Artificial Intelligence (AI) is not a new technology but as we start to see it’s influences in our daily lives through interactions with our phones, televisions, computers, and other ‘smart’ devices, discussions are arising in how to implement AI into the pharmaceutical industry-specifically pharmacovigilance. While the current narrative estimates full implementation to be 3-5 years away, I believe not only will implementation occur sooner, but also improving compliance with regulations will demand faster adoption. It is not a matter of waiting for the technology to be available, other industries have already adopted and are benefiting from AI, but a matter of access and implementation. To understand how pharmacovigilance processes can benefit we first have to better understand the terms we keep hearing: Artificial Intelligence, Machine Learning, and Automation.
People often use the terms Artificial Intelligence with Machine Learning interchangeably, but there is a difference. Artificial Intelligence is a broad concept of machines designed to perform tasks in a manner we consider ‘smart’. Machine Learning is an application of Artificial Intelligence based on the idea of humans providing a machine access to data and the machine will learn what to do with the data. Automation is another broad concept of systems that can perform a procedure or process without human assistance. When people refer to automation in the case of workflow automation, they are most likely referring to Robotic Process Automation (RPA). RPA is clerical process automation and works by the system developing an action list by watching the user perform a task in the Graphic User Interface ( like the computer window you are viewing now) then performs automation by repeating the same tasks observed.
The pitfalls of RPA is that while the system can perform automation, the system doesn’t have any self-learning capabilities. Machine Learning has great promise but as the system learns; the system creates their own algorithms and makes its own decisions the developers are not able to understand or explain why the system did what it did. This is the phenomenon of the ‘black box’ in Machine Learning. If the user cannot explain the reasoning of the system’s action, then how can the system be audited? Until this ‘black box’ issue is addressed the use of true Machine Learning in Pharmacovigilance may be limited to certain tasks.
Back to access and implementation: one has to realize that the scientists employed by pharmaceutical companies aren’t usually data scientists who can build and implement these types of technologies. The reason why certain industries already have AI capabilities is that they have invested in obtaining top talent. It’s a little different in the pharmaceutical industry where I’m not sure millions of dollars will be spent on recruiting and hiring top data scientists. There is much more of a demand than a supply, and top technology firms such as Google, Apple and Oracle have the best resources to recruit the best. So what does that mean for the pharmaceutical industry? Do we have to continue to outsource for technical support and hope our current vendors can hire the best? Not necessarily, there are current options to help those implement AI now. As large companies start to open their resources for developers to train their own machine learning models we are seeing more opportunities to take advantage of this technology. The challenge is that most of these opportunities still require the need for developers and data scientists-which for the pharmaceutical industry still means outsourcing.
We at MobiDox Health Technologies are developing a platform with tools to help PV business users implement AI and automation in their workflows. Our AI platform runs on rules and logic-based reasoning which you define. No technical experience necessary!
We just sent you an email. Please click the link in the email to confirm your subscription!
OKSubscriptions powered by Strikingly