Metaform and XP&DLand announce MetaPujo to bring Durga Puja to Metaverse

September 26, 2022

Metaform and XP&DLand have announced a MetaPujo today, which will foray to bring Durga puja festivities of Kolkata to the metaverse. Deshapriya Park, Ballygunge Cultural, Ahiritola Sarbojanin, and Tala Prattay pandals will be accessible via three-dimensional (3D) twins on the metaverse.

MetaPujo will introduce four NFTs of Durga idols from the most visited Durga pandals of Kolkata. The platform has planned centers around 3D recreations of the pandals, where users in a metaverse platform, ‘Spatial,’ can enter a shared social space where people from worldwide can come together, interact, and even take photographs.

Users have to create a meta-realistic avatar of themselves, with the platform accessible to all through simple smartphones, tablets, and wearables. As stated by Sukrit Singh, co-founder of Metaform and XP&DLand, the platform aims to bridge the gap between devotion and technology.

“This initiative aims to give Durga an address on Web3.0, where the devout from across the world can enter a three-dimensional recreation of the pandals and participate in darshans from the popular pandals of Kolkata. While doing so, attendees could claim four NFTs of her idols that have been created especially for the current occassion, each correlating to one of the pandals,” he added.

“We aim to democratize the metaverse. One doesn’t have to be in Kolkata to celebrate the pujas now. Meta pujos will allow people from across the world to enter pandals as family and friends, separated by physical distance but united in celebration.”

“In addition to presenting these experiences, introducing the concept of digital ownership, we intend to drop digital collectibles or NFTs, which can be purchased. These digital tokens will generate a value over time given their real-life utility. In the future, the plan is to monetize the collectibles and drive community with utility (whether for community, experiences, or collectibles) at a later stage with the intent to build a business,” Suveer Bajaj, co-founder of Metaform and XP&Dland, stated.

Meta and Google reduce staffing to cut costs amid economic downturn

Sahil Pawar

September 26, 2022

Meta and Google reduce staffing to cut costs amid economic downturn

Meta and Google are reducing staff to cut costs amid the economic downturn. The companies have put some employees on traditional 30 to 60 days lists to find a new role within the company or leave.

Wall Street Journal reported that Meta is planning to cut costs by at least 10% in the coming months and has put more and more workers whose jobs are being eliminated on its traditional 30-day list.

Moreover, Google’s parent Alphabet has reportedly employed a similar approach, typically allowing workers 60 days to apply for a new role if fired.

Meta has a long history of eliminating employees whose roles are subject to termination if they cannot find a new job internally at the company within a month. Meta had a total of 83,553 employees at the end of this year’s second quarter.

Last month, Google fired about 50 workers at its incubator Area 120 and gave them more 30 days to find another job. A Google spokesman said that nearly 95% of employees found new roles within the notice period.

Alphabet CEO Sundar Pichai aims to make the company 20% more efficient, hinting at job cuts. The tech giant recently canceled the projects at its in-house research and development (R&D) division called Area 120.

Tools to translate coding languages

Gareema Lakra

September 25, 2022

Today, people use different coding languages for better results or just the zeal to learn something new. The history of coding languages is quite interesting and dates back to the early 1800s. Many modern coding languages have roots in Ada Lovelace’s first machine algorithm developed in 1843. Until now, nearly 9000 coding languages have been created, among which only the most popular ones that count to 150 are vastly used. Coders prefer some coding languages for a specific field like Python for developing websites and software, R programming for statistics, etc. With the development in technology, the coding languages are evolving too, which enhances the way of working, outputs, and the user experience. The translation of a code script or to translate coding languages is required when companies switch from one language to another, as Commonwealth bank of Australia did from COBOL to Java which cost them $750 million in five years. With AI’s application, transpilers are developed as coding languages translators to save money and time.

What is a coding languages translator?

A coding language translator is a tool that translates a program code from the source code into the destination code. Translating coding languages is like the same process of translating languages from one to another but complicated. In the process, we alter the data but want to preserve the data structure as far as possible. Theoretically, a piece of code is translatable to any other language if the programming languages are Turing complete. Turing completeness is a property in computability theory, where you can perform any computation on anything that any other computational method is capable of. We use compilers to perform translations on coding languages, but compilers are frequently used to change code so that a machine can interpret it. It requires a subset of compilers called transpilers for the code to be human-readable. Transpiler (aka source-to-source compiler) is a translator capable of translating between programming languages that operate on the same level of abstraction. Unfortunately, creating a transpiler in practice is challenging since different languages have different syntaxes and rely on various platform APIs and standard library functions.

