Artificial Intelligence: It’s Time

The technology is ready for business. Is your business ready?

By Brittany Swanson

January/February 2019

People may continue to call artificial intelligence and machine learning emerging technologies for decades, but the technology is ready to implement today. In order to avoid falling behind, businesses need to start moving on plans for AI and machine learning now.

Oracle Magazine sat down with Ian Swanson, vice president of product management AI and machine learning for Oracle Cloud, to talk about enterprise AI and machine learning today: adoption challenges, ways to succeed, and how Oracle supports innovation.

Oracle Magazine: AI and machine learning, in particular, have been emerging technologies for some time. What is the state of these technologies in the enterprise today?

Swanson: In terms of plans to build out AI and machine learning capabilities, companies today stand at a crossroads. On the one hand, by the end of the next decade, these technologies won’t be optional—McKinsey reports that AI will have an impact of US$13 trillion on the global economy. Nonadopters will see significant decreases in cash flow. Yet, the challenges of supporting AI at scale—such as complex infrastructure and technology requirements and significant competition for limited talent—have, in many cases, caused companies to hesitate or fumble.

The good news is that the rise of solutions with integrated AI capabilities is helping to accelerate the journey to AI maturity.”

I talk to a lot of executives who are sold on the value of AI and machine learning but are unsure of how to scale. That pretty much defines the enterprise AI landscape right now. My former company, DataScience.com, which was acquired by Oracle in the summer of 2018, was focused on mitigating the issues that popped up along the way to enterprise AI: siloed teams, disparate tools, and limited computing power. Rarely did I encounter someone who didn’t see the potential value; often I heard, “How do I do this well?”—especially if they were looking to build and deploy AI from start to finish.

The good news is that the rise of solutions with integrated AI capabilities is helping to accelerate the journey to AI maturity. The goal of enterprise AI and machine learning is not to achieve the kinds of feats you see in sci-fi movies but to automate redundant and time-consuming tasks, create more-personalized customer experiences, and improve operations.

Oracle Magazine: You mentioned infrastructure and technology challenges, among others. What keeps companies from capitalizing on the promise of AI?

Swanson: Let’s say your data science team builds a machine learning model that returns results to customers in under one second. That’s great—until you realize the fire hose of new data feeding the model causes it to decay just three hours after deployment. By the 24-hour mark, your model is useless. This is not hypothetical; this is a real scenario that played out at a multinational oil and gas company.

In order to keep this extremely valuable system up and running, the model in question needed to be retrained constantly—and as close to the time new data was produced as possible. This is the kind of challenge people don’t talk about much when they wax poetic about AI. However, as you get into enterprise-scale projects that involve huge volumes of data, you’re likely to run into a problem that seems, at first, completely insurmountable. You’re certainly not going to have the time to manually retrain models every three hours. But if you stop thinking about ad hoc fixes and start thinking about the system as a whole, there are plenty of solutions available.

Oracle Magazine: How do businesses overcome the talent challenges, the data volumes, the model retraining, and other requirements to succeed with AI and machine learning?

Swanson: What I always tell business leaders is this: Yes, you have to consider your talent needs, but you must also look at your technology and processes. At the beginning of the data science boom, everyone was rushing around trying to hire data scientists, with the mistaken impression that they’d solve all of a company’s data problems. But the outcome was usually a bunch of reports that got lost under all the paperwork on someone’s desk— or, even more disappointing, a machine learning model that would live and die on a data scientist’s laptop because it was never deployed into production.

Oracle Autonomous Database is another great example of how Oracle is infusing its products with machine learning capabilities.”

If you want to build a sustainable system that delivers predictions directly into applications, you’re looking at a problem that demands significant infrastructure capabilities, along with people who can move big data around and a way to push data science projects into a live production environment. You’ll also be contending with a lot of tools. Data scientists are highly loyal to open source tools, so expect to support a wide variety of them.

Oracle Magazine: What is Oracle doing to help customers rev up their AI capabilities?

Swanson: Oracle offers solutions that touch every stage of AI maturity. As I mentioned before, some companies are in a place where they can’t support custom AI. Oracle has brought machine learning–powered applications, intelligent user experience (UX) components, and conversational agents to enterprise resource planning (ERP), customer experience (CX), human capital management (HCM), and supply chain management (SCM). That means, for instance, that Oracle customers today can use the Oracle Customer Experience Cloud suite to generate offer recommendations—rather than building their own recommendation engine from scratch.

Organizations that want to build AI capabilities in-house can take advantage of Oracle product offerings in the cloud that allow them to build, train, deploy, and manage their own machine learning models. Along with the AI product offerings, Oracle Cloud Infrastructure adds significant value. AI and machine learning projects regularly suffer from issues related to a lack of computational power; deep learning models, for example, often require GPU instances. Oracle offers the latest GPUs and also resolves other issues such as latency problems and high costs for heavy workloads by providing the best price per performance, low-latency data centers, and clustered networking.

And, of course, Oracle Autonomous Database is another great example of how Oracle is infusing its products with machine learning capabilities: The database automatically upgrades, patches, and tunes itself, saving time and reducing costs for users, with the help of machine learning.

Oracle Magazine: How does Oracle support the future of AI?

Swanson: Oracle has a product stack that can support enterprise AI—all stages of it—today. That’s a big deal. At the end of 2018, most research and advisory firms were reporting that companies now know they need to embrace AI—evolve or die—but they haven’t laid all the groundwork, especially in terms of infrastructure and process. Oracle’s portfolio is poised to accelerate their journey to AI maturity on a massive scale.

Next Steps

LEARN more about AI and machine learning.

EXPLORE Oracle AI products.

Photography by Raffi Alexander