Reliability is a fundamental challenge in operating large-scale machine learning (ML) infrastructures, particularly as the scale of ML models and training clusters continues to grow. Despite decades of research on infrastructure failures, the impact of job failures across different scales remains unclear. This paper presents a view of managing two large, multi-tenant ML clusters, providing quantitative analysis, operational experience, and our own perspective in understanding and addressing reliability concerns at scale. Our analysis reveals that while large jobs are most vulnerable to failures, smaller jobs make up the majority of jobs in the clusters and should be incorporated into optimization objectives. We identify key workload properties, compare them across clusters, and demonstrate essential reliability requirements for pushing the boundaries of ML training at scale. We hereby introduce a taxonomy of failures and key reliability metrics, analyze 11 months of data from two state-of-the-art ML environments with over 150 million A100 GPU hours and 4 million jobs. Building on our data, we fit a failure model to project Mean Time to Failure for various GPU scales. We further propose a method to estimate a related metric, Effective Training Time Ratio, as a function of job parameters, and we use this model to gauge the efficacy of potential software mitigations at scale. Our work provides valuable insights and future research directions for improving the reliability of AI supercomputer clusters, emphasizing the need for flexible, workload-agnostic, and reliability-aware infrastructure, system software, and algorithms.

The organizational use of artificial intelligence (AI) has rapidly spread across various sectors. Alongside the awareness of the benefits brought by AI, there is a growing consensus on the necessity of tackling the risks and potential harms, such as bias and discrimination, brought about by advanced AI technologies. A multitude of AI ethics principles have been proposed to tackle these risks, but the outlines of organizational processes and practices for ensuring socially responsible AI development are in a nascent state. To address the paucity of comprehensive governance models, we present an AI governance framework, the hourglass model of organizational AI governance, which targets organizations that develop and use AI systems. The framework is designed to help organizations deploying AI systems translate ethical AI principles into practice and align their AI systems and processes with the forthcoming European AI Act. The hourglass framework includes governance requirements at the environmental, organizational, and AI system levels. At the AI system level, we connect governance requirements to AI system life cycles to ensure governance throughout the system's life span. The governance model highlights the systemic nature of AI governance and opens new research avenues into its practical implementation, the mechanisms that connect different AI governance layers, and the dynamics between the AI governance actors. The model also offers a starting point for organizational decision-makers to consider the governance components needed to ensure social acceptability, mitigate risks, and realize the potential of AI.

In the legal world, and in particular the world of electronic discovery, artificial intelligence (AI) has been around for more than a decade. It is no longer unusual or controversial for organizations to use AI technologies in litigation, especially where large or complex data sets are involved. Legal teams now routinely turn to AI to defensibly accelerate the process of identifying documents likely to be responsive to requests for evidence. Innovations like technology assisted review (TAR), for example, rely heavily on machine learning and natural language processing to make connections and identify patterns within a body of data in a matter of seconds. This is work that would take even the most qualified human reviewers many, many hours to do manually, and with less accuracy.

In the legal world, and in particular the world of electronic discovery, artificial intelligence (AI) has been around for more than a decade.

The CIO's primary focus during this pandemic-induced work-from-home stretch has been to keep hardware healthy and data safe. But in America's factories, warehouses and distribution centers, where "work from work" is almost always the only option, the technology directives understandably center on protecting human capital. You might think the recipes for success between the two scenarios are very different. But it might surprise you to learn that many of the ingredients are very similar. Fortunately for Trilogy Health Services, which runs 28 senior living facilities across four midwestern states, the company already had nearly 80 percent of its 15,000 employees on its internal communication app when the virus hit.

In 2013, IBM and University of Texas Anderson Cancer Center developed an AI based Oncology Expert Advisor. According to IBM Watson, it analyzes patients medical records, summarizes and extracts information from vast medical literature, research to provide an assistive solution to Oncologists, thereby helping them make better decisions. According to an article on The Verge, the product demonstrated a series of poor recommendations. Like recommending a drug to a lady suffering from bleeding that would increase the bleeding. "A parrot with an internet connection" - were the words used to describe a modern AI based chat bot built by engineers at Microsoft in March 2016. 'Tay', a conversational twitter bot was designed to have'playful' conversations with users. It was supposed to learn from the conversations. It took literally 24 hours for twitter users to corrupt it.

That said, technologies--such as big data, cloud computing, supercomputers, artificial intelligence (AI), robotics, 3D printing, thermal imaging and 5G--are being used to effectively complement the traditional methods of increased hygiene, self- and forced quarantines, and enforced global travel bans. Having enforced traditional measures in place, for instance, police officers in China now wear AI-powered helmets that can automatically record the temperatures of pedestrians. The high-tech headgear has an infrared camera, and sounds an alarm if anyone in a radius of 16ft has fever. Equipped with the facial-recognition technology, it can also display the pedestrian's personal information, such as their name on a virtual screen. Officials at railway stations, airports and in other public areas in India, too, are using smart thermal scanners to record temperatures from a distance, thus helping in identifying potential coronavirus carriers.

TYSONS, VA – Appian (NASDAQ: APPN) today announced the latest version of the Appian Platform. The new release of the low-code application development platform increases the speed and business impact of low-code automation for developers, administrators, and end-users. The latest version delivers enhancements to Appian AI, further-expansion of Appian's Connected Systems architecture, integrated Health Check in every application, and simplified DevOps, making it easier than ever to develop, deploy, change, and manage Appian applications. Appian AI, a fast way to add best-of-breed artificial intelligence to any Appian application, now offers Google Cloud Translation as a Connected System. Customers can enable any app to detect languages and translate text with no coding. In addition, this release provides an updated Google Cloud Vision Connected System which now offers integration with Optical Character Recognition (OCR).

An increasing number of software coding tasks are being handed off to AI functions, especially in ... [ ] the low-code & no-code arenas. What software needs now is more AI. This is the universal mantra that every software application development and data platform company will beat out relentlessly throughout 2020. The rise of Artificial Intelligence (AI) and the Machine Learning (ML) processes that feed it and make it smarter has driven every software industry specialist to look for avenues where it can surgically enhance its products and services with additional layers of automated intelligence. Low-code software company Appian has clearly been drinking from the same AI Kool-Aid pot as everybody else; the company's latest platform release features a range of AI assists designed to make low code software development a more intelligently abetted process.

April last year, a medical device powered by artificial intelligence (AI) received approval from the US Food and Drug Administration (USFDA), marking a historic moment in healthcare globally. The IDx-DR, a software algorithm that uses AI to analyse images of the eye using a camera, achieved an 87.4% accuracy rate while detecting'more than mild' diabetic retinopathy, a condition where high blood sugar levels damage the blood vessels in the retina. For IT services firms, which are already developing AI and machine language (ML) tools for other uses and industries, extending AI and ML capabilities to healthcare is a fairly non-complex process, and comes with a large upside. Rather than doing it entirely on their own though, these companies are partnering hospital chains and niche players in the field to acquire the required domain expertise. For instance, Japanese technology firm NTT DATA Services tied up with Pune's Deenanath Mangeshkar Hospital last year to use an AI-based solution to diagnose emphysema, a chronic condition of the lungs.