A 3D-printed Facebook logo is seen placed on a keyboard in this illustration taken March 25, 2020. REUTERS/Dado Ruvic/Illustration

Sept 9 (Reuters) - Facebook Inc (FB.O) is developing a machine learning chip to handle tasks such as content recommendation to users, The Information reported on Thursday, citing two people familiar with the project.

The company has developed another chip for video transcoding to improve the experience of watching recorded and live-streamed videos on its apps, according to the report.

Facebook's move comes as major technology firms, including Apple Inc (AAPL.O) Amazon.com Inc (AMZN.O) and Alphabet Inc's (GOOGL.O) Google, are increasingly ditching traditional silicon providers to design their own chips to save up on costs and boost performance. (https://reut.rs/3E0NlVN)

In a 2019 blog, Facebook said it was building custom chip designs specially meant to handle AI inference and video transcoding to improve performance, power and efficiency of its infrastructure, which at that time served 2.7 billion people across all its platforms.

The company had also said it would work with semiconductor players such as Qualcomm Inc (QCOM.O), Intel Corp (INTC.O) and Marvell Technology (MRVL.O) to build these custom chips as general-purpose processors alone would not be enough to manage the volume of workload Facebook's systems handled.

However, The Information's report suggests that Facebook is designing these chips completely in-house and without the help of these firms.

"Facebook is always exploring ways to drive greater levels of compute performance and power efficiency with our silicon partners and through our own internal efforts," a company spokesperson said.

Reporting by Chavi Mehta in Bengaluru; Editing by Anil D'Silva

Our Standards: The Thomson Reuters Trust Principles.

See the original post:
Facebook developing machine learning chip - The Information - Reuters

The public has lost confidence in our public health agencies and officialsand also our political leadersthanks to confusing and contradictory pronouncements and policies related to the coronavirus pandemic.

But don't think for a moment that leadership failures in health care are confined to COVID-19. Two other recent stories highlight how both government and industry are missing an opportunity to use new machine learning technology to deliver better health care at a lower cost.

In the government's case, Lina Khan's activist Federal Trade Commission (FTC) is opposing a merger between Illumina and GRAIL, two medical technology companies. The latter company, which spun off from Illumina in 2015, has developed a test capable of providing early detection of 50 different types of cancer. Bringing GRAIL back under the Illumina framework would get this life-saving technology to market faster and more efficientlybut the FTC couldn't let a little thing like saving lives get in the way of its anti-business agenda. To its credit, Illumina closed the deal in August without waiting for approval from the FTC or European regulators.

Unfortunately, the government isn't the only entity that too often makes decisions without fully understanding how disruptive innovation really works. The frustrating case of Epic Systems' Early Detection of Sepsis model shows the industry itself can fall prey to a Luddite mindset.

Epic's model is designed to detect and prevent sepsis, a leading cause of death that also accounts for 5 percent of U.S. hospitalization costs. Many of these deaths and the associated costs can be prevented with early diagnosis. Epic is so committed to helping hospitals reduce sepsis-related deaths that it developed and gives awayfor freean AI-powered early-warning model that helps alert doctors and nurses when a patient might need a second look.

The sepsis algorithm has produced encouraging results with customers, but is facing criticism in the health care industry press. In one peer-reviewed study, Prisma reported a 22 percent decrease in mortalitywhich could translate to millions of lives saved if it were implemented globally. More recently, in a controlled clinical trial, MetroHealth found a meaningful reduction in mortality and length of stay, and reduced the time to antibiotic treatment of septic patients in the ED by almost an hour.

Critics of Epic's algorithm have claimed that it is not yet good enough at detecting sepsis cases. But this accusation reveals a misunderstanding of the technology involved. Both the GRAIL and Epic models utilize machine learninga form of artificial intelligence that compares information from a test or a patient's medical record against vast amounts of data about previous cases with known outcomes. By its nature, machine learning gets better as it acquires new information. Failure to account for this fact has led many in government and media to dismiss promising innovations.

Machine learning cannot replace the expertise of a human doctor or nurse, but it offers those human health care providers a powerful new tool to see patterns and evidence they could never notice on their own. These cutting-edge technologies have the potential to save millions of lives and drastically reduce the cost of health care, if we let them.

Getting these complex algorithms right takes time, but we cannot allow the perfect to be the enemy of the already excellent. In the GRAIL case, the government needs to do what it always needs to do, and just get out of the way. In Epic's case, the industry needs to develop a deeper appreciation for how disruptive innovation works and work closely with bold creators to bring revolutionary technologies to life.

Steve Forbes is Chairman and Editor-in-Chief of Forbes Media.

The views expressed in this article are the writer's own.

