Digital measurement solutions (DMS) can be found in various forms, mostly in the form of sensors and wearables. As they have a wide range of applications, they can be ...Read more
In 2020, EU Science Hub published a report on the use of Artificial Intelligence (AI) in Medicine and Healthcare. The expected economic impact of AI is huge. In the US alone, the application of AI in Healthcare is expected to save $150 billion by 2026. These applications are expected to impact the entire clinical practice, enabling improved clinical data analysis, personalized medical interventions, intelligent companion caretaker robots and much more. Despite this projected impact, definitions of what AI and related technologies are, are rather vague and often confusing, hindering productive discussion and adoption.
In this blog post, we aim to demystify some commonly used terms related to AI. To make this fun, we are going to brainstorm on how to make a robot caretaker and explain what AI related technologies we will need along the way. This caretaker will move around in the physical world through robotics, make autonomous decisions using AI and acquire rules to mimic human behavior through traditional programming or machine learning (ML).
Designing a virtual caretaker using robotics
According to the online Encyclopedia Britannica, Robotics is defined as “the design, construction and use of machines to perform tasks traditionally done by human beings.” Robots are already used to perform simple repetitive tasks in industry or to perform tasks in conditions that are inhospitable for human beings. According to this definition, the caretaker we have in mind is a robot that will take over caretaking tasks from humans. It will have to move physically throughout the world and have certain cognitive abilities, such as recognizing patients, connecting them to medicines and so on. It is with this cognitive aspect that Artificial Intelligence can be of use.
Autonomous movement and decision making using Artificial intelligence
Our caretaker can be designed to be controlled externally by a human user or be able to perform its tasks without human intervention. These tasks include deliberate movement, recognizing patients and deciding which patient to give which type of medication. Programming these “human-like” behaviors and others like speech recognition into a machine to achieve independent operation fall under the umbrella of AI. AI is defined as the ability of a digital computer or computer-controlled robot to perform tasks commonly associated with intelligent beings. It is important to note the difference between robotics and AI. Robotics always involves movement of a machine and physical interaction with the world, but it does not require autonomy, since robots that are operated by humans exist. On the other hand, AI always requires autonomy, but does not require physical interaction with the world as demonstrated by automated trading bots in finance or chess playing software that beat world champions. Our ideal caretaker will nevertheless draw from both.
Acquiring rules to mimic human behavior: Traditional programming vs Machine Learning (ML)
Now we have decided that our caretaker will move physically through the world (robotics) and will operate autonomously (AI). The next question is how to allow our caretaker to operate autonomously. Here, we have two choices.
The first approach involves predefining a set of rules that our virtual caretaker draws from to guide its actions. For instance, if the caretaker sees a patient then it will give the medication. This is the traditional programming approach. This used to be the dominant form of producing an AI, before the advent of ML algorithms. Such AI applications are sometimes called expert systems.
The second approach would use ML. ML is defined as the discipline concerned with the implementation of computer software that can learn autonomously. Using this approach, we would not program any rules inside of our virtual caretaker. Instead, we would program the ability to learn from example. This opposition between ML and traditional programming leads to an important notion of when to use one over the other:
Use traditional programming whenever it is practically feasible to define a limited set of rules that do not change over time. Use ML when this is not possible, but you do have a lot of data available to learn from.
Through this example of our caretaker, we have now learned the major differences between robotics, AI, traditional expert systems and machine learning. In summary, robotics involves the creation of machines that perform human tasks. This ability to physically perform human tasks, does not necessarily require the robot to also have the ability to mimic human cognitive abilities. AI involves the creation of machines that mimic human cognitive abilities, which might or might not involve physical interaction with the world. Finally, expert systems and machine learning represent two distinct programming approaches to mimic human abilities and achieve AI. The former uses a set of predefined rules to guide decisions, whereas the latter uses a set of data to infer these rules for itself. From that, we derived the relative strengths of expert systems and machine learning.
Are you interested in how your data can help your business thrive in these uncertain times? At Modis Life Sciences, the IT Solutions team has extensive experience working in the pharma/biotech/medtech sector and understands your needs. With a strong expertise in managing, analyzing and visualizing your data interactively, we help democratize analytics insights and put crucial business insights at the fingertips of every client.
For more information on our IT solutions, please contact Sven Hofman at firstname.lastname@example.org.
If you have any questions regarding this blogpost, you can contact email@example.com .