Where does one possibly begin?
We find ourselves in an era in which AI is being deployed within every sector, even healthcare, an industry known to be the ‘laggard’ in adopting new technologies. Along with the increased prevalence of AI comes a growing awareness among clinicians of the need to become familiar with this emerging technology.
We are Whitley and Christy, pharmacists with a passion for digital health and innovation, especially as it relates to artificial intelligence and machine learning. Realizing the challenges of tackling the subject of AI, we have made it our mission to help demystify the topic and make AI accessible to all pharmacists. When it comes to learning about AI, we have found that it can be challenging knowing where to begin with such a vast topic. With all the articles and educational resources available online to choose from, how do you know which ones will meet your needs and which ones to invest your time in? To help answer these questions, we created an AI Roadmap.
It is our vision, through this educational series, to guide others through the intricacies of AI and its applications. Much of the information provided has been gleaned through our own first-hand exploration of numerous AI resources. However, please keep in mind that this is not meant to be an exhaustive or comprehensive resource guide; rather, it aims to provide a suggested approach on how to focus your time and energy so that you get the most out of your time. We hope you will find some value in the lessons we learned from our own educational journeys
Also, don’t forget to check out our Resources page! We have put together a list of fantastic resources to help you learn more about AI and stay up to date.
Part I: Start Here
If the terms ‘artificial intelligence,' ‘machine learning,’ or ‘algorithms’ seem like another language to you, you are not alone. These are common buzzwords seen almost everywhere now, but what do they really mean? Here we offer a high level overview of AI from a healthcare perspective, where it came from, and how it fits in with what you already know.
Part I: The Framework
What is an AI model? From a clinical perspective, it is helpful to first think about models as being composed of 3 major components:
The type of inputs (or data) that go into the model.
The model’s algorithm(s)
The type of output (or inference the model makes)
Part I: Input
It’s all about the data….
Many types of data can be used in a model, including unstructured data such as images or natural language, as well as structured, or tabular data. The ability to look at unstructured data is a major distinction between machine learning and basic statistical analytics. However, machine learning usually still involves considerable data pre-processing, and often the data are transformed from one form to another.
Part I: Algorithm Learning Styles
When we talk about algorithms, there are two main things to consider: the learning style and the algorithm “type.” We will talk about learning styles first, which is a way to describe how the algorithm uses data to gain information. There are four major learning styles — supervised, unsupervised, semi-supervised, and reinforcement.
Part I: Algorithm Types
There are numerous types of machine learning algorithms. While we will not go into depth on the topic at this time, we do want to mention a few of the most common ones you may see — neural networks, support vector machine, decision trees, and naive Bayes.
Part I: Output
When we think about the output of the model, it all depends on how the model was designed to provide information. It can do this in many ways. If a model is a classifier, it might be designed to classify multiple elements or only one element. For example, if a binary model were designed to determine if an image contained a dog, the model will only provide one of two outputs — “yes” or “no.”
Part II: AI in Healthcare
While we have touched on a lot of information about AI models, ultimately, how can you start applying this knowledge to help you sift through all the news and data about emerging AI tools and applications in healthcare?
How do we go about gaining a high level understanding of the potential clinical utility of a model?
Part III: Policy, Regulations, And Standards
The disruptive potential of artificial intelligence and the role it should play within healthcare are issues that governments and organizations alike are grappling over how to handle. Regardless of the industry, we have to ensure that regulations are in place to guide the creation, and subsequent deployment, of AI models.
Part III: Aligning on a Vision and Purpose
When we look at the global conversations around AI taking place among governments and the private sector, there is a lack of clarity and consistency on what the guiding principles should be for lawmakers, researchers, and businesses.
Part III: Creating A Strategy
The "Guidance for Regulation of Artificial Intelligence Applications" draft rule lists key considerations for other government agencies when creating AI regulation.² Among the top considerations are the need to promote public trust in AI and to minimize or eliminate regulation where possible.
Part III: Putting Regulation in Place
With any new technology, unintended consequences are bound to occur and the stewardship of that technology includes vigilance and active surveillance to identify signals of a potentially undesirable impact. This is a reason why diversity and representation is so critical within AI.
Part III: A Shared Language
Even if all the necessary policies and regulations existed, it is still not enough for a given field to advance and mature. This is where standards come into play. Standards play a fundamental role in everyday life and are necessary to ensure the quality, safety, and functionality of almost everything we interact with in our daily lives.
Part IV: Why Transparency Matters
At what point can a machine be trusted to make important decisions? As George E.P. Box said, “All models are wrong, but some are useful.” It is imperative that we know how a model could be wrong in order to minimize risk. If we do not know how a model came to the conclusion it did, then we do not know when it is or is not appropriate to use, or when it will or will not be useful.
Part IV: Interpretability vS. Explainability
Not all models are interpretable (i.e. intuitively understandable). Therefore, we need another way to try to make sense of them, and that is where explainability comes into play. Can we explain a complex model by breaking it down into smaller parts that are each simple enough to be interpretable?
“What all of us have to do is to make sure we are using AI in a way that is for the benefit of humanity, not to the detriment of humanity.”
Tim Cook, CEO of Apple