Friday Nov 10, 2017
Dr. Elam,
Thank you for the opportunity to participate in the Accounting Advisory Board, and the most recent meeting! I hope my contributions were helpful. Nevertheless, there was much more I wanted to contribute about Artificial Intelligence, but time did not permit. If I may, I would like to share a couple of things that are important to your students’ concept of this oft misunderstood technology.
First, it is essential that your students understand the difference between Automation and Artificial Intelligence. Automation is everywhere. Artificial Intelligence is rare. Automation and Artificial Intelligence are fundamentally different. While both use computers, that is about the only similarity. To use an old phrase, it is like “comparing apples and oranges.” Automation helps us humans manage monotonous, repetitive tasks, and it is very rules-based. Think, if “x” then “y.” In other words, it is smart enough to follow orders.
On the other hand, Artificial Intelligence, like humans, is horrible, and I can’t stress this enough, it is horrible at following rules. Rules will predictably and reliably corrupt Artificial Intelligence. Artificial Intelligence (AI) is, like humans, designed to identify patterns, learn from experience, and select appropriate responses in situations that are relevant to its experience. It gives a computer the ability to learn without being explicitly programmed. In other words, it can mimic what a human might think, say or do.
Second, it is important to understand why true AI is still so slow to be adopted in the real world. The biggest problem with the adoption of AI is ROI. Most companies find that, while they may find the promise of AI attractive, they simply don’t have the necessary resources ($$$$$$$). First, acquiring a truly smart AI “engine” is expensive. Second, the “smarter” the AI engine, the more up-front time, effort and people it requires to plan and filter what is fed into the AI engine, so that the AI is not corrupted. It is surprisingly easy to corrupt any AI by feeding it seemingly innocuous information or context that leads it astray. This planning and filtering requires a lot of time, people and effort. Monitoring and filtering information and context as you go along requires more time, people and effort (think more money). Unlike Automation, which you can program and “forget,” AI requires considerable time and attention.
A real-world example of the time, attention and money that is required with AI, is the story about IBM’s Watson that I mentioned during the meeting. While Watson identified, in less than a second, malware on several computers that anti-virus software had missed, the task of bringing Watson to that point was both arduous and expensive. IBM fed Watson massive amounts of information from the Internet and the Dark Web. They also spent considerable time and effort examining and filtering the information before they fed it to Watson. (Their Vice President would not tell us how much time and effort were involved, but only that it was “considerable.”) In addition, they fed Watson a lot of information from their expensive computer event log Content Collector. Only then was Watson ready to find the malware without being programmed or told to do so. Certainly, that is not the end of the story. I’m sure Watson is doing so much more. IBM’s Vice President told us that they are offering “Watson as a Service” – no doubt to help with the ROI!
In short, AI is very different from Automation in nature, function and cost. While Automation is quite prevalent and useful in the Accounting field, AI may be more suited to other business fields and endeavors.
While this is a bit long (my apologies), I hope you find it helpful.
John
John Wallace, CISA, CRISC | 
Resources Global Professionals
Consultant
Consultant
Comments