Whether it’s neural networks, natural language processing, machine learning or computer vision, companies increasingly use artificial intelligence (AI) to improve enterprise solutions. All AI applications are data-driven and therefore dependent on high-quality data. In my previous post, I used an Internet search engine as an example. In this post, we’ll examine some more examples of how data quality improves AI.
The better job data engineers do when cleaning and maintaining training data, the more time data scientists will have to focus on model development and refinement. And, as those models are deployed across the organization, their output may feed other predictive models. Data quality mitigates the risk that erroneous data will cascade across the predictive ecosystem. Data lineage also plays an important role in helping predictive modeling reach its full potential by ensuring the data feeding its algorithms is well-understood. As we peer through the predictive looking glass into the future, we see that the quality of past data – and the ongoing quality of present data presently passing through those models – make predictions more accurate.
From high-frequency stock trading to medical diagnosis to self-driving cars, AI is becoming a data-driven decision maker employed in many enterprises across industries. To automate any decision, the data drivers must be reliable. Gathering the right data in the right formats and quantity – and with the right level of quality – is essential to automation. Companies like Google, Facebook and Amazon dominated their industries because they cultivated massive, high-quality data sets. Among other uses, this data fed automated online advertising (Google), automated social media marketing (Facebook) and automated e-commerce recommendations (Amazon).
Ours is an online and mobile world. Many of us are practically helpless if separated from our smartphones for too long. Today, web portals and mobile apps have become the primary point of interaction between an enterprise and its customers, employees and partners. My bank, for example, recently launched an AI-driven virtual financial assistant for its mobile banking app that helps me access balance information, transfer money between accounts and pay bills. It relies in part on natural language processing to allow interaction via voice commands and text messages. One of the biggest data quality challenges here – apart from making sure the financial data is accurate – is being able to continuously improve understanding of everyday speech. We haven’t quite achieved banking by emoji yet, but smiley-faced dollar signs will only be possible when data quality is thumbs-up.Find out how SAS Data Management works with artificial intelligence