All AI is machine learning, but not all machine learning is AI. Make sure you know the difference before investing.

The race toward real artificial intelligence (AI) is officially underway, and much like the 1960s race to the moon, the desire to finish first has led to a plethora of misleading information. One unfortunate trend that has developed is AI washing. AI washing describes how vendors falsely label technology as “artificial intelligence” when it is machine learning at best, or in the worst case, the same software algorithms they have always used, just rebranded. This has led many organization leaders to become skeptics of all technologies labeled as AI. However, there are effective strategies to recognize true AI compared to machine learning (ML) capabilities, as well as some key questions to ask that can help adequately vet providers.

Aspects of true AI

Many company leaders don’t know the difference between ML and true AI, which is a root cause of industry skepticism. Consider this idea first: All AI is machine learning, but not all machine learning is AI. Most forms of machine learning are too simple to be considered true AI. Standard machine learning utilizes a series of algorithms developed using an input data set which then produces known outputs. This type of technology then picks up underlying patterns overtime and eventually makes simple predictions about new input date. Therefore, every action, decision and production are defined. Now, take this idea and flip it upside down. True artificial intelligence is made up of the same machine learning algorithms, but none of the outcomes are defined and there is an unlimited number of possible outcomes that can be produced.

What does this type of scenario look like? Machine learning data sets will be developed and used as training data to teach the more advanced system the initial commands, but once the surrounding environment begins to change, true AI will use capabilities similar to our intelligence to make its own decisions. True AI runs on an incredibly complex system that can produce unparalleled value for any organization. Currently, most AI implementations are described as “Narrow AI,” which performs a specific task more accurately and efficiently than any human could.

The right way to vet a provider for AI

The second primary source of skepticism stems from the fact that executives, IT teams and consumers have limited trust in providers and their willingness to be honest about their product’s complete capabilities. The term AI washing is rooted in this assumption since it’s easy for providers to claim their software has AI capabilities, but it is difficult to execute these functionalities. According to Gartner, there are over a thousand software vendors that describe themselves as artificial intelligence vendors or claim their products utilize AI. However, most of these businesses do not deploy true AI. Even consumer products are facing this AI craze, with one example coming this year at CES. Oral-B unveiled their Genius X toothbrush that claimed to function with AI capabilities — but, once experts dove deeper into the technology, it was discovered that the toothbrush did nothing more than provide simple feedback to the user describing if they are brushing their teeth for the proper length or consistency. While this is a simple example of a consumer product touting AI, the reality of AI washing is all too real with business-to business technologies as well.

To properly vet a provider, remember that true AI will get smarter over time. Come up with leading questions around this fact. For example, ask the vendor to explain exactly how their software gets smarter over time and what decisions it can make. Also, be sure to get information on how much human interaction is needed and if the software can complement the already existing human workforce. There’s a good chance the vendor is committing AI washing if they can’t fully answer these questions.