Why it is difficult to translate coding languages?

Translation of coding languages is a tedious task that includes preserving the exact execution semantics and general code similarities at the same time. Besides, both the software you are working on and the languages to be translated makes the process challenging. When an application has many external dependencies that are difficult to replace in the target technology stack, the challenges are different than for an app that implements everything internally. You see, there can be several subtitles in either the source or the destination code, which need to be satisfying or may not matter in translation. Look at the ‘+’ operation in Python, C, and C#; the operand is the same operation yet performs differently. That is, mathematically, the addition is correct in Python but can expand more than 32 bits instead, it stays to 32bits only in C, and in C#, it may throw an exception depending on the mode. Another reason is the calling of library functions. When you translate programming languages with different semantics, for example, Python to C#, the Python functions in your code will not work in C# because they aren’t available in C#. Over time, many researchers came up with the idea of translating coding languages. Here is a list of tools for translating coding languages.

Read more: Top AI chatbot companies in India

List of coding languages translators

Facebook’s TransCoder AI

TransCoder AI is a programming languages translator system developed by Facebook researchers, which can translate code among C++, Java, and Python languages. It is based on an unsupervised machine learning algorithm to perform translation and is one of the best Python transpilers. It was trained on over 2.8 million open-source projects and outperformed existing code translation systems using rule-based translation methods. TransCoder was first proposed in the paper, ‘Unsupervised Translation of Programming Languages, ‘ published on arXiv. The paper talks about transcompliers and how TransCoder works. The algorithm used in TransCoder is inspired by the neural machine translation (NMT) system to programming languages. The algorithm in TransCoder identifies common elements between input and output languages called tokens. Tokens can be defined as the keywords in programming languages like ‘for,’ ‘if,’ ‘else,’ ‘try,’ etc., and the mathematical digits and operators. Some tokens are the common strings that are present in the code. Then, the translation quality is improved by the algorithm using the back-translation method. The back-translation method induces to build source to destination code and destination to source code simultaneously and are coupled together at the end to give the final output. You can use the TransCoder by following the steps in the repository.

The testing for accuracy was done using 852 parallel functions in C++, Java, and Python from GeeksforGeeks. The computational accuracy was calculated while translating between C++, Java, and Python based on the BEAM seach of N=25, which is listed here.

Computational accuracy of translation between:

C++ to Java – 74.8%
C++ to Python – 67.2%
Java to C++ – 91.6%
Java to Python – 68.7%
Python to Java – 56.1%
Python to C++ – 57.8%

TransCoder works on a transformer-based sequence-to-sequence architecture consisting of an attention-based encoder and decoder. It follows three principles, one cross-lingual masked language model pretraining, denoising auto-encoding, and back-translation. Below is an illustration presented in the paper mentioned above.

_{Source: arxiv.org}

IBM’s CodeNet

CodeNet is a work-in-progress project of IBM that has aimed to teach AI to code. At IBM’s Think 2021 conference, Arvind Krishna, CEO of IBM, revealed the project CodeNet, which is a two-year effort of Dr. Ruchir Puri and an IBM research team. The project CodeNet is a large-scale dataset with approximately 14 million code samples written in over 50 programming languages. The wide variety of languages and coding problems CodeNet has contains over a hundred solutions each for 80% of the problems. The idea behind project CodeNet is to enable researchers and developers to help in code search, code completion, code-to-code translation, and a combination of other use cases. A detailed description of project CodeNet is done in this paper, ‘CodeNet: A large-scale AI for code dataset for learning a diversity of coding tasks.‘ CodeNet is similar to ImageNet, a computer vision dataset having 14 million images scattered across 20,000 categories. Though the operation of CodeNet to translate code-to-code is not ready yet, it is a promising project and expects to perform well.

Google’s GWT

GWT, Google web toolkit, is a development toolkit for building and optimizing complex browser-based applications. It is an open-source set of tools allowing developers to create and maintain JavaScrip front-end applications. GWT’s objective is to make it possible to construct high-performance web apps productively without the developer having to be an expert in JavaScript, XMLHttpRequest, or browser quirks. GWT is a popular coding languages translators that work to create and debug AJAX apps in Java, and when pushed to production, the compiler converts your Java application to browser-friendly JavaScript and HTML.

GWT has two major components:

GWT SDK: It contains the Java API libraries, compiler, and development server. It allows the user to write client-side applications in Java and deploy them as JavaScript.
Plugin for Eclipse: It provides IDE support for GWT and app engine web projects.

