Outcomes and Artificial Intelligence
- September 17, 2017
- 0 Comments
Ray Wang wrote a great post on the subject of successful AI projects and the outcomes they seek almost a year ago, but when I revisited it today, it seems just as relevant – it has aged well so far.
Rather than focus on the more dystopian or utopian visions of the disruptive nature of AI, Ray focuses on practical business value that we can drive with AI:
- “Perception describes what’s happening now.” – making sense of the information coming at us
- “Notification tells you what you asked to know.” With some additional AI, however, notifications can be tuned automatically.
- “Suggestion recommends action. Suggestions build on the past behaviors and modify over time based on weighted attributes, decision management, and machine learning.” – think supervised learning, and other techniques that involve human intervention but which can improve our decision making or speed it up.
- “Automation repeats what you always want. Automation enables leverage as machine learning matures over time and tuning.” – this calls to mind RPA to me- automatic repetition of something we want to get done efficiently.
- “Prediction informs you what to expect.” – compare to #1. Now vs the near future. Keeping in mind that prediction is still constrained by what has happened before and how current events correlate with past events. In a sense, prediction might be better stated as “how recent events line up with previous history patterns that imply future events are more likely to happen (or less likely).”
- “Prevention helps you avoid bad outcomes.” Understanding likely outcomes (prediction) allows us to bake in prevention mechanisms. This is a key selling point of our approach to complex financial onboarding for example – understanding when there is a risk to the business and recommending course of action to mitigate.
- “Situational awareness tells you what you need to know right now. Situational awareness comes close to mimicking human capabilities in decision making.” – I haven’t been seeing a lot of this as independent decision making, versus augmenting human decisions. The decisions valuable enough to throw situational awareness at also usually require some kind of human governance. However there are important exceptions in trading and the like.