Tech News

PostgreSQL announces JDBC 42.2.0, pg_back version 1.4, and  repmgr 4.0.2, new processor and tools, and more in today’s top stories around machine learning, deep learning,and data science news. 1. PostgreSQL releases JDBC 42.2.0, pg_back version 1.4, and  repmgr 4.0.2 PostgreSQL announced various releases in line, namely the PostgreSQL JDBC version 42.2.0, the pg_back version 1.4, and the 4.0.2 version of its tool repmgr. Let’s have a look at each of the release: PostgreSQL JDBC version 42.2.0: The changes within this JDBC version include: Support SCRAM-SHA-256 for PostgreSQL 10 in the JDBC 4.2 version (Java 8+) using the Ongres SCRAM library. PR 842 Make SELECT INTO and CREATE TABLE AS return row counts to the client in their command tags. Issue 958 PR 962 Support Subject Alternative Names for SSL connections. PR 952 Support isAutoIncrement metadata for PostgreSQL 10 IDENTITY column. PR 1004 Support for primitive arrays PR#887 3e0491a Implement support for get/setNetworkTimeout() in connections. PR 849 Make GSS JAAS login optional, add an option "jaasLogin" PR 922 To know more about this release visit the link here. 2. pg_back version 1.4 : pg_back is a simple bash shell script for PostgreSQL that can dump all your databases to files. It can operate on standby servers and takes care of pausing and resuming WAL(Write-Ahead Logging) replay. The 1.4 release brings support for PostgreSQL 10. To know more about the features visit the GitHub repo. 3. Repmgr version 4.0.2: repmgr is PostgreSQL’s most popular tool for replication and failover management. The repmgr 4.0.2 has some usability enhancements and some minor bug fixes. The enhancements are as follows: Recognize the -t/--terse option for repmgr cluster event to hide the Details column Add "--wait-start" option for repmgr standby register Add %p event notification parameter for repmgr standby switchover For a detailed list of the bug fixes visit the release notes here. 2. Microsoft releases Cumulative Update 10 for SQL Server 2014 SP2 Microsoft released a Cumulative Update 10 for SQL Server 2014 SP2, which is CU10 - 4052725. These Cumulative updates (CU) are now available at the Microsoft Download Center. This update includes fixes that were released after the release of SQL Server 2014 SP2. To know about these fixes in detail visit the Microsoft support blog. 3. Introducing PolySwarm: A Decentralized Cyber Threat Intelligence Market powered by Blockchain PolySwarm, a decentralized threat intelligence market has been designed to reset the threat caused by the current cybersecurity ecosystem. PolySwarm is backed by Ethereum smart contracts and blockchain technology. The PolySwarm platform: Defines a real-time threat detection ecosystem that encompasses enterprises, consumers, vendors and geographically diverse security experts. Will lower the barriers to entry, provide broader coverage options, discourage duplicative efforts and ensure interoperability among products and threat intelligence feeds. Benefits of PolySwarm include: A legitimate income stream for security experts Puts the user first by providing them with timely access to broad, crowdsourced security expertise and triage. No future data is required to accurately classify a threat. In order to confirm “ground truth” or third-party trustless authorization it takes into account real expertise. To read about the platform’s working in detail click on this link. 4. Videantis launches a new visual Processor and tools for Deep Learning Videantis, a German intellectual property supplier, launched its sixth-generation v-MP6000UDX visual processing architecture and v-CNNDesigner tool. The new v-MP6000UDX visual processing architecture : Adds deep learning capability to a solution that combines computer vision, image processing and video coding from a single unified SoC platform. Can be configured from a single media processor core to up to 256 cores, and is configurable based on the company’s own programmable DSP architecture. Includes an extended instruction set optimized for running convolutional neural nets (CNN), increases the multiply-accumulate throughput per core eightfold to 64 MACs per core, and extends the number of cores from typically 8 to up to 256. The v-CNNDesigner tool : Enables easy porting of neural networks that have been designed and trained using frameworks such as TensorFlow or Caffe. Analyzes, optimizes and parallelizes trained neural networks for efficient processing on the v-MP6000UDX architecture. Fully automates the task of implementing a neural network on the low power processing architecture. To read about both the architecture and the tool in detail visit the link given here. 5. Pegasus Solutions announces analytics solution to improve hoteliers' revenue and profits Pegasus Solutions plans to launch a new cloud-based business intelligence solution, Pegasus Analytics. The brand new analytics solution would be demonstrated at HEDNA Global Distribution Conference, which will be held on January 29-31 in Austin, TX. The analytics solution is designed to aid hotels to improve revenue, profitability and market share through a combination of visually rich, interactive context-driven dashboards, powerful data discovery capabilities, data and human knowledge. Pegasus Analytics: Supports revenue management and marketing goals by aggregating and pairing data collected and stored across different data sources, including CRS, PMS and Google Analytics, delivering actionable insights for impactful decision making. Eliminates the intensive tasks of accessing, inputting and analyzing data, and unlocks business-critical insights that hoteliers can immediately act upon. Enables hoteliers to communicate insights and data directly inside the platform through built-in collaboration and messaging features. This in turn reduces the time required to act upon the data. Hoteliers can now log in to analyze, interact and annotate historic and future performance data by channel, room type, rate, travel agent, consortia and other sources of business information.    

MicroStrategy’s new connectors, Kubernetes scales to 2500 nodes, TensorboardX, and more in today’s top stories around machine learning, deep learning,and data science news. 1. MicroStrategy announces new connectors to data discovery vendors Microsoft Power BI, Qlik Technologies, and Tableau Software MicroStrategy, a leading worldwide provider of enterprise analytics and mobility software announced new connectors at its annual user conference MicroStrategy World™ 2018, which began on 15th January, 2018. These brand new connectors help promote a more open and powerful analytics ecosystem, which enables business users of different discovery tools to bridge applications across their enterprise. Users of Power BI, Qlik, and Tableau can now connect to their MicroStrategy systems and experience the best of both worlds, i.e. the individual features they want from data discovery tools and a robust enterprise environment with governance, security, and scalability for their analytics applications. To know more, visit the link here. 2. OpenAI to scale Kubernetes to 2500 nodes OpenAI decides to scale Kubernetes to 2500 nodes. Kubernetes’ fast iteration cycle, reasonable scalability, and a lack of boilerplate makes it ideal for most of their experiments. OpenAI now operates several Kubernetes clusters (some in the cloud and some on physical hardware), the largest of which they have pushed to over 2,500 nodes. This cluster runs in Azure on a combination of D15v2 and NC24 VMs. To know more about how OpenAI scaled Kubernetes and the breakages caused, visit the link here. 3. Tensorboard-PyTorch plugin for graph visualization of your model The name TensorboardX means tensorboard for X. This means, the package can be used by other DL frameworks such as mxnet, chainer as well. The tensorboard-pytorch plugin is a module for visualization with Tensorboard. It allows you to watch tensors flow without actually using Tensorflow. The Tensorboard-PyTorch plugin: Helps you write Tensorboard events with a simple function call Supports five visualizations: scalars, images, audio, histograms, and the graph To know more about TensorboardX visit the Github repo. To know more about Tensorboard’s pytorch plugin visit the link here. 4. Distilled Analytics announces Behavioral Analytics Model for financial services Distilled Analytics today launched its Behavioral Analytics Model, a state-of-the-art, AI-driven data platform that aims to change how asset managers, banks, and nonbank financial institutions manage risk and grow revenue. The company predicts financial actions using a suite of tools to provide deep insights for financial professionals. The AI platform ingests unstructured and structured data, and analyzes complex patterns to calculate outcomes in areas such as economic development, public health, and identity theft. Some key offerings are: Distilled IDENTITY provides more accurate and timely anti-money laundering (AML) and know your customer (KYC) compliance that eases the burden of observing conflicting regulatory requirements, in addition to helping financial institutions better manage their client relationships. Distilled IMPACT delivers quantification of nonfinancial activities to enable asset growth of impact investing managers. Distilled CREDIT improves understanding of credit risk to enable expanded lending portfolios, identification of unserved or underserved customers, and reduced defaults. 5. Enterprise Blockchain all set to hit the markets this 2018 There have been a lot of enhancements coming in enterprise blockchain platforms. According to Coindesk, 2018 could be the year enterprise blockchain goes live and businesses can move from experimenting to production. In various customer conversations during the 2017, the challenges to moving successful successful PoCs to pilots and further to production boils down to five areas of enterprise-grade requirements: performance at scale operational resilience security and confidentiality supportability and management enterprise integration. To read about this news in detail visit the link here.

