SAS Steers Toward Stronger AI Data Lifecyle Via Viya

AI needs history. To build the new artificial intelligence functions that we seek to put inside our modern enterprise and consumer applications, AI needs a base of historical data upon which to operate. While the Silicon Valley start-ups gain a lot of attention (and will inevitably continue to do so), it’s worth remembering the “family” of (usually competing) firms that have been wrangling with data to create analysis intelligence way back when it was just called data mining.

Now boasting almost fifty years of operational health (the company was founded in 1976 by Dr James Goodnight and he is still CEO), SAS Institute is today more commonly known as SAS.

Quietly reinventing its core technology proposition through the various ages of client-server, early networks and disaggregated computing, the first era of the web and onward into cloud computing and today’s fascination with generative AI (good old fashioned predictive and descriptive AI does still exist), SAS has moved with the times and continues to point toward new weather patterns on the data landscape. Nowhere is this truth born out more clearly than its work to create SAS Viya, the company’s flagship data and AI platform.

Among the more important developments seen this year has been the arrival of the SAS Viya Workbench developer environment for building AI models. Described as a “lightweight developer canvas”, this is a self-service on-demand compute environment for conducting data preparation, exploratory data analysis and developing analytical and machine learning models.

What Is Lightweight Technology?

Technology vendors use the term lightweight to denote software tools that are built with fewer lines of code, a smaller memory footprint, tools that come with prebuilt scalable infrastructures to get started more quickly, or ones that are more nimble in terms of form, function and their ability to work inside tight spaces (edge computing across the Internet of Things for example) and SAS Viya Workbench checks all those boxes. Viya Workbench allows developers and modelers to work in the language of their choice, initially SAS and Python, with R available by the end of 2024.

A casual glance across the technology trade newswires will often bring up the term “developer happiness” and workflow productivity. SAS appears to have put the generation-Z factor into generative AI and thought about these factors (i.e. SAS Viya Workbench on-demand as already noted), but it’s also self-provisioning and self-terminating, so therefore requires minimal IT support. The dedicated analytical environment features customizable CPU/GPU compute power to match the needs of any given project. Models and other results can then be used in SAS Viya for data management, governance and operational deployment.

If that’s not happiness, then it’s somewhere close to quite convenient and rather nice, so make sure you stop and have a decent lunch and eat some greens.

Cutting Edge Cloud Compute

SAS insists that the business case for Viya Workbench is strong i.e. AI developers and modelers want to work with modern, open source packages and cutting-edge cloud computing functions, but they are also under pressure to deliver fast results and manage costs.

“The many challenges developers face aren’t just minor annoyances – they are obstacles that prevent questions from being answered and work from getting done,” said Jared Peterson, senior vice president of platform engineering, SAS. “Viya Workbench provides maximum flexibility and results by allowing developers to use their language and integrated development environment of choice, tailor compute power up or down to meet the needs of the project, and ultimately boost their productivity and efficiency. They can work faster, be more creative and take more risks – which, let’s be honest, is not only what’s expected but it makes the job more fun.”

Additional AI developer happiness offerings currently in development from SAS include Viya Copilot, a technology for software engineers and data scientists designed to shoulder the administrative analytics tasks involved with solving industry-specific business problems. The tasks here would typically include code compilation, essential code commenting processes (i.e. annotations and documentation) and other actions designed to create streamlined code interpretation. With data scientists typically (we might even say “always”) spending more time on data wrangling, data ingestion and crucial data preparation tasks than they would like to, this service may prove popular.

The SAS Viya total toolset is large and we need to stop somewhere, but let’s make sure we also note SAS Data Maker, a technology which has been described as a trusted synthetic data generation tool. SAS Data Maker is built with a suite of synthetic data algorithms that can produce synthetic data for situations where original data is not available (perhaps because of personally identifiable information concerns, perhaps due to governance or cross-border restrictions between states or countries), but where analytical processing using carefully created synthetic data is still eminently useful. Synthetic data is also much-needed for AI analysis related to extremely uncommon events where we still want to perform modeling, such as medical environment analysis of patients with a rare condition, tumor or infection.

Keen to validate the use of its tools in the context of this discussion, SAS enlisted technology analyst house Futurum Group to evaluate the latest AI developer and data science tools. An analysis document has emerged this September 2024 which suggests that the SAS Viya data platform helps users execute the lifecycle of collecting data, building models and deploying decisions 4.6 times faster than “selected competitors” (see below) to increase innovation, speed up decision-making and drive revenue growth.

The Futurum Group analysis says Viya was compared to a leading commercial environment and non-commercial open source environments including Jupyter Notebook with MLFlow and Python Libraries.

How Do We Measure Productivity?

The U.S. Bureau of Labor Statistics defines productivity as a comparison between input and output. In the case of Viya versus selected competitors, input factors measured encompassed data, labor, tools & technology, infrastructure and time. In his scenario, output includes business decisions, insights data products and innovation. But why is productivity important and how can a data and AI platform like Viya make a difference?

According to Forrester, as AI teams are comprised of data scientists, AI experts, business and technology leaders, organizations need AI/ML platform tooling that maximizes productivity for each role while enabling friction-free collaboration.

“Testing showed that an end-to-end data and AI lifecycle can be achieved with more than 4x greater productivity in SAS Viya than in competitive solutions,” said Russ Fellows, VP and analyst at The Futurum Group. “The ability to quickly begin working, together with SAS Viya’s productivity enables AI teams to rapidly produce business results and insights from their data.”

Futurum studied Viya’s speed in 2023; the productivity study is a continuation of their research. According to Futurum, Viya excels for a variety of technical and non-technical users. For example, data engineers are 16 times more productive accessing, preparing and governing data with Viya; data scientists are 3.5 times more productive building, optimizing and validating models; MLOps engineers are 4.5 times more productive automating, monitoring and retraining models; and business analysts and other non-technical staff can complete 86% of data lifecycle tasks using Viya, compared to 56% in the commercial environment and 47% in the non-commercial environment.

Our Next AI Goals

All of which really starts to beg the question – what do we want next from our AI tools, applications and data services?

It’s a tough question to answer, but look at the facts i.e. the chatbots and user copilot technologies will continue to gain plenty of attention for most of the rest of this decade. The embedded AI on your smartwatch that you can interact with via Natural Language Processing enables you to ask what the weather is like, set an alarm or even turn a light on. That won’t be old hat for a while, but it will become an assumed function sooner or later.

Nobody gets excited about Uber cars arriving, automated Google Chrome updates, Zoom calls or Bluetooth connectivity anymore, right?

Comments from Jay Upchurch, executive vice president and CIO at SAS may point the way, “AI runs faster and more efficiently on Viya, so teams learn and see results faster. Viya provides agility and resiliency that empowers an organization to see opportunities before competitors do whether those opportunities involve approving customers for car loans, keeping trains safe or distributing merchandise from a retail distribution center.”

Perhaps it’s these big picture industry-specific AI enhancements that impact a country’s core infrastructure and entire economy that we hanker for most (remember how annoyed we were when supply chains broke during Covid-19?) as they start to manifest themselves. For that level of AI to really engage and penetrate, we’d need technology vendors with solid experience and maturity. A hero may rise.