The American and Chinese tech giants are racing to patent machine learning algorithms. While so far, much of the work has ended up in academic communities and has been open sourced, as larger revenues start to become associated with AI deployments, the fangs of the FAANGs will come out. Further, when we look at Asia vs. American patent deployments, which we did in our Augmented Finance analysis, Asian patents start to outpace those of the West. So is it all worth it? When the tech hits enterprise deployments however, we hit a snag. See below two reports on the topic.
In the first, Bain points to slow adoption of AI and robotic process automation in corporate finance departments, due to very human factors -- there are too many tools, they tools are not integrated, they make too many errors, and user interfaces are hard to understand. This is in part due to how AI is sold and deployed to local environments through a bespoke consulting model. A better approach is an API integration into workflows behind the scenes as a service.
The second report, from O'Reilly, shows the state of adoption in enterprise of machine learning (ML) and associated headcount. We highlight the chart breaking out adoption by stage and geography: (a) companies exploring the technology having implemented anything yet but at least thinking about it; (b) early adopters have been working with a system for at least 2 years; and (c) sophisticated users have had something in place for 5 years. Though the sample size of 10,000+ is quite large, we are surprised to see Western countries lead Asia in enterprise ML adoption. Perhaps the difference lies in whether the software faces into the corporation to make it more cost effective, or whether the software faces the public and acts as the interface. We need to look no further than a dystopian NYTimes article describing the current state of machine vision monitoring of Chinese citizens by various Skynet companies to see that Western society is quite far behind. Perhaps thankfully!