Whether we like it or not, software algorithms in the form of artificial intelligence (AI) programs are affecting more and more aspects of our lives. They determine what news and entertainment we are offered to look at, what goods we’re recommended to purchase, even what people we might like as friends (on Facebook). Businesses increasingly use them to determine who to hire, what to sell, and what kind of marketing and advertising outreaches they will do.
Much of this is in pursuit of customer “personalization” and, of course, profitability, on the part of the businesses that want to sell us things.
But what are the deeper ramifications? Research suggests that the technology is not mature enough—yet—to take on the ambitious “moon shots” we’re seeing in, say, being able to diagnose and recommend treatment plans for cancer, or create truly safe self-driving cars. While celebrating the advancements in technology that can help us—both in our business and personal lives—we need to be cautious about judiciously applying the technology at the right place at the right time.
AI is (mostly) here today
The hype around AI has been deafening as of late. Experts made enthusiastic predictions about AI’s growth for both business users and consumers. And although only 17% of developers worked with the technology in 2017, more than three-fourths planned on using AI or machine learning in 2018.
A Gartner report predicted that AI would create 2.3 million jobs by 2020. And Deloitte performed a survey that found three-fourths of executives believe that AI would transform their companies within three short years.
Yet when Deloitte looked closely at the actual AI projects that were likely to succeed, it seemed that most of them were “low-hanging fruit” and simply making basic business processes more efficient.
- RPA-based AI is by far the most common type of AI (47%) and includes extracting data from emails to update addresses in customer files or automating invoice processing. These types of initiatives are becoming more and more common.
- Cognitive insight automation (38%) is used to discern patterns in big data. Also called machine learning, Deloitte says this is the type of AI used in “recommendation engines” used by the like of Amazon and Netflix, or the AI technology that lies at the heart of identifying credit fraud in real time.
- Cognitive engagement AI initiatives are the least common (just 16% of AI projects) that use machine vision, intelligent agents, and machine learning to actually make judgments and decisions, such as answering customer service questions intelligently to diagnosing medical conditions, to autonomous cars. These types of projects tend to not currently be ready for prime time.
AI Algorithms in our lives today
You might not realize it, but AI is already surrounding you. Here are just a few examples of AI in your everyday business and personal lives.
Keeping your email box spam-free. You probably take for granted that you’re not overwhelmed by spam. But it’s no accident. Consider that spam makes up almost 50% of all email sent in a given month, and you might wonder why you don’t see more of it. That’s because of AI algorithms. Simple rules-based filters—for example, telling your email box to delete messages that contain “Nigerian prince,”—simply don’t cut it, because spammers simply avoid phrases that get tagged. Spam filters today must be intelligent, and learn to keep one step ahead of the spammers. Through its machine-learning technology, for example, Google manages to keep 99.9% of spam out of your inbox.
Minimizing risk for banks and financial services firms. Consumers’ FICO scores, which most banks use to make decisions on whether to extend credit, and to determine the risk of individuals, are based upon machine-learning algorithms. According to a study done by MIT, this type of AI reduces banks’ losses on poor customers by as much as 25%.
Recommending what you should buy next. Amazon keeps AI neural network algorithms a deep, dark secret, but it helps to make sure that when you search on such generic terms as “shampoo” or “vacuum cleaner” that you get something very specific to your needs that Amazon has learned over time. Why do this? Simply: It pays off for the online retail giant—by some estimates, as much as a 30% rise in sales.
AI’s limitations—and some amusing failures
When we think about AI we have to keep in mind that this is still a relatively new technology—one that is still under development. That said, here is some evidence that the audacity of algorithms has yet to be aligned with what they can do without errors.
Apple Face ID fooled by a cheap mask. Just one week after Apple released the iPhone X, a Vietnamese security firm was able to unlock the phone’s much-touted AI-based facial recognition screen unlocking system. Using a mask printed with a 3D printer that cost about $150 to make, the firm convinced an iPhone X that it was human, and unlocked the phone.
Alexa threw a party and only the police showed up. Neighbors in a quiet German neighborhood complained to the police last November of a loud party in a neighboring apartment late at night. The culprit was an Amazon Echo, which had spontaneously started blasting music while its owner was out. The poor man was not only stuck with irate neighbors, but a large locksmith bill, after the police shut Alexa down and changed the locks.
AI makes online shopping too easy. In January 2018, San Diego news channel CW6 reported that a six-year-old girl asked Alexa for a $170 dollhouse—and Alexa promptly ordered one for her. Then, when the TV anchor inadvertently repeated the girl’s words (“I love the little girl saying, ‘Alexa order me a dollhouse,’”) some viewers found their Alexas also ordering the same toy model.
AI will change our lives
Wired magazine co-founder Kevin Kelly has predicted that AI is poised to become the new infrastructure powering a fourth industrial revolution: the data revolution. This is almost certainly true—we’re already seeing the effects—and be prepared to be amazed as we continue on this trajectory.