Buckle up, Guys. Jupyter Notebook version 5.3.0 is here! Jupyter Notebook, the popular language-agnostic HTML notebook application for Project Jupyter, is now available in version 5.3.0. The Notebook is an open-source web application for creating and sharing documents that contain live code, equations, visualizations and narrative text. It can be used for data cleaning and transformation, data visualization, and machine learning to name a few. The new version includes a myriad of bug fixes and changes, most notably terminal support for Windows. It also includes support for OS trash. So now the files deleted from the notebook dashboard are moved to the OS trash as opposed to being deleted permanently. Other changes include: A restart and run all button to the toolbar. Programmatic copy to clipboard is now allowed. DOM History API can be used for navigating between directories in the file browser. Translated files can now be added to folder(docs-translations). Token-authenticated requests cross-origin allowed by default. A “close” button is displayed on load notebook error. Action is added to command palette to run CodeMirror’s indentAuto on selection. A new option is added to specify extra services. Shutdown trans loss is now fixed. Finding available kernelspecs is now more efficient. The new version uses requirejs vs. require. It also fixes some ui bugs in firefox. It can now compare non-specific language code when choosing to use arabic numerals. Save-script deprecation is fixed. Moment locales in package_data are now included. The new version now has Use /files prefix for pdf-like files. The feature of adding a folder for document translation is now available. Users can now set the password, when login-in via token. Other minor changes can be found in the changelog. Users can upgrade to the latest release by pip install notebook --upgrade or conda upgrade notebook. It is recommended to upgrade to version 9+ of pip before upgrading notebook. Fun Fact: Jupyter is a loose acronym meaning Julia, Python, and R. These programming languages were the first target languages of the Jupyter application, but nowadays, the notebook also supports many other languages.  

Google’s Cloud AutoML, Uber AI Labs’ Neuroevolution and Safemutation, G Suite’s new security center, visual tools for Azure data factory, and more in today’s top stories around machine learning, and data science news.   1. Google’s Cloud AutoML lets businesses train Machine Learning models without having to code. Google has launched their Cloud AutoML service to make machine learning accessible to all enterprises. This service will help businesses with limited ML expertise start building their own high-quality custom models by using advanced techniques like learning2learn and transfer learning from Google. Here are a few advantages: Cloud AutoML provides businesses with a more accurate model even if the business has limited machine learning expertise. It provides a faster turnaround time to production-ready models. It is easy to use with a simple graphical user interface for specifying data. Their, first Cloud AutoML release would be Cloud AutoML Vision, a service for creating custom ML models for image recognition. With Vision, businesses can easily upload images, train and manage models, and then deploy them directly on Google Cloud. 2. Uber AI Labs open sources neuroevolution and safe mutation implementations to solve reinforcement learning problems Uber AI labs announced today that it is open sourcing two implementations, namely Neuroevolution and safemutations. These will make it easy for researchers to train their deep neural networks in order to solve reinforcement learning problems. Neuroevolution introduces a Deep genetic algorithm(GA), which involves a simple parallelization trick that allows training deep neural networks with GAs. These GAs are surprisingly competitive with popular algorithms for deep reinforcement learning problems, such as DQN, A3C, and ES, especially in the challenging Atari domain. Also, interesting algorithms developed in the neuroevolution community can now immediately be tested with deep neural networks, by showing that a Deep GA-powered novelty search can solve a deceptive Atari-scale game.  To know more about this in detail, read the research paper. Safe mutation (SM) operators aim to find a degree of change within the mutation operator itself that does not alter network behavior too much but still facilitates exploration. The safe mutation through gradients (SM-G) operator increases the ability of a simple genetic algorithm-based neuroevolution method to find solutions in high-dimensional domains. These domains are those which require deep or recurrent neural networks or domains that require processing raw pixels. By improving the ability to evolve deep neural networks, this new safer approach to mutation expands the scope of domains amenable to neuroevolution. To read more on this, visit the research paper. 3. G Suite launches a new security center with unified dashboard, analytics, & recommendations Google has introduced a new security center for G Suite. This tool brings together security analytics, actionable insights, and best practice recommendations from Google for protecting an organization and its data and users. It consists of three main parts. A unified dashboard, which will get insights into suspicious device activity and visibility into how spam and malware are targeting users. It will also provide metrics to demonstrate security effectiveness. Security analytics, where admins can examine analytics to flag threats. Security health, which reduces risk by providing security health recommendations. It analyzes existing security metrics and gives customized advice to secure users and data.   4. Microsoft announces visual tools for Azure Data Factory v2 Microsoft has added new visual tools for Azure Data Factory (ADF) v2. ADF v2 public preview was announced at Microsoft Ignite on Sep 25, 2017. ADF visual tools provide a simple and intuitive code free interface to drag and drop activities on a pipeline canvas, perform test runs, debug iteratively, and deploy and monitor pipeline runs. It also provides guided tours on how to use the enabled visual authoring & monitoring features including support for all control flow activities running on Azure computes. Moreover, users can validate their pipelines to know about missed property configurations or incorrect configurations. More details are available on their official blog. 5. A new release of SQL Operations Studio is now available Microsoft has announced the January release of their SQL Operations Studio. SQL Operations Studio was announced for public preview at Connect () conference in November 2017. Some major repo updates and feature releases include: HotExit feature is now enabled to automatically reopen unsaved files. Saved connections can now be accessed from Connection Dialog. SQL editor tab color now matches the Server Group color. Run Current Query command fixed. Broken pinned Windows Start Menu icon fixed. Saved Server connections are available in the Connection Dialog. Drag-and-drop breaking scripting bug is fixed. Missing Azure Account branding icon is added. For a complete list of updates, refer to the changelog.

Jupyter Notebook v5.3.0, Apache Solr 7.2.1 and Lucene 7.2.1, Kibana v5.6.6 and v6.1.2, Filament’s new blockchain software, and more in today’s top stories around machine learning, blockchain, and data science news. 1. The much-awaited Jupyter Notebook version 5.3.0 released Project Jupyter has released version 5.3.0 of their popular Jupyter Notebook. The Jupyter Notebook is an open-source web application for creating and sharing documents that contain live code, equations, visualizations and narrative text. The new version includes a myriad of bug fixes and changes, most notably terminal support for Windows and support for OS trash. So now the files deleted from the notebook dashboard are moved to the OS trash as opposed to being deleted permanently. Users can upgrade to the latest release by pip install notebook --upgrade or conda upgrade notebook. It is recommended to upgrade to version 9+ of pip before upgrading notebook. Full details of the release can be found in the changelog. 2. Apache Solr 7.2.1 and Apache Lucene 7.2.1 are now available Apache has released version 7.2.1 of Solr, their open source NoSQL search platform. Solr features include full-text search, hit highlighting, faceted search and analytics and, rich document parsing, to name a few. The new release includes 4 bug fixes since the last update. Overseer can never process some last messages. Rename core in solr standalone mode is not persisted. QueryComponent's rq parameter parsing no longer considers the defType parameter. NPE in SolrQueryParser fixed when the query terms inside a filter clause reduce to nothing. Apache have also announced the release of Lucene 7.2.1. Apache Lucene is a full-featured text search engine library written in Java. With this release, advanceExact on SortedNumericDocValues can be produced by Lucene54DocValuesProducer. Further details of changes are available in the changelog. 3. Kibana 5.6.6 and 6.1.2, the data visualization plugin for Elasticsearch, released Kibana, the popular data visualization plugin for Elasticsearch, has released two new versions 5.6.6 and 6.1.2. These releases of Kibana include an important security fix. The previous versions of Kibana (5.1.1 to 6.1.2 and 5.6.6) had a cross-site scripting (XSS) vulnerability via the colored fields formatter. This vulnerability could allow an attacker to obtain sensitive information from or perform destructive actions on behalf of other Kibana users. This issue has now been solved in the new version 6.1.2 and added to v5.6.6.  Both the releases are available on the downloads page. 4. Filament releases new software and Blocklet Chip hardware solutions based on Blockchain technology Filament, a provider of blockchain solutions for IoT and the enterprise, announces new software and Blocklet Chip hardware solutions, which are based on next-gen Blockchain technology. The new software and hardware solutions enable devices to securely interact and transact against a blockchain.   Filament’s new trusted application software and Blocklet Chip, which are currently in beta, are designed to communicate and interact with multiple blockchain technologies natively. The software is implemented on existing hardware and will deliver a secure distributed ledger technology solution. The Blocklet Chip will allow industrial corporations and enterprises to seamlessly extract the value of recording and monetizing data assets, at the edge of the network, on the sensors themselves. 5. Bitcoin Cash launches a Native Bitcoinj Development Branch Bitcoin Cash now has its own Bitcoinj development branch. Bitcoinj was the second Bitcoin implementation and the first to target SPV (Simplified Payment Verification) light wallet functionality. Most Android wallets make use of Bitcoinj. The goal of the new development branch is to unify all forks for BCH. The development team wants to expand its functionality and limit it not just on SPV wallets. This Branch will also allow more mobile wallets to support Bitcoinj Cash or switch over from BTC to BCH full-time. Plans to revive the Lighthouse project are also in the wake due to this new implementation. With BCH, crowdfunding through Bitcoin can also become viable again. The team also hopes to engage more community members to contribute to this particular branch of development.

