From Succeeding with AI by Veljko Krunic
This article discusses how to get practical business results from AI and how this book will help you learn how to do it.
Guiding AI to business results
I don’t promise results from buying frameworks that solve all your problems, nor am I believer in catalogues of case studies showing you how someone else made money using AI. If you’re hoping that there are some parts of AI algorithms that you poorly understand but which will make you a ton of money, you’ll be disappointed at the end of the road. I call this approach the “rainbow and unicorns approach to AI.” Instead of having a clear idea of where we want to take our business, we too often abdicate responsibility to AI and wish that some mystic part of AI would solve our problems. We replace initiative and understanding of the world around us with the hope that data mining with better AI algorithms will show us the way forward.
This hope is misguided—you can’t get business results with AI by relinquishing control to algorithms, metrics, and AI frameworks. AI isn’t a silver bullet whose application always makes you money. On the contrary, it’s much easier to use AI to lose money than to use AI to make it. Thinking about AI as a black box that can make correct business decisions on its own is a sure way for executives and their budgets to be parted.
NOTE AI can’t exercise good judgment, and most common AI methods used today (such as deep learning) can’t determine causal relations. AI methods are quantitative methods which require the right metrics to drive them. All AI algorithms know how to do is maximize a specified metric, without knowing why they’re maximizing it. The context and the purpose for which AI should maximize that metric have to come from a human.
Humans alone have the capacity and skillsets to define the metrics that link business and technology. Humans must supplement AI in the areas where AI is weak. You make AI applicable to your business problem by engineering the proper link between business and technical metrics. Only humans can design metrics that describe a lifetime of human wisdom in a way that quantitative algorithms can understand.
On the technical side, AI solutions developed in isolation from human judgment don’t work today. On a more philosophical level, there’s another reason to not blindly follow AI—we, as humans, possess agency. AI isn’t an unstoppable force which is affecting us all, and we’re not along for the ride. The AI revolution isn’t happening. It’s us, people like you and me, who are creating that revolution. We must exercise our agency. We must decide what the results of applying AI will be. On the societal level, we can guide AI toward the goals we desire, instead of letting the AI revolution careen toward destinations we fear.
We don’t know when (or if) AI will reach Artificial General Intelligence (AGI) level, in which AI matches (or exceeds) human capabilities. Until AGI arrives, we human beings need to guide business and society and to understand that AI is a tool we need to wield without being overawed by it. Companies like Google have demonstrated that you don’t need AGI to build a great business; you need a smart person who knows how to link business and technology. AI doesn’t make you money—people make you money!
That’s all for this article.
As you learned in section 2.7, there are no unicorns among data scientists and data engineers. As for the rainbow part of the “rainbow and unicorns”—there also is no pot of gold to be found if you walk to the end of the AI rainbow.