As organizations continue to produce vast quantities of data, they increasingly need platforms that allow them to analyze, store, and extract meaningful insights from that data. And research from analyst firms like Gartner offer an important way for organizations to evaluate and compare cloud data warehouse providers.

Earlier this year, Gartner named Google a Leader in the 2019 Gartner Magic Quadrant for Data Management Solutions for Analytics (DMSA) (report available here). This evaluation covers Google Cloud’s core data analytics offerings, including BigQuery, a serverless, managed data warehouse, Cloud Dataproc, a managed Spark and Hadoop service, and Cloud Dataflow, which enables you to stream and batch-process your data. Here are a few takeaways:

Simplicity and speed

BigQuery’s performance permits complex queries on large-scale data sets to return in seconds, and a substantial number of BigQuery customers maintain data warehouses that store more than 50 terabytes (and a few customers now use more than 100 petabytes). More than half of these customers are loading data either continuously or many times per day. These customers value the ability to extract, transform, load, and analyze their data on a serverless platform, all without maintaining any underlying infrastructure.

A versatile serverless data warehouse

One of BigQuery’s major advantages is its ability to allow customers to address a wide variety of use cases—from a traditional data warehouse to data science. Over the past year, we’ve worked hard to introduce new features in BigQuery like data types for financial and monetary uses, BigQuery GIS for geospatial data, and machine learning capabilities through BigQuery ML. BigQuery’s continuous ingest capabilities make it suitable for an operational—or a real-time—serverless data warehouse.

An expanding ecosystem

With overall increased market adoption, our analytics offerings continue to benefit from a fast-growing partner ecosystem of service providers, and business intelligence (BI) and data integration vendors. In particular, in 2018 we expanded our partnerships with established industry providers, including Confluent, Dell Boomi, Informatica, Looker, Reltio, Tableau, and ThoughtSpot.

More and more organizations are finding value in Google Cloud’s serverless data warehouse and analytics offerings. If you’d like to learn more, you can download a complimentary copy of the Gartner Magic Quadrant for Data Management Solutions for Analytics on the Google site (requires an email address).

_{Gartner does not endorse any vendor, product or service depicted in its research publications, and does not advise technology users to select only those vendors with the highest ratings or other designation. Gartner research publications consist of the opinions of Gartner’s research organization and should not be construed as statements of fact. Gartner disclaims all warranties, expressed or implied, with respect to this research, including any warranties of merchantability or fitness for a particular purpose.}

According to a National Academy of Medicine discussion paper, social determinants of health (SDoH) account for upwards of 80% of a population’s health outcomes. Breaking this down further, 50% come just from socioeconomic and physical environment factors such as education, employment, income, family and social support, community safety, air and water quality, and access to housing and transit. SDoH determine access to and quality of healthcare, and are contributors to a system of disparate access to care.

We believe that to more efficiently provide comprehensive healthcare access to more people, providers will leverage technology to blend in-person and virtual modes of care delivery.

In this article, Amwell and Google Cloud examine five ways telehealth – as part of a provider’s overall care model – can help democratize access to healthcare. (graphic url) 

Remove distance as a barrier to care

8.6 million Americans live more than 30 minutes from their nearest hospital. Long drives can deter patients from seeking care or maintaining routine visits. On the flip side, 92% of Americans nationwide have access to wired broadband in the home or through mobile broadband. Therefore, having a virtual visit just a click away can help remove barriers to care, like distance.

Eliminate the risk of unnecessary exposure

Virtual care means that patients don’t have to worry about potential exposure in transit, while sitting in waiting rooms, or from direct interactions during in-person health visits. This is particularly relevant when it comes to those with chronic diseases or underlying conditions. Telehealth provides patients who may be more susceptible to diseases with access to continuous healthcare without putting them at higher risk for developing more severe symptoms. Virtual visits can occur in the comfort –and safety– of patients’ homes.

