AI vs. machine learning vs. algorithms: Providence exec explains the differences, their healthcare applications – Becker’s Hospital Review
Digital tools are rapidly changing the way healthcare services are delivered, but technology jargon isn’t always widely and accurately understood. Algorithms, artificial intelligence and machine learning are imperative to digitally transforming healthcare, but the differences between these three terms can be murky to some.
The terms are broken down below, according to Maryam Gholami, chief product officer at Renton, Wash.-based Providence’s Digital Innovation Group.
Algorithms are a critical component of getting computer systems to perform any task.
“In order to get [computers] to do anything meaningful for us, we need a method to communicate to machines how to process the inputs and signals from the surroundings and produce the desired outcomes,” Ms. Gholami told Becker’s. “We can refer to this ‘method’ as an algorithm.”
Algorithms can be designed in a way that is procedural and deterministic, or they can be designed to learn over time. There is also a wide range in which their complexities can vary.
Algorithms can give machines step-by-step instructions to follow and can also provide a blueprint that gives them past real-life examples of how a task was performed. Sometimes, algorithms are asked to do a job that humans haven’t done before or solve a problem for which humans do not yet know the right solution, according to Ms. Gholami.
Algorithms can also be deployed when the creators have a desired outcome but don’t have much data from the past to share and aren’t sure how it should be achieved. In these cases, creators start algorithms on the job and reward them every time they get closer to achieving the desired outcome so they can learn the right patterns over time.
AI is computer systems’ ability to perform tasks that usually require human intelligence, such as visual perception, speech recognition, decision-making learning and creativity.
“AI is not a single technology,” Ms. Gholami said. “It’s rather an umbrella term that refers to a system of specific problems, the data associated with the problems and the different types of methods that can be used to process the data and perform tasks that mimic human-like intelligence or even significantly surpasses it.”
Ms. Gholami said AI’s development comprises three stages. The first is narrow AI, when the machine has mastered a single task, and the second stage is general AI, when the machine is considered to be as smart as human. The final stage is super AI, when the machine becomes smarter than all humans in the field.
Machine learning is a subset of AI in which algorithms learn and improve themselves as they get used and gather more data.
Neural networks and deep learning are subsets of machine learning algorithms that are the most sophisticated and human-like in the way they learn, Ms. Gholami pointed out.
“With the increase in capturing data more intentionally and ideally in the cloud, and also the increase in computational power to process this information for training the algorithms, machine learning is becoming a much more viable option in solving many healthcare problems,” Ms. Gholami said.
She also said machine learning is commonly used in healthcare settings to automate administrative tasks such as documentation, diagnostic image analysis and voice recognition and natural language processing for virtual assistants.
More articles on artificial intelligence:
IBM’s retreat from health business spotlights AI challenges in healthcare: 4 things to know
How hospitals are using AI to teach physicians to better express empathy
5 data issues limiting AI’s potential in healthcare
© Copyright ASC COMMUNICATIONS 2021. Interested in LINKING to or REPRINTING this content? View our policies by clicking here.