Not All AI Is Really AI: What You Need to KnowA wide range of technology solutions purport to be "driven by AI," or artificial intelligence. But are they really? Not everything labeled AI is truly artificial intelligence. The technology, in reality, has not advanced nearly far enough to actually be "intelligent." "AI is often a sensationalized topic," said Neil Morelli, chief industrial and organizational psychologist for Codility, an assessment platform designed to identify the best tech talent. "That makes it easy to swing from one extreme reaction to another," he said. "On the one hand, fear of AI's misuse, 'uncontrollability,' and 'black box' characteristics. And on the other hand, a gleeful, over-hyped optimism and adoption based on overpromising or misunderstanding AI's capabilities and limitations." Both can lead to negative outcomes, he said. Much of the confusion that exists over what AI is, or isn't, is driven by the overly broad use of the term, fueled to a large degree by popular entertainment, the media and misinformation. [Want to learn more about the future of work? Join us at the SHRM Annual Conference & Expo 2021, taking place Sept. 9-12 in Las Vegas and virtually.]
What Is AI, Really?"Much of what is labeled as 'artificial intelligence' today is not," said Peter Scott, the founding director of Next Wave Institute, a technology training and coaching firm. "This mislabeling is so common we call it 'AI-washing.' " The boundaries have often shifted when it comes to AI, he said. "AI has been described as 'what we can't do yet,' because as soon as we learn how to do it, we stop calling it AI." The ultimate goal of AI, Scott said, "is to create a machine that thinks like a human, and many people feel that anything short of that doesn't deserve the name." That's one extreme. On the other hand, most of those in the field "will say that if it uses machine learning, especially if it uses deep learning, then it is AI," he said. Officially, "AI is a superset of machine learning, which leaves enough wiggle room for legions of advertisers to ply their trade, because the difference between the two is not well-defined." Jeff Kiske, director of engineering, machine learning at Ripcord, agrees. Most of what is called AI today could better be referred to as "machine learning," he said. This, he added, is how he prefers to refer to "cutting-edge, data-driven technology." The term machine learning, noted Kiske, "implies that the computer has learned to model a phenomenon based on data. When companies tout their products as 'driven by machine learning,' I would expect a significantly higher level of sophistication." Joshua A. Gerlick, a Fowler Fellow at Case Western Reserve University in Cleveland, said that AI "is an incredibly broad field of study that encompasses many technologies." At the risk of oversimplification, he said, "a common theme that differentiates a 'true' from a 'misleading' AI system is whether it learns from patterns and features in the data that it is analyzing." This is the promise of many use-cases in HR for machine learning that actually don't rise to the level of true artificial intelligence. Implications for HR For example, Gerlick said: "Imagine a human resources department acquiring software that is 'powered by AI' to match newly hired employees with an experienced mentor within the organization. The software is programmed to find common keywords in both the profiles of the mentees and potential mentors, and a selection is obtained based upon the highest mutual match." While an algorithm is certainly facilitating the matching process within the software, Gerlick said, "it is absolutely not an AI-powered algorithm. This algorithm is simply replicating a process that any human could accomplish, and although it is fast, it does not make the matchmaking process more effective." A truly AI-powered software platform, he said, would require some initial data—like profiles of previous mentee-mentor pairs and whether the outcomes were successful. It would then learn the factors that led to a successful pairing. "In fact, the software would be so sensitive that it might only be applicable to identifying successful mentee-mentor pairs at this one specific organization," Gerlick said. "In a roundabout way, it has 'learned' how to understand the unique culture of the organization and the archetypes of individuals who work within it. A human resources executive should find that the AI-powered software platform improves its effectiveness over time—and hopefully exceeds the success of its human counterparts, leaving them the time to undertake more complex initiatives." Christen da Costa, founder of Gadgetreview.com, said he thinks the term "AI" is thrown around far too readily. "Most automation tools, for example, are not what I would call AI," he noted. "They take in information fed to them by the user and look for cases that match it. Over time they learn the user's preferences and become better, but that's algorithmic learning. While it can be an aspect of AI, it does not an AI make." Does it matter? It can. When HR professionals are considering adopting new technology, it's important to not be confused—or swayed—by lofty tech terms that tend to be thrown around far too frequently. It's also important to not be overly enamored of, or potentially misled by, the lure of "artificial intelligence." "Thoughtful readers and observers of AI in HR would be wise to remember that AI systems help perform manual, repetitious and laborious tasks in HR," Codility's Morelli said. "However, the range and scope of these tasks are probably narrower than some vendors and providers lead people to believe." There is no AI system that understands, perceives, learns, pattern-matches or adapts on its own, he said. "Instead, it needs human-labeled and curated data as a starting point. For this reason, users and evaluators should apply more scrutiny to the training data used to teach AI systems," he said, "especially the data's origin, development and characteristics." "When skeptical over whether a technology is truly 'powered by AI,' consider asking a few simple questions," Gerlick suggested: Is this technology using data in a way to improve its predictive capabilities? Is this technology improving upon the effectiveness of what a human can already do? If the answers to those questions are yes, he said, "then artificial intelligence might be lending a helping hand." Lin Grensing-Pophal is a freelance writer in Chippewa Falls, Wis.
Tags: ai services