In the excellent Prediction Machines Agrawal, Gans and Goldfarb take an economics-based view of the changes that will be brought on by Artificial Intelligence (AI). Essentially some basic economic principles can help us see what will change. The argument is that AI can predict and this will become cheaper with increasingly effective machines. This will have knock on effects: “The drop in the cost of prediction will impact the value of other things, increasing the value of complements (data, judgment and action) and diminishing the value of substitutes (human prediction), (Agrawal, Gans, and Goldfarb, 2018, page 19-20). Those things that allow you to get the best out of the cheap AI, such as the data AI needs to generate a decent prediction, will become more valuable. On the other hand some things will become next to useless. Who needs a person to predict what will happen when an AI can do it much better by faithfully mapping to patterns seen previously without the mistakes made by human beings when we try and concentrate for too long.
The example of “anticipatory shipping” is an interesting one. At some point it’ll generate more profit for Amazon to ship what it thinks you want and need (and accept that this will generate greater return costs) than wait for you to ask for an item. I think this is great but I can imagine some people might find it a little creepy.
The authors suggest that AI is really about prediction. This is where it benefits business. “At the center of the AI canvas is prediction. You need to identify the core prediction at the heart of the task” (Agrawal, Gans, and Goldfarb, 2018, page 140). Doing so forces you to clarify what you mean. A business school needs to specify (to do this the school must decide in advance) what they mean by “the best student” in order to benefit from using a prediction machine in recruiting. Do you want students who are great in class? Donate more after graduation? Get high-profile jobs that bring kudos to the school? The machine needs to know what you want in order to know what to recommend.
But humans will still be needed. Spreadsheets made those who could use them more efficient and so increased the value of such people (rather than stealing their jobs). Some jobs will go away, others will become more value depending upon whether they can be replaced by, or are enhanced by, the availability of cheap effective prediction machines. I’m just hoping marketing academics are in the later category.
Read: Ajay Agrawal, Joshua Gans, and Avi Goldfarb (2018) Prediction Machines: The Simple Economics of Artificial Intelligence, Harvard Business Review Press