Introduction to Machine Learning

When you watch a video on Youtube, there are millions of videos, but only a select few will get recommended for you to keep watching. How are they chosen? And why are they so familiar and beguiling?

Similarly, when you visit EBay or Amazon, the site will recommend products it thinks you might like and display them on the landing page to get your attention. How does it know what you’re interested in?

Now skip ahead a few years and imagine you’re being taxied along in your self-driving car. Suddenly, something streaks across the road in front of the car. How does the car know whether  the object is a paper bag, or a human being?

The algorithms that are in charge of all of these complex processes are not built directly by people,  they’re the result of something called machine learning.

Why is machine learning important?  It helps us detect patterns and solve problems in new ways. Already, machine learning algorithms can play Chess, Go and video games with superhuman ability, and can detect forms of cancer with greater accuracy than many human doctors. They’re also behind facial recognition software and page ranking algorithms, and yep, they’re the things telling your Tesla that it’s a human on the road, not a bag, therefore your car needs to stop. With every passing year, AI’s and machine learning algorithms are infused more into the fabric of our lives. We rely on them, we interact with them, we cooperate with them. So what exactly are they and how do they work?

There are a few ways that a machine can ‘learn’ to do these things.

Genetic Breeding Model

Although people cannot build the algorithm to specifically solve a complex problem, we can build an algorithm that ‘randomly’ builds multiple algorithms. This is very similar to how evolutionary theory works by introducing random genetic mutations.

Successful algorithms and copied and rebuilt while unsuccessful algorithms are discarded.

Eventually after thousands or millions of iterations, the resulting algorithms will be very competent in passing the test that the human has set.

Deep Learning Model

Similar to the genetic breeding model, deep-learning has some key differences in the way the programmers design the algorithm-builder.

Instead of building multiple algorithms, there is only one algorithm with hundreds or thousands of adjustable parameters.

The parameters of this algorithm are then tweaked and tested thousands or millions of times in different combinations until it is successful.

Machine learning and artificial intelligence is already a huge part of our society and they will be a huge part of many of the robots of tomorrow. If you’d like to introduce your kids to coding and robotics to ensure they’re at the forefront of technology, then be sure to check out our courses below.

Beginner Robotics

Advanced Robotics

Parents and Toddlers Programming

Mixed Reality Coding

Leave a comment

Please note, comments must be approved before they are published