On February 9, 2017, two technology market leaders made announcements: SAP unveiled its next-generation intelligent ERP system, and Nvidia announced that demand for artificial intelligence (AI) applications was driving demand for its graphics platform. On the face of it, these announcements were business as usual – routine sound bites that proliferate in the tech news landscape. Look a bit deeper, though, and you realize that this day marked a profound shift in both the way businesses use technology and the implications for the rest of us.
For decades, developing a computer that could think has been the Holy Grail of technology. And while we have made tremendous progress in our ability to process vast amount of data, the “thinking” part has remained mostly elusive. We are still stuck between two diametrically opposed visions of AI – on one side, it’s the smart but deeply dystopian world of HAL (“2001: A Space Odyssey”) and on the other, it’s a simpler world of devices like Alexa or Siri playing songs or ordering items for us.
Until now, that is.
Enterprise AI is finally coming of age, and cognitive computing is finding its way into real products that companies can use to improve business processes. Simply speaking, cognitive computing refers to self-learning systems that mimic the way the human brain works. This evolution has been a long time coming, but once we had the compute capacity (thanks to the emergence of the cloud) to handle large volumes of data cost-effectively, it was only a matter of time before self-learning algorithms would begin to mature. These enabled data-mining techniques, predictive analytics, and pattern recognition to power automated systems that learn from a new and changing data environment and solve problems without human intervention. Add to those systems the ability to express commands in normal spoken language, and the interaction between a machine and a human becomes fairly seamless. Say hello to HAL 2.0.
In the SAP announcement, the “intelligent” piece of “intelligent ERP” comes from a digital assistant (SAP CoPilot) of which users can ask questions and to which they can give commands via voice, text, or gestures, just as they would to a human assistant. That informal and unstructured conversation is then contextualized, analyzed, and used to present the user with the specific business outcome they seek. Not only that, but by deeply understanding users and organizations in the context of their business, SAP CoPilot can adjust quickly and support users by calling their attention to things they might have otherwise missed.
At SAP Startup Focus, our job is to build and scale SAP’s innovation ecosystem, driven by the power of startups. As part of that effort, we work with more than 5,200 startups from more than 55 countries that are building the next generation of solutions to drive businesses forward, including a handful that are specifically working on enterprise AI. While AI is currently in a nascent stage, there is little doubt that it is going to play a critical role in applications as diverse as predicting heart attacks, managing invasive species, or preventing financial fraud.
One of the startups that is part of our program is a company called Feedzai that uses the power of machine learning, a form of AI, to help payment providers, banks, and retailers prevent fraud in omni-channel commerce. According to the 2016 AFP Payments Fraud and Control Survey, 73 percent of companies were victims of attempted or actual payment fraud, compared to 60 percent two years ago. For companies with a revenue of $1 billion or more, the number that were victims stands at a staggering 78 percent. Newer business models are constantly evolving, from instant delivery of goods to virtual cash to digital downloads, and as digital payments become ubiquitous, so does the risk of online fraud. According to LexisNexis Fraud Multiplier, in 2015, every $100 of fraud cost a merchant $223 in actual costs.
Until recently, financial institutions relied on a more rigid rules-driven model e.g. if a credit card was physically scanned a few minutes ago in Los Angeles and then two hours later it surfaced in Mexico, that would be a possible trigger for fraud. Relying on this model doesn’t work anymore because fraudsters change their game faster than anyone can update the rule book (In the credit card example, it could just be a false positive because the customer could have simply traveled across the border in that time). Fundamentally, the pathology of fraud mimics the behavior of good customers – on the surface, there is hardly any difference between your best customer and your worst nightmare. This is where machine learning can help, by quickly learning, adapting, and providing real-time results that can stop fraud before it happens, while minimizing false positives.
Galaxy.ai is another up-and-coming player, this one in the healthcare space, developing software that uses the power of AI to create a disease prognosis before a patient shows any symptoms. Early detection naturally saves on healthcare system costs, and improves patient outcomes significantly. Galaxy’s core approach is in inputting vast amounts of data into its system, then training it to look for patterns, in much the same way that the brain of a young medical student or intern might work. The company still has some way to go before it can develop algorithms that precisely mimic the neurons of a human brain: As Galaxy’s website admits, it has “built flying machines that do not actually flap their wings but mimic the general principles of flying. We might do the same on our path to build artificial intelligence brains.”
No technology advancement is complete without it having a certain dark side. From the days when the textile power loom displaced the weaver or when the automobile displaced the horse and buggy, we humans have generally muddled through technology’s impact on our jobs. As automation eliminated jobs, we tried to move up the food chain and find a function that would be less affected by automation or outsourcing. But when the scale and impact of AI can put at risk the jobs of even the most highly qualified knowledge workers (in the Galaxy.ai case above, we won’t eliminate the need for radiologists but will certainly need a lot fewer of them), that is where the calculus changes completely.
According to a recent survey by the World Economic Forum, by the year 2020, robotics, automation, and AI will lead to a net loss of 7.1 million jobs in the top 15 major industrialized and emerging economies. The irony is that a majority of the jobs lost will be at the lower end of the skills (and income) spectrum where people can least afford to be rendered redundant. The survey’s somewhat better news is that over the same time frame two million new, high-paying jobs will be created in the areas of computing, math, and engineering.
Maybe now is the time for us to actively consider STEM, not just for our kids but for ourselves as well.
Join the conversation @SAPStartups and/or follow me @BansalManju.