What Question are You Trying to Answer?

Posted on August 30th, 2017 by Chip Davis

Rise of the Machines – The past two years have revealed a prolific rise in machine learning and artificial intelligence. Platforms that automate creation of analytical insights are on a very steep evolutionunnamedary curve and the range of subject matters that can be probed with these tools is attracting huge amounts of investment capital.

Many years ago I heard an extremely successful enterprise software sales person make the following observations: (i) a lot of money has been made from “adequate” software, and (ii) most software is bought because it can do “one thing.” It is the second observation that has stuck with me the most.

The Implicit Challenge – The notional allure of artificial intelligence is the initial belief that it can replace non-artificial intelligence (i.e. people). The mechanization of data prep and measurement does displace human tolling, however, measurement requires constant adaptation (a highly-developed human skill).  Where it becomes tricky is when the intelligence requirement pertains to questions oriented toward pre-existing scientific disciplines (e.g. oil & gas exploration and production). By their nature and history, these disciplines want to know specifically how an answer is derived – they will not accept “black box” conclusions.

What artificial intelligence is trying to do is mimic and augment real intelligence.  Machine learning is a scaling technique for taking the best minds (subject matter experts) and imparting their own insights into a system from which all may benefit. The competency of this technique is only as good as the subject matter experts regarding the question at hand. How good are they at answering my particular “scientific” question? What background makes those best minds reliable scientists? When parsed through this lens, the revelation is a certain practical irony:  artificial intelligence all comes down to people.

What do you think of this post?
  • Awesome (2)
  • Sucks (1)
  • Useful (1)
  • Interesting (1)
  • Boring (0)