See the original post:
Government and Industry May Miss Health Care's Machine Learning Moment | Opinion - Newsweek

This article is part of ourreviews of AI research papers, a series of posts that explore the latest findings in artificial intelligence.

The last decades growing interest in deep learning was triggered by the proven capacity of neural networks in computer vision tasks. If you train a neural network with enough labeled photos of cats and dogs, it will be able to find recurring patterns in each category and classify unseen images with decent accuracy.

What else can you do with an image classifier?

In 2019, a group of cybersecurity researchers wondered if they could treat security threat detection as an image classification problem. Their intuition proved to be well-placed, and they were able to create a machine learning model that could detect malware based on images created from the content of application files. A year later, the same technique was used to develop a machine learning system that detects phishing websites.

The combination of binary visualization and machine learning is a powerful technique that can provide new solutions to old problems. It is showing promise in cybersecurity, but it could also be applied to other domains.

The traditional way to detect malware is to search files for known signatures of malicious payloads. Malware detectors maintain a database of virus definitions which include opcode sequences or code snippets, and they search new files for the presence of these signatures. Unfortunately, malware developers can easily circumvent such detection methods using different techniques such as obfuscating their code or using polymorphism techniques to mutate their code at runtime.

Dynamic analysis tools try to detect malicious behavior during runtime, but they are slow and require the setup of a sandbox environment to test suspicious programs.

In recent years, researchers have also tried a range of machine learning techniques to detect malware. These ML models have managed to make progress on some of the challenges of malware detection, including code obfuscation. But they present new challenges, including the need to learn too many features and a virtual environment to analyze the target samples.

Binary visualization can redefine malware detection by turning it into a computer vision problem. In this methodology, files are run through algorithms that transform binary and ASCII values to color codes.

In a paper published in 2019, researchers at the University of Plymouth and the University of Peloponnese showed that when benign and malicious files were visualized using this method, new patterns emerge that separate malicious and safe files. These differences would have gone unnoticed using classic malware detection methods.

According to the paper, Malicious files have a tendency for often including ASCII characters of various categories, presenting a colorful image, while benign files have a cleaner picture and distribution of values.

When you have such detectable patterns, you can train an artificial neural network to tell the difference between malicious and safe files. The researchers created a dataset of visualized binary files that included both benign and malign files. The dataset contained a variety of malicious payloads (viruses, worms, trojans, rootkits, etc.) and file types (.exe, .doc, .pdf, .txt, etc.).

The researchers then used the images to train a classifier neural network. The architecture they used is the self-organizing incremental neural network (SOINN), which is fast and is especially good at dealing with noisy data. They also used an image preprocessing technique to shrink the binary images into 1,024-dimension feature vectors, which makes it much easier and compute-efficient to learn patterns in the input data.

The resulting neural network was efficient enough to compute a training dataset with 4,000 samples in 15 seconds on a personal workstation with an Intel Core i5 processor.

Experiments by the researchers showed that the deep learning model was especially good at detecting malware in .doc and .pdf files, which are the preferred medium for ransomware attacks. The researchers suggested that the models performance can be improved if it is adjusted to take the filetype as one of its learning dimensions. Overall, the algorithm achieved an average detection rate of around 74 percent.

Phishing attacks are becoming a growing problem for organizations and individuals. Many phishing attacks trick the victims into clicking on a link to a malicious website that poses as a legitimate service, where they end up entering sensitive information such as credentials or financial information.

Traditional approaches for detecting phishing websites revolve around blacklisting malicious domains or whitelisting safe domains. The former method misses new phishing websites until someone falls victim, and the latter is too restrictive and requires extensive efforts to provide access to all safe domains.

Other detection methods rely on heuristics. These methods are more accurate than blacklists, but they still fall short of providing optimal detection.

In 2020, a group of researchers at the University of Plymouth and the University of Portsmouth used binary visualization and deep learning to develop a novel method for detecting phishing websites.

The technique uses binary visualization libraries to transform website markup and source code into color values.

As is the case with benign and malign application files, when visualizing websites, unique patterns emerge that separate safe and malicious websites. The researchers write, The legitimate site has a more detailed RGB value because it would be constructed from additional characters sourced from licenses, hyperlinks, and detailed data entry forms. Whereas the phishing counterpart would generally contain a single or no CSS reference, multiple images rather than forms and a single login form with no security scripts. This would create a smaller data input string when scraped.

The example below shows the visual representation of the code of the legitimate PayPal login compared to a fake phishing PayPal website.

The researchers created a dataset of images representing the code of legitimate and malicious websites and used it to train a classification machine learning model.

The architecture they used is MobileNet, a lightweight convolutional neural network (CNN) that is optimized to run on user devices instead of high-capacity cloud servers. CNNs are especially suited for computer vision tasks including image classification and object detection.

