The algorithms that form the heart of machine learning have been around for decades, but computers have only recently reached the level of processing power needed to use the techniques in practical scenarios.
Artificial Intelligence (AI) programs today can learn to identify objects in images and video, translate between languages, and even master arcade and board games. In some cases, like DeepMind’s AlphaGo program, the AI even exceeds top humans at the task at hand!
What is machine learning?
Machine learning is an application of Artificial Intelligence where we give machines access to data and let them use that data to learn for themselves. It's basically getting a computer to perform a task without explicitly being programmed to do so.
Machine learning explained
Your kids have probably watched a battle bots competition before, right? You know, where robots are coded with an algorithm (a set of instructions that are followed to accomplish a task; it’s a computer’s thought process) to attack and "battle" each other.
Well, if machine learning was used in this situation, the robot itself would make a decision in the moment based on the information it has been given. Meaning, the robot would choose to perform either option A or option B, rather than being told through code to always perform option A no matter what.
So, instead of coding software with specific instructions, machine learning trains an algorithm so it can learn how to make decisions for itself.
Read More: How Does Artificial Intelligence Self-Learn?
How machine learning works
As mentioned, machine learning is all about training an algorithm, and in order to train an algorithm, you need a neural network, which is a set of algorithms inspired by biological neural networks—and modeled after the human brain, which consists of individual neurons connected to each other.
In machine learning, a neuron is a simple, yet interconnected processing element that processes external inputs. A neuron receives data through its inputs, processes the data using weights, biases, and an activation function, then sends the result onward as its output.
Once you've got a neuron that takes input data and outputs a value, you will have to train it by adjusting the weights and biases inside the neuron until the output is ideal.
Machine Learning uses these neurons for a variety of tasks like predicting the outcome of an event, such as the price of a stock, or even the movement of a soccer player during a match. A neuron uses input data from any past events to predict the outcome.
What can machine learning do?
One of the main categories of machine learning problems is supervised learning. These are problems where there is available training data so the program can get feedback on its performance as it learns.
Tasks, like playing games and identifying objects, would fall under supervised learning because the computer is getting feedback as it learns. Did it guess what the object in the image was correctly? Did it get a high score on the game, or lose 10 seconds into playing? Feedback allows it to adjust its decision-making process so it can do better next time.
Two of the most common sub-categories of supervised learning problems are classification and reinforcement learning.
In a classification problem, the program is given a set of inputs and has to learn to classify those inputs correctly, like an email spam filter or image recognition program.
In reinforcement learning, a program (the “agent”) interacts with an environment dynamically, making choices for its next course of action. Based on the current state of the environment, the positive and negative rewards, and actions taken, the agent must learn the best method to accomplish the task.
Reinforcement learning agents can learn to play Ms. Pac-Man, master the games of Go and Chess, compete against pros at Dota 2, and even have started learning to play different types of video games, and more complex strategy titles like Starcraft 2.
Machine learning examples
Machine Learning is used to find solutions to various challenges that arise across a variety of different scenarios and environments.
Smart Cars
Machine learning can evaluate the driving environment and driver condition based on information obtained from different external and internal sensors.
For example, a smart car is able to make an observation and detect an object, and can then identify that object using machine learning. Since there are so many different objects in the world, it would be nearly impossible to explicitly code in what every object is or could be into the car's framework. However, if you teach the car to identify objects through machine learning, it can make those decisions itself.
Music and video recommendations
Kids familiar with music apps have probably wondered how the app can suggest other songs they might enjoy listening to. Same with YouTube—how does it know which video kids might want to view next? All of this is made possible with machine learning. The algorithm is trained with previously watched videos, and then from that info, builds and improves an algorithm that defines the listener's or viewer's taste.
Web search
The process to find results after searching for something in a search engine is incredibly complex and uses machine learning. How does Google know that all the thousands of results listed are related to a search inquiry? No one is manually categorizing everything on the internet—it's all a very advanced form of AI and machine learning that decides which images are "dogs" and "cats" and which articles are related to the "Loch Ness Monster" or "Bigfoot."
Machine learning as a career
So, it should be obvious by now that machine learning is one of the coolest emerging fields in tech—but why else should your child hop in and start learning about it?
In the coming years, many companies like DeepMind and OpenAI hope to solve general artificial intelligence, which is a term for an AI that can learn and perform any task put in front of it. This breakthrough is likely still years in the future, but it has the potential to revolutionize how human beings interact with technology, the job market, and society in general.
In the shorter term, machine learning has practical business applications like analyzing large volumes of data, powering self-driving vehicles, and assisting medical diagnoses. As AI research advances, the number of tasks it can perform will only increase. Companies are already desperate for AI experts and aggressively hiring those with expertise in the field.
Sound like a good fit? Students can take their first steps towards revolutionizing technology and society this summer through any of the cutting-edge courses below!
Intro to Python coding (virtual tech camp)
Machine learning tutoring (1-on-1 online lesson)
Python summer camp (coding for machine learning)
Artificial intelligence summer camp (data sets, probability, and stats)
Machine learning summer camp (for coding deep neural networks)
View all coding courses for kids