You can start using GWT. The first step is to download the SDK and take in-depth tutorials to understand the fundamentals of GWT development.

GitHub Copilot

GitHub Copilot is an AI pair programmer powered by OpenAI Codex, a new AI system created by OpenAI. It helps developers to write code faster with fewer efforts by offering autocomplete-style suggestions as you code. The suggestions are given either while writing the code directly or by writing a natural language comment about what you want the code to do. It is similar to giving commands to write code. GitHub Copilot provides various use cases with coding, including writing code with a simple command in more than one language, creating dictionaries of lookup data, navigating a new codebase with GitHub Copilot Labs, and more.

The GitHub Copilot Labs is a companion extension to GitHub Copilot, which can be installed from the visual studio marketplace. It is a visual studio code extension that contains experimental applications of Copilot, which are ML-powered editor features for developers. One of the features of Copilot Labs is translating coding languages from one to another, which comprises around 60 coding languages to choose from.

The process of coding translation consists of only three steps:

Install the GitHub Copilot Labs extension.
After installing Copilot Labs, open the folder you want to translate, then click on the extension icon.
Now, you can see the window ‘Translate code into:’, highlight the code, select your desired language and click on ‘Ask Copilot’.

Following the steps mentioned above, your code will be successfully translated.

The task of code language translation is challenging, and many will find it amusing to become a reality. We have just a handful of coding languages translators, but we are hopeful and believe the best is yet to come. For now, Facebook’s TransCoder shows remarkable results, successfully understanding the syntax specific to each language and learns data structures and their methods. At higher levels, maintaining a transpiler is more difficult than developing one as new technologies emerge and new features are introduced when programming languages advance. As a result, the work on transpilers is a consistent process and will need more hands on the deck.

Researchers use Deep Learning to Hallucinate synthesis of new proteins

Preetipadma K

September 24, 2022

synthesis of new proteins using hallucination algorithm

In a new study, researchers from the University of Washington School of Medicine assert that deep learning can be used for protein molecule synthesis much more accurately and quickly than previously possible. The scientists hope this revelation will lead to many new vaccines, treatments, tools for carbon capture, and sustainable biomaterials.

Proteins are composed of lengthy chains of amino acids connected by peptide bonds. They are essential for the chemistry of the body as well as the structure of cells and communication between them. A protein’s function is determined by its shape. When protein creation goes wrong, the resulting malformed proteins cause failure to perform their crucial tasks or neurological diseases, like Alzheimer’s, Parkinson’s, Huntington’s, and Lou Gehrig’s (ALS) disease. A protein must “fold” in order to perform its function in the cell. Protein folding is the procedure by which a molecule is changed into a complicated 3D structure that can interact with its target in the cell.

The protein won’t form the correct alignment nor carry out its function inside the body if the folding is interrupted. To better understand how cells work and how misfolded proteins contribute to disease, researchers have been focusing on understanding protein folding. Better protein prediction methods will also aid in developing drugs that can target a specific topological region of a protein where chemical reactions occur.

According to the study led by biochemist David Baker at the University of Washington in Seattle, the shapes that can be found in nature are only a tiny portion of what is thought to be conceivable.

In December 2020, DeepMind’s protein structure prediction tool AlphaFold made headlines when it won the Critical Assessment of Protein Structure Prediction, or CASP, competition. The competition, which is held every two years, assesses advancement in one of biology’s most challenging problems: figuring out proteins’ three-dimensional (3D) structures strictly from their amino-acid sequence. Entries made using computer software are compared to protein structures identified using experimental methods like X-ray crystallography or cryo-electron microscopy (cryo-EM), which shoot X-ray or electron beams at proteins to produce an image of their shape.

In the 1990s, Baker’s team began creating a software called Rosetta that helps with protein folding. The software first determined an amino acid sequence corresponding to the structure researchers had originally envisioned for a novel protein, typically by fusing pieces of other proteins together.

But when created in the lab, these “first draft” proteins rarely folded into the required form and were instead locked in various ways. Therefore, another step was required to modify the protein sequence such that it would only fold into the particular desirable structure. This stage was computationally intensive because it involves modeling every possible folding scenario for various sequences.

That time-consuming procedure has been made instantaneous by employing AlphaFold. In a method known as “hallucination,” which Baker’s team created, scientists input random amino-acid sequences into a structure-prediction network; this changes the structure so that it becomes ever more protein-like, as evaluated by the network’s predictions. In simple words, this method consists of taking pieces of an existing structure and asking the AI to fill in the gaps. In a 2021 study, Baker’s group reported finding evidence that around one-fifth of the tiny, “hallucinated” proteins they produced in the lab resembled the predicted form.

