As we look forward into 2017, intelligence augmentation (IA) will begin to take its rightful place alongside artificial intelligence (AI).
Here’s one of my talks about this. Summary below.
- Most of the successful AI use cases today are fully automated, with no human in the loop. “Which advertisement should I show?”, “What is the best book to recommend?”, “What is the best film to recommend?” are typical examples here.
- The reason: these are the easiest to solve, the low-hanging fruit. They represent problems where data is abundant and the patterns from the past work well in the future.
- But there are thousands of use cases that don’t fit this pattern.
- The future of AI is about a combination of AI and IA: intelligence augmentation. This is when we invite humans into the loop
- The existing machine learning stack (data, algorithms, visualizations) is an important part of IA. But instead of answering the question: “What does the data tell us?” we can answer “If I make this decision today, what will be the outcome?”
- This requires a combination of data with human expertise.
- Human expertise is needed when the data is not adequate, often because the situation has changed or we don’t have data about it. Humans understand systems, but the vast majority of the time we won’t have enough data (from the past) to understand the full picture. So the two must work hand in hand.
Bottom line: the vast majority of the important problems that AI will help us to solve in the future will require humans in the loop. And to do that we must build hybrid models that combine human expertise with machine intelligence. Decision intelligence is, simply put, the science of doing this in a way that minimizes cognitive friction – making the human/machine collaboration as simple and intuitive as possible.
What do you think?