Sometimes “Small Data” Is Enough to Create Smart Products
When thinking about practical applications for artificial intelligence in your business, it’s easy to assume that you need vast amounts of data to get started. AI is fueled by data, and so it only makes sense that the more data you have, the smarter your AI gets, right? Not exactly. When it comes to extracting intelligence by applying AI to data, context matters. In other words, you can build the biggest data lake imaginable, but if you don’t know what you’re trying to find and you don’t have the right data to do it, then you’re not going to get where you want to go. AI is a huge set of technologies, each with a specific, fine-tuned purpose. Companies that can zero-in on the impact they want to see and focus on curating the right datasets mapping to those goals have the best opportunity for generating really impactful results from AI.
When thinking about practical applications for artificial intelligence in your business, it’s easy to assume that you need vast amounts of data to get started. AI is fueled by data, and so it only makes sense that the more data you have, the smarter your AI gets, right? Not exactly.
When it comes to extracting intelligence by applying AI to data, context matters. In other words, you can build the biggest data lake imaginable, but if you don’t know what you’re trying to find and you don’t have the right data to do it, then you’re not going to get where you want to go.
That’s because AI is not some magical black box that can ingest mountains of data and then just spit out results. AI is a huge set of technologies, each with a specific, fine-tuned purpose. Companies that can zero-in on the impact they want to see and focus on curating the right datasets mapping to those goals have the best opportunity for generating really impactful results from AI.
Consider how the United States Postal Service (USPS) automates mail sorting. With the help of machines and advanced optical character recognition (OCR) technology, the USPS can now read and process 98% of all hand-addressed mail and 99.5% of machine-printed mail without human assistance. By linking this technology with a relatively small and finite data set of U.S. zip codes and cities, the USPS can now process upwards of 36,000 pieces of mail per hour. With the USPS facing harsh financial challenges in recent years, the impact of this automation is immeasurable.
Another interesting example of small, high precision data being used to make big gains with AI can be found in the airline industry. In 2015, Boeing launched the Aerospace Data Analytics Lab in partnership with Carnegie Mellon University to develop AI technology for airlines. One such project aims to dramatically reducemaintenance costs with AI by standardizing maintenance logs.
Every aircraft is required to keep highly-detailed maintenance logs. However, when planes travel around the world, communication starts to breaks down. Basic language barriers are the first stumbling block. From there, it only gets worse. Some logs are captured digitally; others are hand-written.