In the past year, a team under the direction of Minkyung Baek, a postdoctoral scholar in the Baker lab, created software that employs deep learning to swiftly and accurately predict protein structures from sparse data. Dubbed as RoseTTAFold, the team developed this AI software as it wasn’t clear when DeepMind would make the AlphaFold software or its forecasts publicly available.

Researchers describe RoseTTAFold as a “three-track” neural network, which means it constantly considers potential three-dimensional structures of proteins, patterns in protein sequences, and how amino acids interact with one another. This architecture enables the network to collectively reason about the link between a protein’s chemical components and its folded structure by exchanging one-, two-, and three-dimensional information.

Rosetta is a revolutionary toolset for protein structure prediction, however, its success rates are really relatively low, i.e., just a small percentage of its designs successfully fold and function as intended. Deep learning models like AlphaFold and RoseTTAFold are stepping up the game!

Baker’s team divided the problem of protein design into three pieces and employed novel software solutions for each to create proteins that go beyond the proteins found in nature.

To begin, a new protein form must be created. The scientists demonstrated how artificial intelligence could create novel protein forms in two methods in an article released on July 21 in the journal Science. The first was “hallucination,” and the second was “inpainting,” which is similar to the autocomplete function seen in current search bars.

Baker and his colleagues believed they could create self-assembling proteins using hallucination that would form variously sized and shaped nanoparticles. However, none of the 150 designs worked when scientists taught microbes to create them in the laboratory.

To speed up the process, a deep-learning tool was developed simultaneously by Justas Dauparas, a machine-learning expert. This main objective was to handle the so-called inverse folding issue, which is identifying the protein sequence that matches a given protein’s overall structure. According to the September 15 edition of Science, the ProteinMPNN approach for designing protein sequences is based on deep learning and has exceptional performance in both silico and experimental tests. By changing sequences while keeping the overall shape of the molecules, it can serve as a “spellcheck” for designer proteins developed with AlphaFold and other tools.

The researchers evaluated and improved the predicted sequences using protein structure prediction algorithms and laboratory protein synthesis. Next, the scientists used X-ray crystallography to confirm the protein structures and cryo-electron microscopy to determine the proteins’ shapes. The researchers identified the structures of 30 of their novel proteins, and 27 of them matched the AI-led designs.

ProteinMPNN has a sequence recovery rate of 52.4% on native protein backbones, compared to 32.9% for Rosetta. In creating the molecules experimentally, Baker and his team had significantly more success when they used this second network on their hallucinated protein nanoparticles.

Finally, using AlphaFold, the team independently determines if the proposed amino acid sequences will likely fold into the desired shapes. According to Baker, ProteinMPNN is to protein design what AlphaFold was to the prediction of protein structure. The team explains that while protein structure prediction software is a part of the solution, it cannot provide any original ideas by itself. In order to discover any new functional proteins, you would need to comb through billions of sequences, even if you had the ideal technology for predicting how protein sequences fold.

A group from the Baker lab demonstrated in a different study published on September 15 in Science that combining ProteinMPNN and the other new machine learning methods could consistently produce proteins that worked in the lab. According to the author, you need to understand these molecules in the real world rather than just relying on the computer to build proteins correctly. According to project scientist Basile Wicky, the team discovered that proteins generated using ProteinMPNN were significantly more likely to fold up as planned, and we could construct highly complicated protein assemblies using these approaches.

According to Baker and his colleagues, synthesizing a brand-new protein in the laboratory is the true test of deep learning-based protein structure prediction approaches. This is evident from their early failure to create hallucinated protein assemblies. The scientists created a variety of novel proteins, including nanoscale rings that they anticipate could be repurposed as components for innovative nanomachines. The rings, which had widths about a billion times smaller than a poppy seed, were examined using electron microscopes.

Top Artificial Intelligence Books

Disha Chopra

September 24, 2022

Artificial intelligence is an incredible technology that accomplishes several human tasks much faster and more efficiently. With more processing power, AI offers several benefits that make it an excellent tool for modern organizations. The innovations of AI-based technologies are applied universally to deliver healthcare-related services, manufacturing processes, banking, and financial services, life sciences, and more. Given the efficiency of artificial technology, many newcomers and existing companies are shifting their focus to incorporating AI in their operations. To keep up with the trends, professionals seek to know more about the technologies by referring to several artificial intelligence books. This blog will provide a list of some worthy AI books. But, let’s first learn what artificial intelligence is.