Extend access to specialized care

55% of preventable hospitalization or mortality in rural settings is due to lack of access to specialty care. With telehealth, physical proximity to specialized services–typically in urban areas–can be reduced as a limiting factor. Virtual solutions grant everyone access to top specialists, regardless of location.

Save time and money

By augmenting in-person visits with telehealth applications, providers can benefit from greater efficiencies in scheduling, helping to improve their bottom line, and add more flexibility to their workday. Meanwhile, patients can spend less on travel and childcare, limit time taken off work, and save on other costs associated with in-person visits — for a savings of $35 to $690 per visit. In a survey by the COVID-19 Healthcare Coalition, 76% of the 2,000+ patients surveyed across the US responded that transportation was removed as a barrier, 65% reported they no longer had to take time off from work for an appointment, and 67 percent reporting lower costs than an in-person visit.

Combat physician shortage and fatigue

Research shows there will be a shortage of more than 100,000 doctors by 2030. In the midst of a physician shortage, telehealth can help improve care delivery and make it more efficient, ensuring more people can still have their healthcare needs addressed. Additionally, by integrating intelligence such as case triaging along with telehealth into a virtual care model, providers can help reduce clinician burnout.

Amwell and Google Cloud are partnering to deliver transformative telehealth solutions that will make it easier for more patients to receive care and improve patient and clinician experiences across the continuum of care. One example includes embedding real-time captioning and translation services powered by Google Cloud’s AI and NLP technologies within the Amwell platform to increase health access and understanding for more people.

As physicians, we are excited that the future of healthcare will continue to blend cloud technologies like EHR-integrated telehealth platforms, AI, healthcare-trained virtual agents along with in-person care to create an integrated hybrid care model that will improve patient outcomes and unburden providers, all while expanding access to broader patient populations.

To learn more, download the whitepaper “Healthcare’s Virtual Transformation,” written in conjunction with Becker’s Hospital Review.

^{1: “Social Determinants of Health 101 for Health Care: Five Plus Five,” National Academy of Medicine, October 2017}

The latest round of MLPerf benchmark results have been released, and Google’s TPU v4 supercomputers demonstrated record-breaking performance at scale. This is a timely milestone since large-scale machine learning training has enabled many of the recent breakthroughs in AI, with the latest models encompassing billions or even trillions of parameters (T5, Meena, GShard, Switch Transformer, and GPT-3). 

Google’s TPU v4 Pod was designed, in part, to meet these expansive training needs, and TPU v4 Pods set performance records in four of the six MLPerf benchmarks Google entered using TensorFlow and JAX. These scores are a significant improvement over our winning submission from last year and demonstrate that Google once again has the world’s fastest machine learning supercomputers. These TPU v4 Pods are already widely deployed throughout Google data centers for our internal machine learning workloads and will be available via Google Cloud later this year.

Figure 1: Speedup of Google’s best MLPerf Training v1.0 TPU v4 submission over the fastest non-Google submission in any availability category – in this case, all baseline submissions came from NVIDIA. Comparisons are normalized by overall training time regardless of system size. Taller bars are better.¹

Let’s take a closer look at some of the innovations that delivered these ground-breaking results and what this means for large model training at Google and beyond. 

Google’s continued performance leadership

Google’s submissions for the most recent MLPerf demonstrated leading top-line performance (fastest time to reach target quality), setting new performance records in four benchmarks. We achieved this by scaling up to 3,456 of our next-gen TPU v4 ASICs with hundreds of CPU hosts for the multiple benchmarks. We achieved an average of 1.7x improvement in our top-line submissions compared to last year’s results. This means we can now train some of the most common machine learning models in a matter of seconds.

Figure 2: Speedup of Google’s MLPerf Training v1.0 TPU v4 submission over Google’s MLPerf Training v0.7 TPU v3 submission (exception: DLRM results in MLPerf v0.7 were obtained using TPU v4). Comparisons are normalized by overall training time regardless of system size. Taller bars are better. Unet3D not shown since it is a new benchmark for MLPerf v1.0.²

We achieved these performance improvements through continued investment in both our hardware and software stacks. Part of the speedup comes from using Google’s fourth-generation TPU ASIC, which offers a significant boost in raw processing power over the previous generation, TPU v3. 4,096 of these TPU v4 chips are networked together to create a TPU v4 Pod, with each pod delivering 1.1 exaflop/s of peak performance.

