Takeaway: Two of technology’s biggest trends – machine learning and cloud – are teaming up, and it’s sure to cause some innovation (and some disruption) in the enterprise.
Much of the cloud’s brief history has been characterized by the race to provide bulk compute and storage services at the lowest price point. The thinking was that once the enterprise becomes accustomed to the cloud as a cheaper alternative to traditional data infrastructure, it will then be on the path to consume more specialized services that generate higher revenues.
Heading into the new year, it seems this strategy is paying off better than most people had expected. Not only has the enterprise become increasingly willing to move critical workloads to the cloud, but it is also looking to tap an increasingly diverse portfolio of intelligent and cognitive services that simply do not exist anywhere but the cloud at the moment.
A case in point is Amazon’s P3 instances, which the company has recently upgraded with the new Nvidia Volta GPU. As HPC Wire points out, Amazon is bypassing the current Pascal line of accelerators in favor of the Volta 100, which offers 12 times the throughput of the Pascal for applications like deep learning training and inference. Each P3 instance is now backed by the Intel Xeon E5 and up to eight V100s, each of which provides more than 5,000 CUDA cores plus 640 Tensor cores to deliver upwards of 125 teraflops and mixed-precision performance. P3 instances are currently available in the U.S. East and West regions, as well as the EU and Asia Pacific regions via on-demand purchase or reserved or spot pricing.
Meanwhile, Google is turning its AI prowess toward tailored solutions for key industry verticals like healthcare. The company is forging deep ties to key applications developers though its Launchpad Studio machine learning platform, which seeks to cultivate start-ups that have the potential to vastly improve – or disrupt, depending on your point of view – established business processes. Among the first takers are Augmedix, which is using the Google Glass platform to automate prescription processing, and BrainQ, which is using neural networks and machine learning to customize the treatment of brain and spinal injuries. Other projects include advances in plug-and-play wearable technology and enhanced computer vision capabilities that may help researchers understand the biomechanics of infection. (Get the basics on machine learning in Machine Learning 101.)
Read Also: What is Artificial Intelligence(AI)?
For a company like Microsoft, which has a strong presence both in the cloud and the data center, AI is an effective tool to help customers make the most of hybrid infrastructure. EWeek reports that the company has added AI capabilities to the SQL Server 2017 platform, along with Linux support and DevOps-friendly application and container tools. At the same time, the Azure cloud is available to take on high-scale workloads in what General Manager John Chirapurath calls a “data plus AI” strategy. The goal is to leverage services like Azure Machine Learning in support of Hadoop and other big data workloads to allow the enterprise to quickly ramp up IoT and digital transformation strategies on the infrastructure they deem most appropriate for their needs. (Learn more on big data in the cloud in The Cloud: The Ultimate Tool for Big Data Success.)
Read Also: Artificial Intelligence vs Machine Learning
Even leaders in the “race to the bottom” pricing wars of the past are starting to see the benefits of a more intelligent service level. Storage specialist Box recently unveiled the new BoxSkills framework designed to help customers increase the value of the data they’ve placed in Box repositories. The system uses machine learning and other tools to manage metadata, trigger workflows, apply policy governance and perform a host of other functions to convert simple bulk storage into a functional business asset. Key solutions within the new platform are image, audio and video intelligence, which add context to uploaded content for improved search and retrieval, as well as the Box Graph tool that continuously learns how people and content interact to enable more predictive, personalized and contextualized experiences.
To be sure, the enterprise is likely to build out its own AI capabilities over time, but this will take some time due to the normal refresh cycles of various hardware and software platforms. The cloud is delivering AI now, and at both scale and price points that allow even small businesses to start crunching data like they were members of the Fortune 100.
As organizations come to depend on digital services not merely as value-adds to existing products but as core revenue-generators themselves, maintaining an advantage over competitors will come down to how well they can utilize the data at their disposal. And since volumes, which are already at record levels, are set to explode once again, only an intelligent, automated and highly orchestrated analytics ecosystem will be able to keep up with the load.
For the enterprise, then, AI in the cloud represents the only viable option at the moment, both in terms of the speed at which intelligent capabilities must be deployed and the scale at which they are expected to operate. And the smarter the cloud becomes, the more appealing it is for the kinds of workloads that are coming to define next-generation data services.