What is AI?

Artificial intelligence is the simulation of human intelligence with computers. It is a technology that enables machines, especially computers, to perform human tasks. AI requires a strong foundation of structured hardware and software for building and training algorithms using R, Python, and more. AI systems ingest vast amounts of data for training, analyzing this data to find patterns and, ultimately, using the patterns to predict results and future states. Learning more about artificial intelligence can be challenging if you do not have the correct reference or guidance. Several books talk about AI, but which one is right for you? Have a look at this list of best AI books.

List of top Artificial Intelligence Books

We have divided the list into two parts: one enlists the best artificial intelligence books for beginners, and the other is for people with some experience and ideas about AI.

Best artificial intelligence books for beginners

Architects of Intelligence: The truth about AI from the people building it

Written by Martin Ford, this AI book collects in-depth, one-on-one interviews with some bright minds in the AI community. Ford is a New York Times bestselling author and futurist who speaks at conferences and interviews experts to know more about the future of AI and automation. In this book, he talks about his conversations with 23 researchers and entrepreneurs working on robots and artificial intelligence. Yann LeCun (Facebook), Andrew Ng (AI Fund), and Stuart Russell (UC Berkeley) are a few among 20 others who spoke the ‘truth about AI’ in this AI book. It is a digression from usual AI books about AI projects and technical details.

Link for the book – Architects of Intelligence: The truth about AI from the people building it

Artificial Intelligence Basics: A Non-technical Introduction

The most efficient way to understand artificial intelligence is to start with the basics. You will find several artificial intelligence books that discuss AI for beginners. Still, this book, ‘Artificial Intelligence Basics: A Non-Technical Introduction’ by Tom Taulli, teaches you how to implement AI in your organization. It also teaches you how to strategize, set realistic goals, and deal with inherent risks like biases, employee resistance, and data quality. The book will introduce you to several concepts like machine learning, natural language processing, deep learning, and many more. The AI book is indispensable for giving you a headstart in artificial intelligence.

Link to the book – Artificial Intelligence Basics: A non-technical introduction

Artificial Intelligence: A Modern Approach by Peter Norvig and Stuart Russel

Once you have a basic understanding of the subfields of AI, it is logical to proceed and learn the ‘common framework’ of how it works. AI is a big field, and so is this AI textbook. The new edition focuses on a modern approach that cultivates an idea of an ‘intelligent agent,’ and AI is the study of such agents that perceive the environment and perform accordingly. Intended for undergraduates, it is one of the best books for AI that will enable you to understand the logic, probability, and continuous mathematics that form the foundation of AI systems.

Link to the book – Artificial Intelligence

Python Machine Learning: Machine Learning and Deep Learning with Python, Scikit-Learn, and TensorFlow – 3rd Edition

Artificial intelligence works by training specific algorithms to perform human tasks. If you wish to learn Python for artificial intelligence, it is one of the best books on artificial intelligence. Written by Sebastian Raschka, this will help you understand artificial intelligence conversationally. The 3rd edition is a step-by-step guide to help you build your ML systems and has been updated to include TensorFlow 2.0, new Keras API features, and the latest additions to Scikit-Learn. Compared to the previous edition, it has been expanded to cover reinforcement learning and sentiment analysis, a sub-field of natural language processing.

Link to the book – Python Machine Learning

Artificial Intelligence Engines: a Tutorial Introduction to the Mathematics of Deep Learning

Deep learning is a part of artificial intelligence based on artificial neural networks, and this domain enables computers to undertake classification tasks directly from texts, images, or sounds. Studying deep learning can be challenging as it involves leveraging data analytics to work faster than human minds. But Artificial Intelligence Engines: a tutorial introduction to the mathematics of deep learning is an ideal AI book for beginners. The book explains neural networks informally at first, followed by extensive mathematical analyses, and has been written informally so that readers do not have a hard time comprehending the concepts.

Link to the book – Artificial Intelligence Engines: a tutorial introduction to the mathematics of deep learning

Read More: Top Machine Learning Books

Some other books on Artificial Intelligence

Life 3.0: Being a Human in the Age of Artificial Intelligence

Written by Max Tegmark, it is one of those books about AI that posit a hypothetical scenario wherein artificial intelligence has exceeded human intelligence and has overtaken society. While other AI textbooks talk about the benefits one can reap with artificial intelligence, ‘Life 3.0’ talks about how the technology might be able to redesign itself someday. The first few chapters talk about the origin of intelligence millions of years ago and project future developments. It discusses the issues and describes several potential outcomes that could be achieved by integrating humans and machines. Both positive and negative scenarios have been portrayed in terms of Friendly AI or an AI Apocalypse. This book will give you a different perspective on how AI impacts your life.