OpenAI releases a python/Tensorflow package, Gradient checkpointing! Gradient checkpointing lets you fit 10x larger neural nets into memory at the cost of an additional 20% computation time. The tools within this package, which is a joint development of Tim Salimans and Yaroslav Bulatov, aids in rewriting TensorFlow model for using less memory. Computing the gradient of the loss by backpropagation is the memory intensive part of training deep neural networks. By checkpointing nodes in the computation graph defined by your model, and recomputing the parts of the graph in between those nodes during backpropagation, it is possible to calculate this gradient at reduced memory cost. While training deep feed-forward neural networks, which consists of n layers, we can reduce the memory consumption to O(sqrt(n)), at the cost of performing one additional forward pass. The graph shows the amount of memory used while training TensorFlow official CIFAR10 resnet example with the regular tf.gradients function and the optimized gradient function. To see how it works, let’s take an example of a simple feed-forward neural network. In the figure above, f : The activations of the neural network layers b : Gradient of the loss with respect to the activations and parameters of these layers All these nodes are evaluated in order during forward pass and in reversed order during backward pass. The results obtained for ‘f’ nodes are required in order to compute ‘b’ nodes. Hence, after the forward pass, all the f nodes are kept in memory, and can be erased only when backpropagation has progressed far enough to have computed all dependencies, or children, of an f node. This implies that in simple backpropagation, the memory required grows linearly with the number of neural net layers n. Graph 1: Vanilla Backpropagation The graph above shows a simple vanilla backpropagation, which computes each node once. However, recomputing the nodes can save a lot of memory. For this, we can simply try recomputing every node from the forward pass as and when required. The order of execution, and the memory used, then appear as follows: Graph 2: Backpropagation with poor memory By using the strategy above, the memory required to compute gradients in our graph is constant in the number of neural network layers n, which is optimal in terms of memory. However, now the number of node evaluations scales to n^2, which was previously scaled as n. This means, each of the n nodes is recomputed on the order of n times. As a result, the computation graph becomes much slower for evaluating deep networks. This makes the method impractical for use in deep learning. To strike a balance between memory and computation, OpenAI has come up with a strategy that allows nodes to be recomputed, but not too often. The strategy used here is to mark a subset of the neural net activations as checkpoint nodes. Source: Graph with chosen checkpointed node These checkpoint nodes are kept in memory after the forward pass, while the remaining nodes are recomputed at most once. After recomputation, the non-checkpoint nodes are stored in memory until they are no longer required. For the case of a simple feed-forward neural net, all neuron activation nodes are graph separators or articulation points of the graph defined by the forward pass. This means, the nodes between a b node and the last checkpoint preceding it need to be recomputed when computing that b node during backprop. When backprop has progressed far enough to reach the checkpoint node, all nodes that were recomputed from it can be erased from memory. The order of computation and memory usage then would appear as: Graph 3: Checkpointed Backpropagation Thus, the package implements checkpointed backprop, which is implemented by taking the graph for standard/ vanilla backprop (Graph 1) and automatically rewriting it using the Tensorflow graph editor. For graphs that contain articulation points or single node graph dividers, checkpoints using the sqrt(n) strategy, giving sqrt(n)memory usage for feed-forward networks are automatically selected. For other general graphs that only contain multi-node graph separators, our implementation of checkpointed backprop still works. But currently, the checkpoints have to be selected manually by the user. Summing up, the biggest advantage of using gradient checkpointing is that it can save a lot of memory for large neural network models. But, this package has some limitations too, which are listed below. Limitations of gradient checkpointing: The provided code does all graph manipulation in python before running your model. This slows down the process for large graphs. The current algorithm for automatically selecting checkpoints is purely heuristic and is expected to fail on some models outside of the class that are tested. In such cases manual mode checkpoint selection is preferable. To know more about gradient checkpointing in detail or to have a further explanation of  computation graphs, memory usage, and gradient computation strategies, Yaroslav Bulatov’s medium post on gradient-checkpointing.

OpenAI’s Gradient-checkpointing, Ethercraft, an Ethereum based RPG, pytorch implementation of faster R-CNN, and more in today’s top stories around artificial intelligence, blockchain, and data science news. 1. OpenAI releases gradient-checkpointing - a package for making huge neural nets fit in memory OpenAI has released gradient checkpointing - a package for fitting bigger Tensorflow models on GPUs. The package is developed jointly by Tim Salimans and Yaroslav and allows trading off some memory usage with computation to make a model fit into memory more easily. The package was able to fit more than 10x larger models onto GPU for feed-forward models, at only a 20% increase in computation time. The memory intensive part of training deep neural networks is computing the gradient of the loss by backpropagation. The package checkpoints nodes in the computation graph and recomputes the parts of the graph in between those nodes during backpropagation. Thus making it possible to calculate the gradient at reduced memory cost.  Using this way, researchers were able to reduce the memory consumption to O(sqrt(n)), at the cost of performing one additional forward pass when training deep feed-forward neural networks consisting of n layers. The complete working of the model can be found at the Github Repo. 2. A new project enables faster pytorch implementation of faster R-CNN There is a new project which allows complete pytorch implementation of faster R-CNN. This project aims to accelerate the training of faster R-CNN object detection models. Here are some highlights: It is pure Pytorch code. All numpy implementations are converted to pytorch Supports multi-image batch training. Supports multiple GPUs training. A multiple GPU wrapper (nn.DataParallel here) is used to make it flexible to use one or more GPUs, as a merit of the above two features. Supports three pooling methods, roi pooling, roi align and roi crop. It is memory efficient. It is faster. More information can be found at the Github repo. 3. Nordcloud teams up with Microsoft Azure to help customers to complete AI projects faster Nordcloud has partnered with Microsoft for jointly deploying Azure AI-based solutions for their enterprise customers. Nordcloud provides public cloud infrastructure solutions and cloud-native application services. It plans to help a much wider range of businesses implement artificial intelligence by leveraging Azure's AI infrastructure and services. Microsoft's Azure cloud computing service offers Platform Services for AI, machine learning, and IoT development. Nordcloud will work with Azure to allow its clients to spot opportunities to build cost-effective, scalable and intelligent AI-powered digital solutions. Customers can now use the entire Azure AI stack to create scalable intelligent services. They can also add smart Azure features to their existing solutions in a fast and agile manner. 4. Microsoft creates AI for state-of-the-art natural language processing A team at Microsoft Research Asia have reached the human parity milestone by creating an AI that can read a document and answer questions as well as a person. Microsoft has submitted a model based on the Stanford Question Answering Dataset (SQuAD). SQuAD is a machine reading comprehension dataset that is made up of questions about a set of Wikipedia articles. Microsoft's model was able to reach the score of 82.650 on the exact match portion. The human performance on the same set of questions and answers is 82.304. Microsoft is already applying earlier versions of the models that were submitted for the SQuAD dataset leaderboard in its Bing search engine, and more complex problems. 5. Introducing Ethercraft, a decentralized Role Play Game based on Ethereum blockchain There is a new Role Play Game (RPG) in town. Ethercraft is a new decentralized RPG game based on the Ethereum Blockchain. Although this game is still being developed, players can already start trading and purchasing items. The game and its development can be divided into three main components: Items: Buying, selling, and trading. Crafting: Combining items according to a recipe The playable video game Players will be able to buy, sell, and trade items beginning immediately. All of the items are independent tokens adhering to the ERC20 token standard. Players can also buy, sell, and trade Gold Pieces. Gold Pieces (XGP) is an ERC20 token, which can be used as an intermediary for everything involving ETH transfer. It also acts as a governance token. The Crafting and the playable video game is currently undergoing development and testing.

