it261: AI v. Human Intelligence: Which Application is Right?

Tuesday, 11 September 2018

AI v. Human Intelligence: Which Application is Right?

When new technology appears and matures, it often spawns new business models. The new business models compete with existing ones, to offer better products, better prices, or better customer experiences, forcing existing businesses to adapt and account for the new competition. AI is one of those technologies that could be the basis of new business models, much the way Amazon leveraged the Internet, writes Raj Ramesh in AI Is Driving New Business Models: How Do We Adapt?

As one of the most powerful technologies humans have invented, AI will spawn new business models that will be vast and deep, the likes of which we have not yet seen.

Ramesh defines AI as commonly used umbrella term for many subdisciplines, including computer vision, machine learning (ML), deep learning, and natural language processing (NLP), among others. Most of these consume large volumes of data, build models, and use those models to make inferences. The models essentially represent the patterns of the data that was used to build the models in the first place.

According to Ramesh, business processes that generate large amounts of high-quality information are good candidates for automation by AI. For example, the vast amount of data collected by an insurance company over many years can be used to build ML models to make inferences about the level of risk new customers pose. However, since business strategy considers current information in addition to past information (as well as informed guesses about the future) business strategy decisions still should be made by humans, writes Ramesh,

… because human creativity is vital, and AI systems are currently unable to formulate strategy there are situations in business where a collaborative solution between machines and humans is optimal. Consider credit card fraud. When a customer’s credit card number is stolen and used for purchases, it is impossible for humans to spot variations in patterns across billions of transactions. It is also not possible to write specific business rules to spot fraud due to the large number of variations. ML algorithms, on the other hand, could be used to create alerts for uncommon charge patterns by customers. One customer may be charging large amounts on a credit card daily, while another might rarely charge large amounts. So the spending patterns are different across different customers. With ML, the system could alert human operators to look at a smaller subset of possible fraudulent transactions to identify fraud.”

Ramesh predicts that with machines taking over many tasks that humans previously performed, companies will change the makeup of their organizations, moving affected employees to areas that require human skills such as imagination and creativity. Humans will focus on things that AI cannot do well today. This shift, with its fundamental organizational changes, will create new business models.