming language and framework without large rewrites." Read next: Modern Cloud Native architectures: Microservices, Containers, and Serverless – Part 1 What's so great about Serverless Docker? Clearly, speed is one of the most exciting things about serverless Docker. But there's more to it than that - it also offers a great developer experience. Johannes Schickling, co-founder and CEO of Prisma (a GraphQL data abstraction layer) said that, with Serverless Docker, Zeit "is making compute more accessible. Serverless Docker is exactly the abstraction I want for applications." https://twitter.com/schickling/status/1029372602178039810 Others on Twitter were also complimentary about Serverless Docker's developer experience - with one person comparing it favourably with AWS - "their developer experience just makes me SO MAD at AWS in comparison." https://twitter.com/simonw/status/1029452011236777985 Combining serverless and containers One of the reasons people are excited about Zeit's release is that it provides the next step in serverless. But it also brings containers into the picture too. Typically, much of the conversation around software infrastructure over the last year or so has viewed serverless and containers as two options to choose from rather than two things that can be used together. It's worth remembering that Zeit's product has largely been developed alongside its customers that use Now. "This beta contains the lessons and the experiences of a massively distributed and diverse user base, that has completed millions of deployments, over the past two years." Eager to demonstrate how Serverless Docker works for a wide range of use cases, Zeit has put together a long list of examples of Serverless Docker in action on GitHub. You can find them here. Read next A serverless online store on AWS could save you money. Build one. Serverless computing wars: AWS Lambdas vs Azure Functions

With the release of Alexa Auto Software Development Kit (SDK) the integration of Alexa into in-vehicle infotainment systems will become easier for the developers. Currently, the SDK assumes that automotive systems will have access to the cloud all the time, but it would be better if Alexa-enabled vehicles have some core functions even when they’re offline. This means that we need to reduce the size of the underlying machine-learning models, so they can fit in local memory. In this year’s Interspeech, Grant Strimel with his colleagues will present a new technique for compressing machine-learning models that could reduce their memory footprints by 94% while leaving their performance almost unchanged. What is the aim behind statistical model compression? Amazon Alexa and Google Assistant support skills built by external developers. These skills have Natural Language Understanding (NLU) models that extend the functionality of the main NLU models. Because there are numerous skills, their NLU models are loaded on demand only when they are needed to process a request. If the skill’s NLU model is large, loading them into memory adds significant latency to utterance recognition. To provide quick NLU response and good customer experience, small-footprint NLU models are important. Also, cloud-based NLU is unsuitable for local system deployment without appropriate compression because it has large memory footprints. To solve this, Grant and his colleagues have designed an algorithm which take large statistical NLU models and produce models which are equally predictive but have smaller memory footprint. What are the techniques introduced? Alexa’s NLU systems use several different types of machine learning models, but they all share some common traits. One common trait that Alexa’s NLU systems share is, extracting features (strings of text with particular predictive value) from input utterances. Another common trait is that each feature has a set of associated weights, which determines how large a role it should play in different types of computation. These weights of millions of features are stored, making the ML models memory intensive. Two techniques are proposed to perform statistical model compression: Quantization The first technique for compressing an ML model is to quantize the feature weights: Take the total range of weights Divide the range into equal intervals Finally, round each weight off to the nearest boundary value for its interval Currently, 256 intervals are used, allowing the representation of every weight in the model with a single byte of data, with minimal effect on the network’s accuracy. The additional benefit is that the low weights are discarded because they are rounded off to zero. Perfect Hashing In this technique we use hashing to perform mapping a particular feature to the memory locations of the corresponding weight. For example, play ‘Yesterday,’ by the Beatles,” we want our system to pull up the weights associated with the feature “the Beatles” — not the weights associated with “Adele”, “Elton John”, and the rest. To perform such mappings we will use hashing. A hash function is a function, which maps arbitrary sized inputs and maps them with fixed sized outputs and also have no predictable relationship to the inputs. One side effect of hashing is that it sometimes produces collisions, which means, two inputs that hash to the same output. The collision problem is addressed through perfect hashing: Source: Amazon Alexa We first assume that we have access to a family of conventional hash functions, all of which produce random hashes. For this we use the hash function MurmurHash, which can be seeded with a succession of different values. We represent the input strings to be hashed by N. We begin with an array of N 0’s and apply our first hash function called Hash1. We change a 0 in the array to a 1 for every string that yields a unique hash value. Next, a new array of 0’s is built for only the strings that yielded collisions under Hash1. We apply a different hash function to those strings. Similar to step 2, we toggle the 0’s to collision-free hashes. This process is repeated until every input string has a corresponding 1 in some array. All these arrays are then combined into one giant array. The position of a 1 in this giant array indicates the unique memory location assigned to the corresponding string. When the trained input receives an unseen input string, it will apply Hash1 to each of the input’s substrings and, if it finds a 1 in the first array, it goes to the associated address. If it finds a 0, it applies Hash2 and repeats the process. This process does causes a slight performance problem, but it’s a penalty that’s only paid when a collision occurs. To know more about the statistical model compression you can visit the Amazon Alexa page and also check out the technical paper by the researchers. Amazon Alexa and AWS helping NASA improve their efficiency Amazon Echo vs Google Home: Next-gen IoT war Diffractive Deep Neural Network (D2NN): UCLA-developed AI device can identify objects at the speed of light