DeepBrain Chain project, AI bot that challenges humans, usql v0.6.0, and more in today’s top stories around machine learning, deep learning,and data science news. 1. DeepBrain Chain, the First AI Computing Platform Driven by Blockchain DeepBrain Chain, is the first AI computing platform driven by blockchain for global AI computing resource sharing and scheduling. As most organizations do not have capital to buy expensive GPU servers,  these companies provide a huge number of GPU servers which are unused or idle for a prolonged period.   The DeepBrain Chain provides a decentralized AI Computing platform, which is low cost, private, flexible, and safe. It serves the interests of several parties such as, the Miner’s main income is rewarded with token from mining, the AI companies just pay small amounts to run. Also, the Chain uses the smart contract in order to physically separate the data provider and data trainer. Thus, it protects the data of the provider. The interests of three major parties can be reconciled with the advanced technology. It is also automatically adjustable; if some nodes of DBC are attacked by hackers, the remaining nodes are working well as usual. DBC makes sure AI factories’ operations will never be interrupted. 2. Alibaba develops an AI bot to challenge humans in comprehension Alibaba, China’s biggest online e-commerce, has developed a deep neural network model, which has out-performed humans in a global reading comprehension test. According to a release, the model has scored higher on the Stanford Question Answering dataset (a large-scale reading comprehension test with more than 10,000 questions). Alibaba’s machine-learning models scored 82.44 on the test, compared with 82.304 by humans, on 11th January.  Si Luo, chief scientist of natural language processing at Alibaba’s research arm, said that, “We believe the underlying technology can be gradually applied to numerous applications such as customer service, museum tutorials, and online response to inquiries from patients, freeing up human efforts in an unprecedented way”. Similar to the model’s performance in the Stanford test, the machine learning model could identify the questions raised by consumers and look for the most relevant answers from prepared documents. Currently, the system only works best with questions that offer clear-cut answers. If the language or expressions are too vague, has grammatical errors, or there is no prepared answer, the bot may not work properly. 3. usql v0.6.0 released A universal command-line interface for SQL databases releases its version 0.6.0 with major updates below. Syntax highlighting Better compatibility with psql commands Homebrew support The release also includes some minor feature additions and a general code cleanup. Know more about this release on GitHub. 4. Cumulative Update #3 for SQL Server 2017 RTM Microsoft has released the 3rd cumulative update for SQL Server 2017 RTM. The major changes include: CPU timeout setting added to Resource Governor workgroup Support for MAXDOP option added for CREATE STATISTICS and UPDATE STATISTICS statements in SQL Server 2017 Improvement in tempdb spill diagnostics in DMV and Extended Events in SQL Server 2017 XML Showplans can now provide a list of statistics used during query optimization in SQL Server 2017 PolyBase technology enabled in SQL Server 2017 Execution statistics of a scalar-valued, user-defined function added to the Showplan XML file in SQL Server 2017 Optimizer row goal information added in query execution plans in SQL Server 2017 Other fixes and updates can be found here. The update can be downloaded from the Microsoft Download Center. Registration is no longer required to download the Cumulative updates. 5. Logz.io: AI-Powered ELK as a Service Logz.io have launched an AI powered ELK as a cloud service solution which offers a fully managed environment and unlimited data with automatic data parsing capabilities. The ELK stack (for Elasticsearch, Logstash, and Kibana — now called the Elastic Stack) is basically used to handle operational data, specifically log files. Although open source, this stack can be hard to implement and manage according to enterprise standard and is often expensive due to its labor-intensive nature. Logz.io’s ELK as a cloud service solution is an enhanced architecture which delivers the advanced log analytics, integration and security that enterprises require. The platform has an ‘intelligence layer’ which applies artificial intelligence to optimize data, establish correlations between new deployments and resulting log errors, and identify undetected patterns in the data, among other uses.

NumPy v1.14, Baidu’s Blockchain as a service, new insights on Intel’s Nervana NNP, and more in today’s top stories around machine learning, blockchain, and data science news. 1. The highly anticipated NumPy 1.14 released NumPy a popular library for Python, updates to version 1.14. The new update contains a large number of bug fixes and new features, along with several changes with potential compatibility issues. A major feature includes stylistic changes in the way numpy arrays and scalars are printed. Here are the top highlights: The np.einsum function uses BLAS (Basic Linear Algebra Subprograms) when possible. genfromtxt, loadtxt, fromregex and savetxt can now handle files with arbitrary python supported encoding. Major improvements to printing of NumPy arrays and scalars. There are also four new functions introduced. parametrize: decorator added to numpy.testing. chebinterpolate: Interpolate function at Chebyshev points. format_float_positional and format_float_scientific : format floating-point scalars unambiguously with control of rounding and padding. PyArray_ResolveWritebackIfCopy and PyArray_SetWritebackIfCopyBase: new C-API functions useful in achieving PyPy compatibility. The complete changelog can be viewed in the NumPy GitHub repo. 2. Baidu launches BaaS (Blockchain as a Service) open platform Baidu has launched a new Blockchain as a Service platform. The news came soon after Tencent, another Chinese tech giant announced its BaaS platform, a couple of months ago. According to the Baidu BaaS website, the blockchain open platform will customize and flexibly configure the attributes, modules, and mechanisms of the blockchain according to the actual business scenario of the enterprise. It supports real-time block writing and query with high concurrency and low latency, and supports multiple replications, multi-instance deployment, and ensures data consistency through consistency algorithm. It ensures safety with the use of asymmetric encryption, signature, certificate authentication, audit nuclear, access control, and other technical solutions, to guarantee data security. 3. A blockchain-powered open multimedia delivery platform by Snap Interactive Snap Interactive, a leading provider of live video social networking applications building on blockchain and other innovative technologies, today announced that it is developing an open source, multimedia delivery platform. This platform will combine STVI’s live streaming video, voice and data routing capabilities with the enhanced security, scalability and cost-effectiveness of blockchain technology. The platform is expected to: Specialize in routing live, rich media content and powering applications that require real time data and video communications Host Backchannel, STVI’s secure video messaging app, which is planned to launch in 2018. This app is anticipated to enable third-party developers for building next-gen social networking, messaging, group collaboration and live video streaming apps using the Company’s blockchain-based services. Provide the developer community with the opportunity to make use of company’s expertise in live video delivery and its recent investment in blockchain infrastructure to enable decentralized media routing. Offer the potential for superior security, scalability and cost efficiency, with token-based incentives, which will drive a large community of contributors of bandwidth and computing resources This decentralized platform is a horizontal extension of the best-of-breed NEM blockchain protocol. 4. The Nervana Intel chip moves from codename to actual product Intel recently announced few insights about its soon to be released Nervana Neural Network Processor (NNP). This AI-based chip was previously code-named as Lake Crest. It is built on a unique architecture, which is designed from the ground up in order to speed up the neural network training and AI modeling. Though complete details of the Intel NNP is yet unknown, we shed some light on the NNP chip architecture: It is a custom ASIC with a Tensor-based architecture placed on a multi-chip module (MCM) along with 32GB of HBM2 memory Each AI accelerator features 12 processing clusters which are paired with 12 proprietary inter-chip links, four HBM2 memory controllers, a management-controller CPU, as well as standard SPI, I2C, GPIO, PCI-E x16, and DMA I/O. Each processing element has more than 2MB of local memory, is software controlled and can communicate with each other using hi-speed bi-directional links. The Nervana NNP has 30MB in total of local memory. The processor is designed to be highly configurable, meeting both mode and data parallelism goals. To know about more features in detail, visit the official Intel NNP link here. 5. A new way to optimize MySQL database performance The IT operations and maintenance developers have found a new way to improve MySQL database performance. Optimizing MySQL database performance has always been a pain point for database administrators, especially in heavy demand public cloud services. According to a report by inside HPC, compiling the MySQL source code using the Intel C++ Compiler, the database performance is improved by 35% as compared with other compilers. This is made possible using Intel’s Interprocedural Optimization feature (IPO). With IPO, the compiler performs a multistep whole-program analysis across functions and procedures and multiple source files. This gives the compiler a complete view of the entire program so it can better implement inlining, constant propagation, dead call/function elimination, and other specialized optimizations. The Intel C++ Compiler can also generate machine code that is optimized to maximize performance. This includes specialized optimizations targeted at the new SIMD instructions and cache structure of the latest Intel CPUs.