Once the model is trained, it is plugged into a phishing detection tool. When the user stumbles on a new website, it first checks whether the URL is included in its database of malicious domains. If its a new domain, then it is transformed through the visualization algorithm and run through the neural network to check if it has the patterns of malicious websites. This two-step architecture makes sure the system uses the speed of blacklist databases and the smart detection of the neural networkbased phishing detection technique.

The researchers experiments showed that the technique could detect phishing websites with 94 percent accuracy. Using visual representation techniques allows to obtain an insight into the structural differences between legitimate and phishing web pages. From our initial experimental results, the method seems promising and being able to fast detection of phishing attacker with high accuracy. Moreover, the method learns from the misclassifications and improves its efficiency, the researchers wrote.

[architecture]

I recently spoke to Stavros Shiaeles, cybersecurity lecturer at the University of Portsmouth and co-author of both papers. According to Shiaeles, the researchers are now in the process of preparing the technique for adoption in real-world applications.

Shiaeles is also exploring the use of binary visualization and machine learning to detect malware traffic in IoT networks.

As machine learning continues to make progress, it will provide scientists new tools to address cybersecurity challenges. Binary visualization shows that with enough creativity and rigor, we can find novel solutions to old problems.

Original post:
Computer vision and deep learning provide new ways to detect cyber threats - TechTalks

The U.S. Department of Energy (DOE) advanced Computational and Data Infrastructures (CDIs) such as supercomputers, edge systems at experimental facilities, massive data storage, and high-speed networks are brought to bear to solve the nations most pressing scientific problems.

The problems include assisting in astrophysics research, delivering new materials, designing new drugs, creating more efficient engines and turbines, and making more accurate and timely weather forecasts and climate change predictions.

Increasingly, computational science campaigns are leveraging distributed, heterogeneous scientific infrastructures that span multiple locations connected by high-performance networks, resulting in scientific data being pulled from instruments to computing, storage, and visualisation facilities.

However, since these federated services infrastructures tend to be complex and managed by different organisations, domains, and communities, both the operators of the infrastructures and the scientists that use them have limited global visibility, which results in an incomplete understanding of the behaviour of the entire set of resources that science workflows span.

Although scientific workflow systems increase scientists productivity to a great extent by managing and orchestrating computational campaigns, the intricate nature of the CDIs, including resource heterogeneity and the deployment of complex system software stacks, pose several challenges in predicting the behaviour of the science workflows and in steering them past system and application anomalies.

Our new project will provide an integrated platform consisting of algorithms, methods, tools, and services that will help DOE facility operators and scientists to address these challenges and improve the overall end-to-end science workflow.

Research professor of computer science and research director at the University of Southern California

Under a new DOE grant, the project aims to advance the knowledge of how simulation and machine learning (ML) methodologies can be harnessed and amplified to improve the DOEs computational and data science.

The project will add three important capabilities to current scientific workflow systems (1) predicting the performance of complex workflows; (2) detecting and classifying infrastructure and workflow anomalies and explaining the sources of these anomalies; and (3) suggesting performance optimisations. To accomplish these tasks, the project will explore the use of novel simulation, ML, and hybrid methods to predict, understand, and optimise the behaviour of complex DOE science workflows on DOE CDIs.

Assistant director for network research and infrastructure at RENCI stated that in addition to creating a more efficient timeline for researchers, we would like to provide CDI operators with the tools to detect, pinpoint, and efficiently address anomalies as they occur in the complex DOE facilities landscape.

To detect anomalies, the project will explore real-time ML models that sense and classify anomalies by leveraging underlying spatial and temporal correlations and expert knowledge, combine heterogeneous information sources, and generate real-time predictions.

Successful solutions will be incorporated into a prototype system with a dashboard that will be used for evaluation by DOE scientists and CDI operators. The project will enable scientists working on the frontier of DOE science to efficiently and reliably run complex workflows on a broad spectrum of DOE resources and accelerate time to discovery.

Furthermore, the project will develop ML methods that can self-learn corrective behaviours and optimise workflow performance, with a focus on explainability in its optimisation methods. Working together, the researchers behind Poseidon will break down the barriers between complex CDIs, accelerate the scientific discovery timeline, and transform the way that computational and data science are done.

As reported by OpenGov Asia, the U.S. Department of Energys (DOE) Argonne National Laboratory is leading efforts to couple Artificial Intelligence (AI) and cutting-edge simulation workflows to better understand biological observations and accelerate drug discovery.

Argonne collaborated with academic and commercial research partners to achieve near real-time feedback between simulation and AI approaches to understand how two proteins in the SARS-CoV-2 viral genome interact to help the virus replicate and elude the hosts immune system.

Read the original here:
Leveraging Artificial Intelligence and Machine Learning to Solve Scientific Problems in the U.S. - OpenGov Asia