Figure 3: A visual representation of 1 exaflop/s of computing power. If 10 million laptops were running simultaneously, then all that computing power would almost match the computing power of 1 exaflop/s.

In parallel, we introduced a number of new features into the XLA compiler to improve the performance of any ML model running on TPU v4. One of these features provides the ability to operate two (or potentially more) TPU cores as a single logical device using a shared uniform memory access system. This memory space unification allows the cores to easily share input and output data – allowing for a more performant allocation of work across cores. A second feature improves performance through a fine-grained overlap of compute and communication. Finally, we introduced a technique to automatically transform convolution operations such that space dimensions are converted into additional batch dimensions. This technique improves performance at the low batch sizes that are common at very large scales.

Enabling large model research using carbon-free energy

Though the margin of difference in topline MLPerf benchmarks can be measured in mere seconds, this can translate to many days worth of training time on the state-of-the-art models that comprise billions or trillions of parameters. To give an example, today we can train a 4 trillion parameter dense Transformer with GSPMD on 2048 TPU cores. For context, this is over 20 times larger than the GPT-3 model published by OpenAI last year. We are already using TPU v4 Pods extensively within Google to develop research breakthroughs such as MUM and LaMDA, and improve our core products such as Search, Assistant and Translate. The faster training times from TPUs result in efficiency savings and improved research and development velocity. Many of these TPU v4 Pods will be operating at or near 90% carbon free energy. Furthermore, cloud datacenters can be ~1.4-2X more energy efficient than typical datacenters, and the ML-oriented accelerators – like TPUs – running inside them can be ~2-5X more effective than off-the-shelf systems. 

We are also excited to soon offer TPU v4 Pods on Google Cloud, making the world’s fastest machine learning training supercomputers available to customers around the world. Cloud TPUs support leading frameworks such as TensorFlow, PyTorch, and Jax, and we recently released an all-new Cloud TPU system architecture that provides direct access to TPU host machines, greatly improving the user experience. 

Want to learn more? 

Please contact your Google Cloud sales representative to request early access to Cloud TPU v4 Pods. We are excited to see how you will expand the machine learning frontier with access to exaflops of TPU computing power!

^{1. All results retrieved from www.mlperf.org on June 30, 2021. MLPerf name and logo are trademarks. See www.mlperf.org for more information. Chart uses results 1.0-1067, 1.0-1070, 1.0-1071, 1.0-1072, 1.0-1073, 1.0-1074, 1.0-1075, 1.0-1076, 1.0-1077, 1.0-1088, 1.0-1089, 1.0-1090, 1.0-1091, 1.0-1092.
2. All results retrieved from www.mlperf.org on June 30, 2021. MLPerf name and logo are trademarks. See www.mlperf.org for more information. Chart uses results 0.7-65, 0.7-66, 0.7-67, 1.0-1088, 1.0-1090, 1.0-1091, 1.0-1092.}

Notified is a leading communications cloud for events, public relations, and investor relations to drive meaningful insights and outcomes. They provide communications solutions to effectively reach and engage customers, investors, employees, and the media.

One of Notified’s Public Relations solutions is the ‘Media Contact Database’ that allows customers to discover media and influencers in a unique media database powered by AI and human-curated research. 

The goal of the initiative is to expand the scope of the AI driven, dynamically discovered influencers, and analyze online news articles using AI/ML technologies to extract entities and classify content. The prior process to extract insights from news articles provided only 30-40% of the desired results, and there were accuracy and stability issues that resulted in a lot of manual intervention.

Journalist Beat

A key outcome of the AI driven process is to identify the ‘Journalist Beat’. A Journalist Beat essentially summarizes the individual’s area of focus such as a sports writer, financial journalist etc. 