Tableau version 10.5, AI-fused Titan supercomputer, IMVR’s Pro-Line and more in today’s top stories around machine learning, deep learning,and data science news. 1.Tableau upgrades to 10.5 with two new additions: Hyper and Linux Tableau 10.5 launched! By upgrading to Tableau 10.5 you automatically have access to the two new additions, Hyper and Linux. Hyper is Tableau’s patent pending data engine technology. Hyper will deliver up to 5 times improved query performance and about 3 times extract creation speed. Also, Tableau 10.5 delivers Tableau server on Linux, viz in tooltip, nested projects and more. Here’s what you can do with the new and upgraded Tableau 10.5: Perform high speed analytics using Hyper. Quick integration of Tableau server into your current Linux processes and workflows seamlessly. Create engaging data visualizations with the help of viz in tooltip feature. Organize your workbooks efficiently with the nested projects Work while you are away from your desk by using the annotate and share feature in the Tableau mobile iOS. Using this, one can quickly send notes or point out to interesting data points. Tableau 10.5 has also added more than 75 native connectors, which gives the ability to connect virtually to any web data through the Web Connector with a new connector to Box. It is also deeply integrated with the AWS platform with an update to the Amazon Redshift connector. With this, you can directly connect Tableau to data in Amazon Redshift and analyze it in conjunction with data in Amazon S3. To know more about this exciting upgrade visit the link here. 2. Apache® Trafodion™ announced as a Top-Level Project by Apache Software Foundation Apache software foundation today announced Apache Trafodion, as a Top-level project. Trafodion, is the first integrated Open Source solution that delivers on the promise of integrated transactional and analytical systems (OLTP/OLAP) for Apache Hadoop. Trafodion builds on the scalability, elasticity, and flexibility of Hadoop. Features of Trafodion include: A fully functional ANSI SQL support that leverages existing SQL skills Distributed ACID data protection, guaranteeing data consistency across multiple tables and rows Compile-Time and Run-Time Optimizers, delivering performance improvements for OLTP workloads Parallel-aware Query Optimizer, supporting large data sets Apache Spark integration, supporting streaming analysis Interoperability with existing Apache Hadoop tools and solutions, such as Hive, Ambari, Flume, Kafka, and Oozie, and Apache Hadoop and Linux distribution neutrality. “We are planning to use Trafodion to expand the business of China Mobile's Big Data platform: our data statistics of 4G real-time business in the country and provinces are more efficient than ever before", said Tianduo Gao, Senior Development Engineer of Software Technology (Suzhou) at China Mobile. To read about this in detail visit the link here. 3. AI-fused Titan supercomputer to accelerate design, training of deep learning networks Researchers at the Department of Energy's Oak Ridge National Laboratory have planned to integrate artificial intelligence and high-performance computing within their Titan supercomputer. By this AI fusion, the supercomputer will achieve a peak speed of 20 petaflops in the generation and training of deep learning networks. To efficiently design neural networks capable of tackling scientific datasets and expediting breakthroughs, the team developed two codes for evolving (MENNDL) and fine-tuning (RAvENNA) deep neural network architectures. These can, Generate and train as many as 18,600 neural networks simultaneously. They can also achieved a peak performance of 20 petaflops, or 20 thousand trillion calculations per second, on Titan (or just under half of Titan's single precision total peak performance). Practically, this means training 40-50,000 networks per hour. With Titan, the team is able to train an unparalleled number of highly accurate networks. It is a Cray hybrid system, meaning that it uses both traditional CPUs and graphics processing units (GPUs) to tackle complex calculations for big science problems efficiently; the GPUs also happen to be the processor of choice for training deep learning networks. Know more about this news in the link given here. 4. Alibaba's AI collaborates with MediaTek for IoT initiatives The e-commerce giant Alibaba, has partnered with MediaTek on an IoT initiative. Under this collaboration, both companies will work on developing smart home protocols, customized IoT chips and AI smart hardware. This partnership was officially announced at the Consumer Electronics Show (CES) 2018, which kicked off over the weekend in Las Vegas. They also announced a Smartmesh connectivity solution supporting many-to-many Bluetooth mesh technology. The solution, is based on a self-developed IoT protocol named IoTConnect and the Bluetooth chip co-developed by both parties. It enables smart home devices to automatically pair with Tmall Genie, their voice-controlled smart assistant. 5. IMVR showcases its “PRO Line” VR Devices with AI capabilities IMVR, announced their IMVR PRO line VR devices including the PRO-BlueSky and the PRO-DG1 at the ongoing CES 2018. These VR devices are equipped with an always-on connection through the AlterSpace immersive environment to AILEENN (The Artificial Intelligence Logical Electronic Emulation Neural Network Engine). The PRO line connects with the AlterSpace Virtual Environment Engine to accurately simulate any experience with its dual screens and military grade, ultra-precise head tracking. It also has a setting featuring of a 130 degree field of view, with adjustable focus and interpupillary distance. The PRO line connects to practically any HDMI output, enabling millions of computers and devices to render VR 360 interactive immersive content. The IMVR PRO line provides high quality VR experience focused on industrial, automotive, aerospace, military, theme park, and medical usage.