Link for the book – Life 3.0: Being a human in the age of artificial intelligence

Applied Artificial intelligence: A handbook for Business Leaders

Many enterprises are leveraging artificial intelligence in their operations. Suppose you are a business owner looking for a practical guide to leverage machine learning techniques in your organization. In that case, Applied Artificial Intelligence is one of the best books on artificial intelligence. This book balances technical details and general content about the impact of artificial intelligence and machine learning technology. The first part of this book talks about the educational background of the state of artificial intelligence today. The second part will walk you through some strategic and methodological steps to implement AI projects. The last section discusses real-world examples of AI applications commonly used for standard business functions.

Link for the book – Applied Artificial intelligence: A handbook for Business Leaders

TensorFlow in 1 Day: Make your own Neural Network

If you are a machine learning enthusiast, TensorFlow in 1 Day: Make your own Neural Network is a great machine learning book for beginners. This book will guide you to build your neural network, a framework that works like the human brain. Written by Krishna Rungta, it aims to educate readers about TensorFlow, an open-source library for deep learning and traditional machine learning applications. The initial chapters discuss deep learning, TensorFlow, and other requisite frameworks. The chapters discuss all the essential packages you will need to build a recurrent neural network (RNN).

Link for the book – TensorFlow in 1 Day: Make your own Neural Network

The Emotion Machine: Commonsense Thinking, AI, and the Future of the Human Mind

The Emotion Machine is a brilliant AI textbook that explains thinking and the human mind by relating it to insightful technologies. Inspired by how human minds always work, Marvin Minsky created this book on artificial intelligence, thinking, and other related areas. Minsky is a computer science and AI expert who developed the book based on intuitions and feelings as forms of ‘thinking.’ The book explains how human minds think and process complex information. Once you understand the thinking process, you can build artificial bits of intelligence and machines to assist you in thinking and making better decisions after following the same patterns.

Link to the book – The Emotion Machine: Commonsense Thinking, AI, and the Future of the Human Mind

Human + Machine: Reimagining Work in the Age of AI

This AI book by Paul Daugherty and H. James Wilson highlights that artificial intelligence is no longer a futuristic concept and is gaining momentum. In Human + Machine, you will read about the paradigm shift of revolutionizing businesses with AI and making them more fluid and adaptive. If you look forward to incorporating technologies in your workplace, reading this book will convince you that AI transforms businesses into hybrid technological organizations that leverage technology to ensure real-time service provision and management.
Link to the book – Human + Machine: Reimagining Work in the Age of AI

Top NVIDIA GTC Announcements 2022

Disha Chopra

September 23, 2022

NVIDIA is a leading company involved with artificial intelligence and computing technologies. The company was founded by Jensen Huang in 1993 and has pioneered accelerating computing to overcome challenges that several firms face. NVIDIA has been hosting the GTC (GPU Technology Conference) since 2009. The conference aims to bring together researchers, pioneers, developers, and several IT professionals to discuss the latest technologies.

The NVIDIA GTC 2022 event commenced on September 19 and ended on September 22. The event hosted several AI experts and tech-invested companies. CEO Jensen Huang addressed several important topics and announcements about AI, Omniverse, GPU chips, NVIDIA’s partnerships, etc.

DLSS 3

DLSS is a neural graphics technology that scales performance by creating new frames and displaying higher resolution. The technology reconstructs temporal images to produce better ones by rendering pixels in fractions to maximize the frame rate. NVIDIA has announced the third generation, DLSS 3 (deep learning super sampling). The sampling framework clubs three innovations, DLSS Super Resolution, DLSS Frame Generation, and NVIDIA Reflex, into one. The updated technology utilizes the latest Optical Flow Accelerator (OFA) from the Ada Lovelace architecture. It is a substantial improvement as now the technology can predict entire frames, not just pixels. Another significant improvement is its ability to enhance CPU-limited gaming experience by boosting frame rates from 64 FPS to 135 FPS. Try the new Portal RTX game to experience the improvement credited to DLSS 3.

GeForce RTX 40 Series GPUs

Two years after the RTX 30 GPU series, NVIDIA announced the brand-new RTX 40 GPU series at the GTC event. The RTX 40 series is built on the company’s new Ada Lovelace architecture and delivers a significant performance improvement. The GPUs will enhance AI-driven graphics and content creation. With 24GB of GDDR6X memory, NVIDIA claims that the 40 series GPUs are two to four times faster than the previous generation and can deliver up to 100 FPS in 4K gaming. The GPUs are also compatible with DLSS 3, NVIDIA’s new deep-learning sampling framework.