Three options were evaluated for the AI/ML process to generate the Journalist Beats :

Option 1:  Topic ML

Unsupervised ML approach to determine the commonly used terms.

• Pro: Common approach to grouping documents and determine similar text
• Con: Unbounded list of text

Option 2: ML Classification

Build classification models (supervised) to map reference articles to ‘Beats’ 

• Pro: Aligns to ‘Research Analytics’ existing processes
• Con: Time to build and maintain ML models for hundreds of beats.

Option 3: GCP Context Classification

Leverage GCP’s Natural Language API for initial classification and as input to Notified single model

• Pro: Aligns to ‘Research Analytics’ without building ML models.

Ultimately the GCP Natural Language API solution was chosen because of the speed of execution and a high level of accuracy with the pretrained models. The Notified team was able to launch the product feature within a few weeks, without ever needing to do extensive data collection and train the models. 

Here is the high level process that was implemented for Journalist Beats.

Since Notified supports curated media contacts globally, news articles were instantly translated to English using GCP Translation API. GCP Natural Language API’s solution to classify text was used to analyze the translated text and generate the list of content categories.

Solution Architecture

Here is a sample solution architecture for the ‘Discovered Journalist’ process.

Three core principles guided the above architecture – Serverless & Fully Managed, Scalability & Elasticity for flexibility and to optimize costs, API led real-time processing.

In addition to the GCP Natural Language API and Translation API below are a few serverless GCP products that were part of the automated solution:

• BigQuery is Google Cloud’s fully managed, petabyte-scale, and cost-effective analytics data warehouse that lets you run analytics over vast amounts of data in near real time.
• Cloud Run is a fully managed serverless platform that can be used to develop and deploy highly scalable containerized applications.
• Cloud Tasks is a fully managed service that allows you to manage the execution, dispatch, and delivery of a large number of distributed tasks.

The powerful pre-trained models of the Natural Language API provide a comprehensive set of features to apply natural language understanding to applications such as sentiment analysis, entity analysis, entity sentiment analysis, content classification, and syntax analysis. 

Notified looks ahead to super-scaling

In an effort to even further improve its best in class ‘Media Contact Database’, Notified looks to super scale the above AI driven Influencer Discovery process to the order of 100+ million news articles per month. It plans to expand the scope of entities extracted from the news articles and provide a news exploration service for its customers by performing intelligent entity-based searches.To watch your markets evolve, see how competitors add AI insights. To actually stay in the market, make AI the main driver of your product road maps. GCP Natural Language API accelerated our ability to adopt AI at scale.Thomas Squeo, CTO, Notified

Acknowledgments

We’d like to thank our collaborators at Google and Notified for making this blog post possible. Thanks to Arpit Agrawal at MediaAgility for contributing to this blog post.

To learn more about how Google Cloud Natural Language AI can help your enterprise, try out an interactive demo and take the next step, visit the product overview page here.

Enterprise agility and the ability to innovate, adapt and respond quickly to the ever-changing risk and regulatory landscape is no longer a choice, but the cornerstone of successful digital transformation and commercial growth. Traditional access to and ways of managing data invariably create challenges in dealing with multiple data repositories, reconciliations, fire-drills, etc.

In response, the move to cloud is increasing significantly. It enables risk analytics and regulatory reporting at scale in a secure environment with data storage, management and encryption capabilities as a standard. In addition, as regulatory reporting requirements become more granular, machine learning can help facilitate new insights and allow for risk management to become more embedded into operational processes.

This webinar will address the day-to-day challenges in risk management and regulatory compliance, while also exploring how technological innovations can provide massive improvements and potential.

Key themes

• Real-life data challenges in the eyes of risk managers: can compliance, fraud detection and identifying liquidity positions be improved through the use of AI?
• Innovative approaches to streamline regulatory reporting to derive deeper customer insights from data at the moment of truth.
• Reimagining operations: how to modernise the data infrastructure to accommodate data explosion, drive flexibility and deliver a more cost effective outcome.