Python 3.7.0a4, New updates to Power BI Desktop, AI platform for cybersecurity, photo-centric cryptocurrency, and more in today’s top stories around artificial intelligence, blockchain, and data science news. 1. Python releases early developer preview of Python 3.7 Python org tease the developer preview of Python 3.7 with the release of version 3.7.0a4, the last of four planned alpha releases. Many new features for Python 3.7 are still being planned and written. The major new features and changes so far include: Coercing the legacy C locale to a UTF-8 based locale A New C-API for Thread-Local Storage in CPython UTF-8 mode Deterministic pyc Core support for typing module and generic types Module __getattr__ and __dir__ Built-in breakpoint() Data Classes Time functions with nanosecond resolution The next pre-release of Python 3.7 will be 3.7.0 beta 1, currently scheduled for 29th January. The official documentation of Python 3.7.0a4 can be found here. 2. New updates to Power BI Desktop Power BI, a powerful suite of business analytics tools, kicked off the new year with new updates to their Power BI Desktop. The updates revolve around 4 categories. Reporting which includes the ability to show and hide pages, automatic date hierarchy, data label and axis formatting, a relative date slicer, and more. Analytics, where users are provided with correlation coefficient quick measure Custom Visuals, which includes PowerApps, TreeViz, Funnel with Source by MAQ Software, Agility Planning Matrix Light, etc. Data connectivity, with support for Azure Active Directory authentication for Azure SQL Database & Data Warehouse connectors. The complete list of January updates can be found here. 3. Kodak launches KODAKOne image rights management platform and KODAKCoin, a photo-centric cryptocurrency Kodak, in a licensing partnership with WENN Digital, have launched KODAKOne, an image rights management platform and KODAKCoin, a photo-centric cryptocurrency. The KODAKOne image rights management platform provides an encrypted, digital ledger of rights ownership for photographers to register both new and archive work that they can then license within the platform. KodakOne will use blockchain technology to provide continual web crawling to monitor and protect the IP of the images registered in the KODAKOne system. They have also launched a new photo-centric cryptocurrency, KODAKCoin, which allows photographers to receive payment for licensing their work immediately upon sale and sell their work confidently on a secure blockchain platform. Shares in Kodak Co. have more than doubled after the launch the KODAKCoin. 4. CUJO AI announces its AI platform for the cybersecurity ecosystem CUJO AI has launched an artificial intelligence platform to help ISPs protect customers smart home environments by using threat intelligence, machine learning, and cloud computing to analyze an automated home’s network behavior. CUJO AI algorithms identify all user devices on the network, troubleshoot possible issues, block known threats and inform the user about the status via an app. The CUJI AI Platform has the ability to control devices, access, and security. According to CUJO AI, Using CUJO AI platform, ISPs can enable full internet security protection for home users without having to replace or introduce new hardware reducing costs thereby delivering value to their customers. More information is available on their official website. 5. Visteon launches DriveCore autonomous driving platform At the ongoing CES 2018, Visteon corporation unveiled DriveCore, their autonomous driving platform. DriveCore is a complete technology platform for developing machine learning algorithms for autonomous driving applications of Level 3 and above. It consists of the hardware, in-vehicle middleware, and a PC-based software toolset. Using a centralized computing approach, DriveCore will provide automakers a fail-safe domain controller, with a high degree of computing power scalability. In addition, it provides support to perform late sensor fusion of data from multiple cameras, Lidar and radar sensors. DriveCoreTM consists of three primary components: Compute is a modular, scalable computing hardware platform designed to deliver 500 gigaflops to 20 teraflops of processing power. Runtime is in-vehicle middleware that provides a secure framework to enable applications and algorithms to communicate in a real-time environment. Studio is a PC-based development environment that enables automakers to create and support an ecosystem of algorithm developers, through an open framework for sensor-based AI algorithm development.

Day 2 of the International Consumer Electronics Show (CES 2018) witnessed major contributions by Intel, Samsung, Qualcomm, Udacity, and LG in incorporating AI in their technologies and products. Here are the top highlights: Intel shows off its neuromorphic AI chip and a 49 qubit quantum chip Intel announced that its Loihi AI chip, launched on September 17, is now fully-functional and ready to be shared with research partners. Loihi is Intel’s first neuromorphic chip, designed to mimic the functioning of neurons and synapses in the brain. They are less flexible and more powerful than most general-purpose chips. The Loihi chips don’t require a huge amount of training data to learn a process and are energy efficient. Currently, its functionality is limited to simple object recognition. However further applications of these chips are likely to be in robotics and self-driving cars. Intel also unveiled its Tangle Lake chip—a superconducting quantum test chip of 49 qubits at CES 2018. The 49-qubit chip builds upon their earlier work with 17-qubit arrays. Intel has also developed packaging to prevent radio-frequency interference with the qubits. They use a flip chip technology that enables smaller and denser connections to get signals on and off the chips. This new announcement has put Intel in a good position with IBM and Google as far as the quantum computing race goes. Qualcomm plans to make a smart speaker development kit and extends its support to popular voice assistants Qualcomm announced their plans for the first half of 2018 at the ongoing CES event. First, they plan to make a smart speaker development kit based on the Qualcomm Smart Audio Platform. The development kit is engineered to help developers and audio manufacturers simplify the development of smart speaker products. The development kit will feature a Wi-Fi certified System-on-Module (SoM) that integrates the key system components. The kit also includes schematics and design files to support easier customization and differentiation in the manufacturers’ products. Additionally, the development kit offers a reference design for smart speaker devices. The company also announced that its Smart Audio Platform will now allow developers to choose which assistant they want to incorporate into their smart speakers. A choice of voice assistants from Amazon’s Alexa, Microsoft’s Cortana, and Google assistant will also allow other hardware manufacturers to more easily build devices that support these virtual assistants. LG plans to intelligently enhance TV images using computer vision LG announced its plans to apply AI to enhance TV images using state-of-the-art computer vision at the ongoing CES 2018. They will apply object-based enhancement to TV images. Applying object recognition AI will help in smoothing color banding and also help in identifying faces in a picture or distinguishing between, say a cat from a dog. Thus every image or scene will be parsed more intelligently. However, this announcement is at a very early stage, as more progress is still to happen. Apart from this, LG also talked about the potential of its new AI platform, ThinQ, for bringing deep learning and interoperability to the company's smart products. LG smart appliances will have the ability to learn habits over time and communicate with each other. Samsung plans to connect all its products with the IoT cloud by 2020 Samsung announced plans to connect 90 percent of their products with the IoT cloud at the CES 2018.  Additionally, the products will have artificial intelligence capabilities through Samsung’s virtual assistant Bixby. Samsung’s SmartThings Cloud service will also be available this spring. The SmartThings Cloud would allow people to control IoT devices from a single app, instead of having one for each gadget. Apart from working with Samsung products, it will also connect with cars running on Harman's Ignite cloud and other products that work with SmartThings. Samsung also unveiled its modular TV, called The Wall with customizable size configurations. It will also have built-in Bixby support for searching for TV shows, movies, weather reports, play songs, show photos from the cloud etc. Udacity and Baidu partner to come up with AI courses for building self-driving cars Baidu, one of the leading AI organizations has partnered with Udacity, an online education platform for building courses together. This collaboration was announced by Baidu’s COO Qi Liu and Udacity founder Sebastian Thrun at the ongoing CES 2018. According to Thrun, the AI expertise required to build self-driving cars is depleting and thus courses and programs like these are necessary to bring more talent in this area. Apart from these, Udacity will also make contributions in Baidu’s Apollo which is an open platform for autonomous driving. In regards to this Udacity will offer an Introduction to Apollo program and help Apollo with talent identification and acquisition. The program will be free of cost to the aspirants and will cover the entire Apollo software and simulation environment with hands-on learning opportunities. More advancements and announcements are bound to continue for the next 3 days as more organizations showcase their innovative products. Keep an eye on our website for further updates.