NVIDIA Drive Thor

Thor is the next-generation system-on-a-chip that centralizes all autonomous vehicle functions on a single computer for better security. NVIDIA unveiled the super chip DRIVE Thor, built on the latest GPU and CPU innovations to deliver world-class performance. Thor is a significant upgrade of the DRIVE Orin, the company’s previous lineup for autonomous vehicle technology. The chip unified several distributed functions like digital cluster infotainment, assisted driving, and parking; for enhanced software iteration and efficient development.

LLMs and H100 GPUs in volume production

The GTC 2022 event attendees also witnessed the new large language models (LLMs) based on deep learning. Huang explained that these LLMs are one of the most vital AI models that run the digital economy. These models are engines of E-commerce, digital advertising, and searching. Following the same, he announced the new NeMo LLM service that makes AI more accessible by adding a conversational layer to the models. The service takes pre-trained models like the GPT-3 or Megatron and constructs a framework around them, saving model training time.

These enormous models require good computational power, for which Huang announced the NVIDIA H100 Tensor Core GPU chips with a next-gen transformer engine. The chips are in full production and will soon begin to be shipped.

NVIDIA DRIVE Concierge and DRIVE Chauffeur

The GTC keynote also talks in detail about the DRIVE Concierge and DRIVE Chauffeur, AI platforms aimed to make driving hassle-free by transforming the digital experience inside the car. With NVIDIA Concierge, people inside the vehicle have continuous access to real-time conversational AI because of DRIVE IX and Omniverse Avatar. Omniverse Avatar enables COncierge to function like a digital assistant and performs tasks like calling, booking reservations, alerting, etc. The company also posted a demo video of the Concierge AI showcasing the assistant on the center dashboard screen and helping the driver reach the destination.

The concierge is integrated with Chauffeur, AI-assisted driving framework, to provide high-quality 4D visualization with low-latency and 360-degree view so that the driver can sit idly while the Chauffeur drives. The chauffeur is based on the NVIDIA DRIVE AV SDK to tackle urban and highway traffic while following safety rules.

Jetson Orin Nano for Robotics

Several GTC announcements cater to robotics development. As per Huang, robotic computers are “the newest type of computers” to enable machines to move through the virtual worlds. He announced the new Jetson Orin Nano, a tiny robotics computer, to bring the DRIVE Orin technology to the market. This tiny computer is NVIDIA’s second-gen processor for robotics and is 80x faster than the previous Jetson Nano. Jetson Orin Nano runs on the Isaac robotics stack based on the ROS 2 GPU-driven framework and also features NVIDIA Isaac Sim, a robotics simulation platform. Huang made another significant announcement about containers for the NVIDIA Isaac platform accessible on the AWS marketplace for robotics developers who use AWS RoboMaker.

Omniverse Cloud

The GTC 2022 event also unveiled NVIDIA Omniverse Cloud, the company’s first Saas (software-as-a-service) and Iaas (infrastructure-as-a-service) framework. Omniverse is a complete suite of cloud-native services, including metaverse applications, robotics, and autonomous vehicle simulation. The Omniverse Cloud is powered by the NVIDIA Graphics Delivery Network (GDN), a distributed data center responsible for delivering high-performance and low-latency graphics in the company’s cloud gaming services. Omniverse Cloud allows people to experience collaborative 3D workflows without local computing infrastructure.

Some other GTC announcements:

NVIDIA announced the second-gen NVIDIA OVX powered by the L40 GPU and designed for building complex industrial applications. The new OVX-fitted systems will be shipped to companies including Lenovo, Supermicro, and Inspur by the first quarter of 2023.
NVIDIA also announced a partnership with The Broad Institute of MIT and Harvard to bring GPU-driven Clara Parabricks software to the Terra biomedical data platform. The company says the partnership will contribute to its deep learning models for identifying genetic disease-associated genetic variants.
NVIDIA announced another collaboration with Booz Allen to use AI by accelerating the GPU power of its cybersecurity platform built on NVIDIA’s Morpheus architecture.

You can refer to the GTC 2022 Keynote for more information.

Crypto mining data center Compute North filed for bankruptcy

Sahil Pawar

September 23, 2022

Crypto mining data center Compute North filed for bankruptcy

Compute North, one of the largest operators of crypto-mining data centers, filed for bankruptcy and revealed that its CEO stepped down as the rout in cryptocurrency prices weighs on the industry.

