Human Assisted AI
Another common trend is pairing humans to evaluate results from Artificial Intelligence (AI). As great and sensational AI has been made out to be recently, it is still long way from having human-like abilities of comprehension, reasoning and intuition. For instance, in radiology, given lymph node cells, AI alone had 7.5 percent cancerous cells detection error rate, where a human pathologist had a 3.5 percent error rate; combined, AI plus human would lower the error rate down to 0.5 percent, an astounding 85 percent reduction in error.
In the short-to-midterm, human-AI teaming will result in better outcomes than either one alone would provide. The human component will be especially important in use-cases where AI would need additional, currently prohibitively expensive architectures, such as vast knowledge graphs, to provide context and supplant human experience in each domain.
More and more systems operate and adapt to new circumstances with little to no human control. Although the literature is focused on autonomous cars (Google, Tesla, and Uber) or autonomous delivery drones (Amazon), this category is much broader, for instance automated financial trading or automated content curation systems, such as creating automated news digests in sports or finance. Autonomy also includes the ability to diagnose and update the internal system, such as in identifying and patching security vulnerabilities, which is key in the world of rapidly evolving cybersecurity threats.
A visible impact of this trend will be in automating away certain parts or whole job categories altogether, while opening new opportunities for more creative roles. Key here will be how Enterprises deal with resource reallocation, workforce re-training, and whether individuals will be afforded those new opportunities.
Checkout our white paper on Four Tenets of Machine Learning and Data Science to learn how data science plays a vital role in unlocking the potential of enterprise data.