You can’t barely escape any mention of “Artificial Intelligence (AI)” or “machine learning” in business articles, LinkedIn posts, or any offerings from firms, big or small. It’s like a magic word that suddenly turns mundane analysis into something that sounds impressive.
Now, I’m not against AI nor machine-learning, but I just want to inoculate you against any misconception and those who want to impress you with smokes and mirrors to mask something that is inherently straightforward (although still impressive in its own right!).
In simple terms …
Ever played “I spy with my little eye” when you were young or with your little ones in your last road trip? You would try to narrow down the options by asking questions to finally arrive at the object to be identified.
Imagine a method of interrogation with quick-fire questions to quickly identify an object, a specific condition, or a desired activity – all in a matter of seconds. The more the program is trained for the identification, the better. Just like your iPhone’s Face ID. Essentially, your iOS would try to match several crucial points of your facial features with a growing data stored on your phone. Whenever you change your hair-do, get new glasses, develop new wrinkles, or face the screen with different expressions, your iPhone would try to match your face with the database. It would calculate the probability that it is you and not a stranger trying to unlock your phone. If it can’t make any definitive conclusion, you will need to provide a positive confirmation by keying in your passcode. Any new information about your face will then be added so the machine learns more about your facial features.
This technique is used either to identify stars and nebulae or to check whether your credit card has been compromised. It’s used by your robot vacuum cleaner to learn about the boundary of your living room. It’s also applied in marketing or business circle to try to predict an ‘optimal’ way to structure your product portfolio, product distribution, budget allocation, media scheduling, or advertising creative. A specific algorithm will scrutinise the data to propose tactics that deliver successful results and those that don’t.
Essentially, this what AI or machine learning is.
I was introduced to AI through the old Byte magazines in the late 1980s. The concept seemed intriguing and fascinating to my young geeky mind. When I did my undergraduate degree in Computer and Information Science in the 1990s, the University of South Australia offered an “Artificial Intelligence” course as an elective in my third year, which I naturally picked. The course took away the mystery and some of the allure but it left me with a realistic appreciation of what it can and cannot do well. A lot has changed in 30 years but undoubtedly, the principles remain.
It’s not an alternative to Real Intelligence
One key limitation of AI or machine learning is that its boundary or scope. An algorithm used to predict product launch success in snacking may fail miserably if you use it for garden fertiliser data – unless the algorithm has been trained to cater for a wide range of categories. It will need more information to be used to improve its learning capability to cover new contexts.
Which brings us to the real limitation of AI. Essentially, machine learning treats every information as cold, hard facts. It treats data without context or qualitative insights that we, humans, are capable of. An AI system that could identify different types of cows with great accuracy may identify Indy, my Staffordshire Bull Terrier, as a malnourished Holstein Friesian cow.
This is the key area that business managers need to be aware of when receiving offers of algorithm-based solutions. Context and insights still need to be human-led. Otherwise, things that make logical sense to the system may be unworkable or just outright wrong. AI still needs to be driven by people with real insights — dare I say, real intelligence.
AI is a means to an end more efficiently. It still needs human brains to refine the results: real intelligence that can apply analogies and experience from unrelated fields, pivot to a different direction, and infuse external context and soft knowledge. Things that do not make sense to a computer algorithm. If a firm tries to bedazzle you with complex terms, and the only explanation you receive is “This is the best strategy for you, based on our algorithm” – you may want to look elsewhere.
A firm that would focus on AI as the main selling point is akin to a restaurant that promotes its state-of-the-art kitchen and its automated appliances. Think about it: when you go to a three-Michelin-star restaurant, you don’t pay a handsome amount of money to get perfectly julienned carrots. You are there to enjoy the mastery of the head chef and their protégés in crafting their dishes. Such chefs are able to work with the simplest tools or in the most sophisticated kitchens.
Ultimately, there may be an AI, machine learning (or deep learning) system that can gather enough information and data points across conditions and topics to give you more holistic recommendation and results – closer to what a human may recommend. In most cases, it still won’t replace a well considered analysis done by real experts.
There’s definitely a place for AI and machine learning. It helps us to accelerate exploration and analysis and be more efficient in our analytical work. However, if your agency relies too much on the artificial for business insights, perhaps work with those that can give you real intelligence.