Ethereum’s defiance of the cryptocurrency downfall, CES 2018 highlights, new ML tools for an analytics software, a new Edge AI Platform & AI Technology for Cross Platform Consumer Devices, in today’s top stories around machine learning, blockchain, and data science news. 1. Ethereum maintains its position while other cryptocurrencies fall Cryptocurrencies Bitcoin and Ripple led the scoreboard of downfall early Monday morning. Ripple started off with  $2.46 which means 10 percent lower on the Bitstamp exchange. Bitcoin and Litecoin were down by 7.6 percent and 6.4 percent respectively, as per Coinbase. Additionally, Bitcoin futures also fell 10.5 percent at the Cboe, and traded at $15,010 as of 4. 15 p.m.(New York time). Popular speculations for the downfall are, Due to capital flows with investors realizing their profits from cryptocurrencies. A data adjustment by CoinMarketCap, the most popular site for cryptocurrency pricing data, removed South Korean exchanges from its site, which have been known to trade much higher than the rest of the world On the other hand, Ethereum still managed to escape the fall and traded at $1,143.56, i.e at a 4.7 percent rise. According to a report, the transaction volume on Ethereum's network doubled which included  periods where over 10 transactions per second were being processed. For comparison, Bitcoin's blockchain network can only handle about three transactions per second. 2. SkyFoundry Releases Machine Learning tools for its SkySpark® Analytics platform SkyFoundry, a leading developer of analytics software, releases Machine learning tools for their  SkySpark® Analytics software, which automatically analyzes building, energy and equipment data to identify issues, faults and opportunities for improved performance and operational savings. The newly released machine learning tools provide support for supervised learning in order to predict and forecast through regression-based approaches, and classification using Support Vector Machine (SVM) techniques. SkySpark, with a combination of full programmability and an extensive library of built-in functions allows customized analytics in order to meet the real-world needs of individual systems and facilities. 3. Rockchip announces 2.4 TOPS Embedded AI Chip Chinese fabless semiconductor company Rockchip launches an AI-focused chip, the RK3399Pro, which consists of an SoC(System on a Chip) and an NPU (Neural processing Unit) Features of the RK3399Pro SoC(System on a Chip): The chip uses a dual-core Cortex-A72 CPU and a quad-core Cortex-A53 CPU as well as a Mali-T860 GPU. It comes with a dual type-C interface, and supports a dual Image Signaling Processor (ISP), 4096x2160 display output, as well as an 8-channel digital microphone arrays input. Features of the RK3399Pro NPU (Neural processing Unit): The Neural Processing Unit (NPU), within the chip, promises an AI performance of up to 2.4 trillion operations per second (TOPS). The NPU’s performance falls in between Huawei’s Kirin 970 AI processor (1.9 TOPS) and Google’s Pixel Visual Core (3 TOPS) It is four times faster than Apple’s Neural  Engine (0.6 TOPS) The RK3399Pro NPU supports OpenVX, TensorFlow Lite, Android’s Neural Network API (NNAPI), as well as the more full-featured Caffe and TensorFlow machine learning framework frameworks. The chip can do both 8-bit and 16-bit computing. The company aims at targeting lower-cost devices with the RK3399Pro. Rockchip will enable developers with a reference design and SDK to get started on their RK3399Pro-based projects. 4. MediaTek shows off its Edge AI Platform & AI Technology For Cross Platform Consumer Devices At CES 2018, MediaTek, a global semiconductor leader, today unveiled its ongoing AI platform strategy in order to enable AI edge computing with its NeuroPilot AI platform. MediaTek will bring AI across its wide-ranging technology portfolio that powers 1.5 billion consumer products a year across smartphones, smart homes, autos and more. This will be achieved by combining hardware, software, an AI processing unit (APU), and NeuroPilot SDK. MediaTek's current AI solutions for voice assistants, Televisions and autonomous cars, will also be showcased at CES where the company is demonstrating the power of AI and how it's redefining today's consumer devices. This year, MediaTek plans to enable their partners and customers with technology advancements consumers demand through the power of AI integration with their chipsets. The MediaTek's NeuroPilot AI platform has some key features: Edge AI Enabler - Creates a strong hybrid of an edge-to-cloud AI computing solution. Edge AI Efficiency - A balance of performance and power efficiency which makes implementing and running AI applications efficient and practical across devices. Enhanced AI - Improvises features and applications that people use every day such as mobile devices, intelligent camera imaging and voice and image detection or recognition. Supports Mainstream AI Frameworks - Operates in concert with existing neural processing SDKs including Google TensorFlow, Caffe, Amazon MXNet, Sony NNabla and more. Software & Hardware Solution - Allows developers access to SOC level functions to build AI applications and solutions across MediaTek chipsets and MediaTek powered devices. "With our broad range of chipsets, backed by our current and newly developed AI framework, MediaTek is poised to be a full ecosystem AI solution provider", said Jerry Yu, Corporate Vice President and General Manager of the Home Entertainment Business Group, MediaTek. 5. Tensorlang: A differentiable programming language which is based on TensorFlow Tensorlang is a programming language for large-scale computational networks (e.g. deep neural nets) that is faster, more powerful, and enjoyable to use. Tensorlang is designed to address a number of requirements such as: Ability to saturate a single machine's local CPU and GPU with linear scaling Seamless scaling to clusters of machines Ability to compile programs to native code that runs fast on major operating systems and mobile devices Native support for symbolic differentiation Easy debugging and actual stacktraces for graph errors Execution model that matches other programming environments (e.g. no delayed execution) A productive REPL environment Compatibility with existing libraries and models To know more about Tensorlang in detail, visit the github link here. 6. Intel,Qualcomm, Samsung, LG, and many other tech players’ announcements at the CES 2018 The International Consumer Electronics Show (CES) 2018 kicked off last weekend. Here are some highlights so far: Intel shows off it’s neuromorphic AI chip and a 49 qubit quantum chip Qualcomm plans to make a smart speaker development kit and extends its support to popular voice assistants Samsung announced that it would connect all its products with AI by 2020. It has also unveiled a high-tech television, The wall. LG plans to intelligently enhance TV images using computer vision. Udacity and Baidu partner to come up with AI courses for building self-driving cars For a detailed catch up on each of these news, you can read the detailed news article in the link above.  