According to a filing, the company filed for Chapter 11 in the U.S. Bankruptcy Court for the Southern District of Texas.

Compute North in February announced a capital raise of $385 million, consisting of an $85 million Series C equity round and $300 million in debt financing. But it fell into bankruptcy as miners struggled to survive amid slumping bitcoin (BTC) prices, rising power costs, and record difficulty in mining bitcoin. The filing is likely to have negative implications for the industry. Compute North is one of the largest data center providers for miners and has multiple deals with other larger mining companies.

“The Company has initiated voluntary Chapter 11 proceedings to provide the opportunity to stabilize its business and implement a comprehensive restructuring process that will enable us to continue servicing our customers and partners and make the necessary investments to achieve our strategic objectives,” a spokesperson told CoinDesk.

The spokesperson added that CEO Dave Perrill stepped down earlier this month but will continue to serve on the board. Drake Harvey, a chief operating officer for the last year, has taken the role of president at Compute North, the spokesperson said.

Samsung launches Samsung Innovation Campus in India to upskill youth in AI, IoT, Big Data, Coding and Programming

Sahil Pawar

September 23, 2022

Samsung launches Samsung Innovation Campus in India

Samsung has launched its global CSR program Samsung Innovation Campus in India, creating industry-relevant skills and job opportunities for youth in technology domains such as Artificial Intelligence, the Internet of Things, Big Data, and Coding & Programming. With this, Samsung is partnering with new India’s growth story and strengthening its commitment to the Government’s Skill India initiative.

The program aims to upskill over 3,000 unemployed youth from 18-25 years of age in future technologies and enhance their employability in the pilot phase. These are critical technology skills for the Fourth Industrial Revolution. Samsung Innovation Campus will also take Samsung a step ahead in its vision of ‘Powering Digital India’ as the country’s most vital partner.

To execute the program, Samsung, India’s largest smartphone and consumer electronics brand, has partnered with India’s premier skill development organization, the Electronics Sector Skills Council of India (ESSCI), a National Skill Development Corporation (NSDC) approved entity. ESSCI will execute the program through its nationwide network of approved training and education partners.

During the program, participants will receive instructor-led blended classroom and online training through approved training and education partners of ESSCI across the country. Youth enrolled in the program will undergo online and classroom training. They will complete their hands-on capstone project work in their selected technology areas from Internet of Things, Big Data, Artificial Intelligence, and Coding & Programming.

They will also be given soft skills training to enhance their employability in relevant organizations. The participants will be mobilized through ESSCI’s education and training partners across India.

Those opting for the AI course will undergo 270 hours of theory training and 80 hours of project work. Those doing the IoT or the Big Data course will receive 160 hours of training and complete 80 hours of project work. Participants taking the Coding & Programming course will do 80 hours of training and be part of a 4-day Hackathon.

Meta sued for bypassing Apple privacy features

Sahil Pawar

September 23, 2022

Meta sued for bypassing Apple privacy features

Meta has been sued for allegedly building a secret workaround that enabkes the company to bypass privacy features Apple launched earlier last year to keep iPhone users from having their internet activity tracked.

Two Meta users filed the lawsuit in San Francisco, where the same claim was made last week, accusing the company of skirting Apple’s 2021 privacy rules and violating state and federal laws limiting the unauthorized collection of personal data.

The accusations are based on a study published by Felix Krause last August. Krause, a former Google employee, argued that Meta exploits the “in-app browser” — a feature that allows Facebook and Instagram users to visit a third-party website without leaving the platform — to “inject” a tracking code that allows the monitoring of all user interactions.

The practice, called Javascript injection, which in most cases is considered a type of malicious attack, enables the tech giant to follow users throughout the web after they click links on the Facebook and Instagram apps.

“This allows Meta to intercept, monitor and record its users’ interactions and communications with third parties, providing data to Meta that it aggregates, analyzes and uses to boost its advertising revenue,” the claimant read.

In response to the allegation, Meta admitted that the Facebook app tracks browser activity but refuted claims that user data was being unlawfully collected.

The lawsuit contends that Meta’s collection of user information via the Facebook and Instagram apps enables the company to get around Apple’s privacy regulations, which require all third-party applications to acquire user consent before tracking users’ online and offline activity.

Starting with iOS 14.5, Apple introduced App Monitoring Transparency, which enables users to choose whether or not to enable app tracking when they first open an app. The feature, according to Meta, has impacted the company’s revenue by more than $10 billion so far.

1...161162163...354 Page 162 of 354