The real-world business of AI
For decades, humans have had an uncertain relationship with artificial intelligence (AI). Computers that can think and learn have been seen both as powerful tools and as existential threats. Those fantasies and fears have long appeared in science fiction, but today, AI is making its mark in the real world.
And the reality is very different. While AI outpaces human performance in complex analytical tasks, humans are still very much running the show. Rather than an epic battle of man versus machine, it’s transformed into a match made in enterprise heaven—a symbiotic relationship with extraordinary potential to improve the way business is done across every industry.
As consumers, we hear “artificial intelligence” and typically think of in-home digital assistants that deliver information and complete tasks within seconds. But these consumer-facing devices pale in comparison to what’s happening in the business world, where AI is dramatically transforming business operations and even saving lives.
Across industries, AI-derived insights are augmenting ordinary human capabilities. Well-trained AI systems help make sense of it all, crunching the numbers from new data streams in real time.
AI helps bank managers stay one step ahead of cybercriminals, who are getting faster and smarter. Servers built for deep learning, machine learning and AI allow banks to track spending patterns and spot financial crimes in real time. As a result, bank managers can quickly implement fraud countermeasures to save countless dollars and reduce customer stress. Through AI contextual analysis, the credit card industry has reduced the rate of false card declines by 80%.1
AI assists physicians in more accurately diagnosing diseases such as melanoma by reviewing vast numbers of image scans to better detect edge cases and the anomalies. While physicians and health care workers can experience fatigue within 20 minutes of looking at images, a computer never gets bored or tired. The physician can instead focus on analyzing the results, interacting with patients and formulating a treatment strategy.
AI can help protect inspectors and improve their productivity, while helping identify patterns that predict failure. For example, telecom companies use drones to inspect high-voltage power lines. AI on servers and a high-speed storage platform work together to process images from the drones on-site. Inspectors spend more time on incident analysis. And telecom companies reduce the cost of downtime and safety risks of inspections.
These business applications for AI all rely on one thing: getting the right data set and knowing how to use it as a competitive differentiator. The key is a solid partnership between a company’s CIO and its chief data scientist. Together, they identify the business’s data requirements, harness their data sources across silos and design IT systems that can organize, store and process the data.
“We see a race developing within industries in which people want to have better understanding of consumer sentiment; better forecasting of demand; and do better-targeted advertising,” says Hillery Hunter, Director, Accelerated, Cognitive, and Infrastructure, IBM Research. “So the first step of AI is to bring your data together.”
AI is here, but can your
data keep up?
AI is here, but can your
data keep up?
Some businesses are already far along the AI adoption curve. Their performance illustrates what the world is like when day-to-day work is bolstered by powerful computer intelligence.
EverString cofounder and CEO J.J. Kardwell is one of the leading advocates for today’s AI revolution. But he’s also aware of a problem that some businesses may overlook in their excitement to get started with AI.
“About two years in, we realized that the biggest blocker for machine learning and applied AI’s success was actually the data itself,” he says. “It is one of the most overlooked, misunderstood and undervalued elements.”
Businesses once worked hard to acquire useful information about their customers and operations. Today, they typically have more data than they know what to do with.
Data security and transparency are also crucial, especially if enterprises outsource their AI processing needs. Questions about where data is coming from, what data sources were used to train AI and what data is being generated by AI will remain ongoing concerns.
Kardwell thinks the next step will be developing control systems that monitor the integrity of the output data AIs are generating. That will help ensure businesses have ownership of the datasets they’re processing. Performance may start to drift heavily, and you may not even know it unless humans are continually monitoring,” he explains.
But managing your data doesn’t have to be a daunting task.
AI strategy depends on data-driven
AI strategy depends
Companies are eager to embrace new AI functions, but for them to successfully implement AI, they need to rethink their IT infrastructure.
“Businesses have been faced with what I call an ‘unoptimized building block,’” says Dylan Boday, Director, Offering Management, Cognitive Systems at IBM. “Clients don't have hardware that's optimized, and then they have software that's unoptimized on top of that. They're really doing a lot of uphill battling.”
In a way, AI could be considered the ultimate reward of the big data era. It employs high-powered algorithms to turn ordinary businesses into iterative, adaptive, learning organizations. To harness the full potential of data-driven AI, most companies first need to upgrade the very foundations of their IT infrastructure.
“There's no way that you can be a player and not be touched by AI. So really the question is, how quickly can you adopt these technologies and bring them to bear on your business?”
“Full adoption of AI is complex,” explains Danny Mu, principal analyst for CIOs at Forrester. “A variety of different storage and server configurations can be used for AI, but containers-based, cloud-native architecture should play an important role in dynamic and large-scale AI applications.”
According to an IDC study of AI adopters, more than 45% of small businesses and 35% of large businesses expect their current infrastructure for AI to last no more than one year.2 And 90% of respondents ran into one or more limitations with their cloud-based systems, while 77% of respondents ran into one or more limitations with their current on-premise AI infrastructure.3
Businesses’ biggest limitation, in most cases, is legacy systems that simply can’t keep up with the processing demands required by AI.
“A key step is the investment in IT infrastructure that will help your data scientists and AI creation folks do their jobs as productively as possible,” says Hunter. This requires an integrated approach to hardware, software and data, including investments in all-flash data storage and GPU-accelerated servers which are designed specifically for AI.
The future is man and machine
The future is
man and machine
AI deviating from its initial programming might sound worrisome, but IT leaders know computers won’t be replacing humans in the workplace.
“There will be a very long cycle where the most value will come out of the hybrid mind—the converged intelligence of humans and machines,” Hunter explains. “We will see humans and machines working more closely together, with humans routinely teaching machines, and machines regularly asking humans questions to enable active learning in large-scale ways.”
Forrester’s Danny Mu dismisses sci-fi fear-mongering because it overlooks the real potential of AI, which is to augment human possibility and business potential in nearly every domain. “Don’t be afraid of AI—embrace its potential,” he insists. “In the long run, there is going to be a mindset revolution around all of this.”
That means there’s one question every IT leader should be asking today: Is my business’s IT infrastructure ready to harness the full power of AI?
1. Felt, Cristine. “Can AI Really Help Minimize Credit Card Fraud?” Clouded (blog), Sept. 8, 2017
2. IDC Video, “Hitting the wall with server infrastructure for artificial intelligence?”, September 2017
3. IDC White Paper, sponsored by IBM, “Hitting the Wall?”, September 2017