The world’s biggest consumer technology show is here! Presenting CES 2018 that commenced last weekend. This new year, multiple tech firms such as LG, Sony, Samsung have launched brand new OLED screen televisions, smart laptops, speakers, and so on with next-gen technologies for their consumers. To know about these in detail, you can visit the link here. In this article, we explore how tech giants such as Nvidia, Intel, and AMD have leveraged AI and ML to launch next-gen products. Let’s take a brief look at each one’s contribution at the CES 2018. Nvidia Highlights at CES 2018 Nvidia unveiled their Xavier SoC(System on a Chip) autonomous machine intelligence processors at CES this year. The Xavier has over 9 billion transistors with a custom 8-core CPU, a 512-core Volta GPU, an 8K HDR video processor, a deep-learning accelerator and new computer-vision accelerators. With all these huge figures, Xavier can crunch more sensor and vehicle data for the AI systems that will power self-driving vehicles. The other striking features of this SoC are, it can perform 30 trillion operations per second using only 30 watts of power, and is 15 times more efficient than the previous architecture. Nvidia also announced three new variants of its DRIVE AI platform. These new variants are based around Xavier SoCs. The three variants include: Drive AR focuses on getting Augmented Reality into vehicles, which can enhance and transform the driving experience. It offers developers with an SDK, which will further enable them to build experiences that leverage computer vision, graphics and artificial intelligence capabilities to do things like overlay information about road conditions, points of interest and other real-world locations using interactive in-car displays. Drive IX would formulate an easy way to build and deploy in-car AI assistants. These assistants will be capable of incorporating both interior and exterior sensor data to interact not only with drivers but also with passengers on the road. The third DRIVE AI-based platform is a revision of its existing autonomous taxi brain, Pegasus. This new version improves on the previously revealed preproduction edition by compiling two Xavier SoCs with two Nvidia GPUs into a package that’s roughly the size of a license plate – down from the trunk-filling physical footprint of the original. Nvidia also announced that it is partnering with two Chinese companies Baidu and automaker ZF, for bringing autonomous driving to roads. Nvidia’s CEO Jensen Huang stated that Nvidia’s Drive Xavier auto compute platform would be used for Baidu’s Apollo Project. The Apollo project offers an open platform for self-driving cars in partnership with a wide variety of automakers, suppliers and tech companies. Huang also revealed that Nvidia will be supplying its self-driving computer hardware to Aurora, a Google start-up. Aurora would build self-driving systems for both Volkswagen and Hyundai, the startup revealed last week. Also, Uber has chosen Nvidia as one of its key technology partners in its fleet of self-driving, specifically to provide the AI computing aspects of its autonomous software. Uber has used Nvidia’s GPUs in both its self-driving ride-hailing test fleet and in its self-driving transport trucks, which are also developed by its Advanced Technologies Group. Intel Highlights at CES 2018 Intel in collaboration with AMD has unveiled new processors with the help of AMD’s Radeon RX Vega M graphics. These new core processors are Intel’s first CPU with discrete graphics included in a single package. This leads to an incredibly thin and lightweight laptops and desktops that are able to provide an impressive gaming performance with an added 4K media streaming. As per Intel, these chips would be the first example of power-sharing across CPU and GPU, the first consumer mobile chips to use HBM2 (the second-generation high bandwidth memory, a faster type of graphics memory), and also the first consumer solution to use Intel EMIB(Embedded Multi-die Interconnect Bridge). To know more about this in detail please visit the link given here. At CES 2018, Intel unveiled its new mini-PC NUC system, formerly codenamed Hades Canyon. This system aims at premium virtual reality (VR) applications. The system comes in two versions, the NUC8i7HVK and the NUC8i7HNK. The NUC8i7HVK: comes with Radeon RX Vega M GH graphics can operate from 1,063MHz to 1,190MHz It has an 8th-gen quad-core 100W Intel Core i7-8809G 3.1GHz with 4.2GHz turbo mode, and is "unlocked and VR-capable". The NUC8i7HNK: comes with Radeon RX Vega M GL graphics with an operating range of 931MHz-1,011MHz. It also has a 65W quad-core 8th-gen Intel Core i7-8705G 3.1GHz CPU with 4.1GHz turbo mode. To know more about this news in detail, visit the link here. AMD Highlights at CES 2018 AMD announced its brand new Ryzen 3 2300U APU chips specifically designed for affordable laptops and Chromebooks. The Ryzen 3 2300U is a full-featured chip featuring 4 cores and 4 threads clocked at a base 2.0GHz and boost 3.4GHz. Its APU comes with full-on Radeon RX Vega graphics powered by six compute units. In addition to the dual-core, Ryzen 3 2200U runs with 4 threads at a standard 2.5GHz frequency that boosts up to 3.4GHz. It also features Radeon RX Vega graphics similar to other APUs in the family but requires only three compute units to power it. AMD announced a new set of Ryzen chips for desktops i.e desktop Ryzen APUs in order to replace its ongoing Athlon chips. AMD’s new APUs are based on the Raven Ridge Architecture, and is a combination of an updated version of Ryzen processor with “discrete-class” Radeon RX Vega graphics. AMD has introduced two chips: Ryzen 5 2400G APU includes 4 cores and 8 threads clocked at a base 3.6Ghz and is boosted with 3.9GHz. On top of the processor, this new chip features Radeon RX Graphics with 11 compute units for playable gaming experiences at 1080p and high-quality settings.  Ryzen 3 2200G is rated for 3.5GHz base and 3.7GHz boost clock speeds. This entry-level APU also comes outfitted with 4 cores, but only 4 threads, as well as just 8, compute units attached to its Radeon RX Vega GPU. AMD also spoke about its new Ryzen 2 which would hit the market around April this year, which would have: A new 12nm Zen architecture, which out-smalls the 14nm transistors of Intel Coffee Lake. This upcoming chip brings higher clock speeds and Precision Boost 2 technology for greater performance and efficiency. To know more about this news in detail, click on the link here. Apart from well-known names such as Nvidia, Intel, and AMD, Ceva, the leading licensor of Signal processing platforms and AI processors, unveiled NeuPro. NeuPro is a powerful and specialized Artificial Intelligence (AI) processor family for deep learning inference at the edge. It is designed for edge device vendors who can quickly take advantage of the significant possibilities that deep neural network technologies offer. NeuPro extends the use of AI beyond machine vision to new edge-based applications including natural language processing, real-time translation, authentication, workflow management, and many other learning-based applications. With 4 more days to go, many such advancements are expected to be announced at the CES 2018. Watch this space in the coming days for more.

Nvidia’s new AI platforms, AI processors by Ceva, a new platform for many agent reinforcement learning, and more in today’s top stories around artificial intelligence, blockchain, and data science news. 1. Ethereum Foundation is looking for outside developers to help them solve Blockchain’s scaling problem Ethereum creators are exploring newer ways to fix the inability of blockchains to effectively scale. They are inviting outside developers to help solve the scaling problem. Until now, Ethereum has explored two possible fixes for the problem. The first solution is sharding which would require a small percentage of nodes to see and process every transaction, allowing many more transactions to be processed in parallel at the same time. The second solution involves creating data-link layers or layer 2 protocols that send most transactions off-chain and only interact with the underlying blockchain in order to enter and exit from the layer-2 system or in case of attacks on the system. A specification for an initial prototype is close to finalized and the next step involves building a reference implementation in python on top of Py-EVM, and a testnet in python. Outside developers are now invited to get involved in this sharding testnet and then the sharding mainnet steps. Ethereum is offering subsidies ranging from $50,000 to $1 million to programmers who can help find the fixes. Interested developers can send their proposals to apply@ethereumresearch.org. For more information visit here. 2. Nvidia announces three new variants of the DRIVE AI platform based around Xavier SoCs At the ongoing CES 2018, Nvidia has announced three new variants of its DRIVE AI platform, which are based around Xavier SoCs. The Xavier autonomous machine intelligence processors are now shipping out to customers, after being unveiled last year. Most of Nvidia’s initiatives, this year,  revolve around self-driving cars and its platform for allowing car manufacturers to build their own. DRIVE AR, the first of the DRIVE AI offerings, aims at enhancing and transforming the driving experience by adding augmented reality into vehicles leveraging computer vision, graphics, and artificial intelligence capabilities. DRIVE IX, the second platform, helps developers build and deploy in-car AI assistants. These AI assistants will interact with drivers as well as passengers on the road by incorporating both interior and exterior sensor data. Apart, from these, Nvidia has launched a revision of Pegasus, it’s autonomous taxi brain. According to them, "it delivers the performance of a trunk full of PCs in an auto-grade form factor the size of a license plate" Nvidia is currently working with at least 25 customers using Pegasus to power their self-driving robotaxi fleet. 3. Volkswagen joins forces with Nvidia to use AI in its new electric microbus Volkswagen joins forces with Nvidia to use it’s Drive IX platform in some of its upcoming vehicles, including the I.D. Buzz electric bus. Drive IX, announced at CES 2018, is a software developer kit that Nvidia created to tap into the power of Xavier. Volkswagen will use it to build features like facial recognition, gesture control, natural language processing, and more in their microbus. Volkswagen will initially focus on building Intelligent Co-Pilot features and using sensor data to make driving easier, safer and more convenient for drivers. Volkswagen will also work with Drive AR, a new augmented reality-based SDK from Nvidia to incorporate augmented reality into vehicles. The partnership between the two companies is also likely to be extended to future vehicles. 4. MAgent: A new platform for Many-Agent Reinforcement Learning MAgent is a new platform to support research and development of many agent reinforcement learning. Instead of using single or multi-agent reinforcement learning, MAgent focuses on supporting the tasks and the applications that require hundreds to millions of agents. Within the interactions among a population of agents, it enables not only the study of learning algorithms for agents' optimal policies but more importantly, the observation and understanding of individual agent's behaviors and social phenomena emerging from the AI society, including communication languages, leadership, altruism. MAgent is highly scalable and can host up to one million agents on a single GPU server. MAgent also provides flexible configurations for AI researchers to design their customized environments and agents. You can read the AAAI 2018 demo paper here. You can also watch the demo video for some interesting showcases here. 5. CEVA brings deep learning at the edge with NeuPro, a family of AI Processors Ceva Inc has unveiled NeuPro, an AI processor family for deep learning inference at the edge, at the CES 2018. It is designed for edge device vendors looking to quickly take advantage of the significant possibilities that deep neural network technologies offer. The AI processors offer performance ranging from 2 Tera Ops Per Second (TOPS) for the entry-level processor and 12.5 TOPS for the most advanced configuration. The NeuPro processor line extends the use of AI to new edge-based applications such as natural language processing, real-time translation, authentication, workflow management, etc. The NeuPro family comprises four AI processors offering different levels of parallel processing: NP500 the smallest processor, targeting IoT, wearables, and cameras. NP1000 targeting mid-range smartphones, ADAS, industrial applications and AR/VR headsets. NP2000 for high-end smartphones, surveillance, robots, and drones. NP4000 for high-performance edge processing in enterprise surveillance and autonomous driving.