If you want your company to be perceived as cutting-edge, throw the word “AI” in the tagline.
The popularity of this term has distracted people from its original meaning. Just because a software solves complex problems in a way that seems inscrutable, doesn’t mean it’s “intelligent”. In fact, its intelligence can often be attributed the old fashioned way – to the intelligence of its designers, usually the software developers. Just like a mechanical clock’s design can be attributed to its intelligent hardware developer (“clockmaker”).
The inner workings of AI aren’t designed by humans. They’re created by algorithms that “learn” from the data the humans provided to them. Humans can influence how the AI works by either:
(A) picking different learning approaches, or
(B) providing different data to learn from.
They can’t, however, directly influence what the AI learns. And, afterwards, they can’t understand in precise detail how it works. This is exactly why AI is a suitable solution for problems we don't know how to solve directly. But we do this at the cost of loss of control.
Let’s take a hiring system as an example. A human can create a rather sophisticated model that makes a calculation based on dozens of factors including, years of work experience, relevance of previous industries, and level of education completed. However, an artificial intelligence model can take into account hundreds of criteria not obvious to the human – perhaps looking at response rate and sentiment in emails, or the success of previous businesses. On the one hand, an AI model could mean less risk for the employer, and more opportunity to people whose resumes would otherwise get overlooked, reducing reductive and unjust decisions. On the other hand, the AI model is a black box – it’s incredibly difficult to tell when a model is, for example, making an unethical choice.
For this very reason, we should be careful when deciding whether an AI solution is the right approach in the first place, for any given problem. AI, in its truest definition, should not be the goal of any business trying to achieve data literacy. It’s just another tool in the data scientists’ toolbelt.