How are Analytics, AI, and ML related?

Are you ever confused about which term to use when talking about analytics, artificial intelligence (AI), and machine learning (ML)?

You're not alone. These terms are often used interchangeably, but they're actually different. Analytics tells you where you've been, AI helps you figure out where you want to go, and ML is the way we'll get there.

In this episode of Qcast, we'll understand what these terms mean and how they're related.

Sahil Brij Malhotra: Hi, I'm Sahil Brij Malhotra, director of business development at Qualetics Data Machines. Welcome to Qcast. Where we bring in our leaders, experts, and guests to talk about AI challenges and opportunities. Today we are going to interact with Mike Fowler, chief commercial officer at Qualetics, and we're going to talk to him about the impact of AI on business growth. Welcome, Mike!

Mike to someone not specializing in AI, you know the terms analytics, artificial intelligence, machine learning, they seem to be used interchangeably, can you describe how they relate to each other in the larger picture of AI solutions.

Mike Fowler: Sure Sahil, so going beyond correlating or analyzing data, artificial intelligence is more involved. So AI is about applying the more involved human intelligence to problems or tasks. It accounts for thought processes and improving on those thought processes with the benefits of technology.

If we use the example of a doctor, who's reading lung x-rays to assess medical conditions, a classic lung x-ray is a monochrome picture showing various shades of white and black to provide an outline of the inner features of a human chest, the size and density of a mass relative to its location within the human physiology are important cues to the doctor in diagnosing an ailment and understanding the shape and the density of different objects, allows the doctor to determine what objects are healthy elements that should be there versus unhealthy elements that require medical attention. Translating the series of logical steps reflected in that doctor's human intelligence process into computer programming that's the process of creating artificial intelligence. 

So, if we stay with this x-ray example, in general, it starts with translating that picture into something the computer understands. So we reduce every pixel of that x-ray image to a data point and we identify that data point's relative lightness or darkness and where it is in relation to every other pixel, and that helps the computer determine shapes and understand healthy shapes versus abnormalities. Each abnormality is then analyzed in its shape density and location in reference to a normal person's physiology, and it's compared to other known and understood abnormalities this allows the computer to predict how closely it matches what are known conditions. In health care situations the objective is not to replace the doctor's role, but to provide effective decision support to help them be more effective in diagnosing x-rays as quickly as possible in the interest of saving more lives, more efficiently. 

If you think about a doctor who's going through a long series of x-rays may be covering multiple patients, the ability to enhance that process to help make sure they don't miss anything to improve on the analyzed results. For instance, in an AI program to color code, under different conditions a different color is used, it makes it easier to call out those abnormalities in an AI  assessment for the doctor to spend more time on understanding the important data in those x-rays.

Sahil Brij Malhotra: That's pretty interesting, Mike. You know, as you mentioned about artificial intelligence. So one thing can be safely assumed that artificial intelligence is not here to replace the doctors but artificial intelligence is here to enable the doctors. Now continuing from this point, Mike, how the journey of AI further continues into machine learning, could you throw some light on that, please.

Mike Fowler: Sure Sahil. So machine learning then is an extension of the AI process. It's meant to mirror the process humans use to improve their knowledge, to learn how to do things better. So human learning occurs in this example when a doctor is in training, they may turn to an instructor or to an experienced mentor for confirmation of their diagnosis, so that they can receive feedback. Or if it's an experienced doctor they may have made a certain diagnosis but biopsy results come back to tell them, that they made an error in the diagnosis or they may be in an operating room they may have diagnosed one condition and in the midst of the operation they found that that diagnosis was wrong, probably on the operating table, they can see why what they may have missed, and that experience allows human learning to translate that feedback experientially, and adjust their logic to be applied in future diagnoses.

So machine learning does the same thing. The AI program is run against a series of x-rays with known diagnoses, the outcome of the program is compared to the correct diagnosis and recorded in such a way that the model learns from correct and incorrect outcomes, it adjusts its logic and then it improves. It's a way of calculating the probabilities of its accuracy. That allows the program's results then to be reflected in the doctor's assessment of just how accurate does this program think that its assessment is and it can focus on those assessments, that may have lower probability versus higher probability to make sure they're applying their human insight to whatever the AI program. We also by the way work to design feedback loops into the AI program, that allows any of the experts reviewing the results to provide feedback to the AI program, and that feedback allows the analysis to improve through this machine learning process constantly getting better and more accurate.

Sahil Brij Malhotra: Thank you, Mike. That provides a very helpful context for analytics artificial intelligence or AI and machine learning or ML. 

Thank you, Mike, for your insightful thoughts on the impact of AI on business growth. We just heard from Mike Fowler, Chief Commercial Officer at Qualetics Data Machines. And we will be back with more thoughts from Mike and our other team members soon.