How Does AI Work?

• How does AI work?
• What is machine learning?
• What are the types of AI? (real vs. science fiction)

How does AI work? A good way to answer the question, “How does AI work?” is to also ask the question, “How do people learn?” We gather or are presented information, with which, we’re able to identify trends and patterns and then draw conclusions. It can be something as simple as learning 2+2=4 or as complex as string theory. Whatever it is, we learn by observing samples. The same is true for AI. An AI algorithm takes in data and stores the information in a database. It collects samples of information and uses them to make observations that lead to conclusions. Ultimately, you could teach an AI system that 2+2=3 if you fed it enough samples that showed this was true.

What is machine learning? The world of AI is incredibly broad and includes many types of learning. One of the most common terms we hear when referring to AI is machine learning. So, when comparing machine learning versus AI, you have to think of one as the parent and the other as the child. In this case, AI is the parent, and machine learning is the child because it’s a subfield of artificial intelligence.

Machine learning involves the same type of process as traditional AI by taking in information, processing it, and making conclusions. The specialty of machine learning comes in the ability of an algorithm to improve the overall function of a machine. We see it every day when we use the predictive text feature on our smartphones, social media, and the recommendations list on Netflix. It’s all about learning how you interact with an interface, then trying to serve you better by adjusting to your tendencies. Let’s take a deeper look at how we interact with AI every day and the ways that it makes your life easier.

What are the types of AI (real vs. science fiction) AI has already been implemented into our daily lives in some ways that you might not realize. On the other hand, some forms of AI are still mostly in the science fiction phase. Let’s take a closer look at these different types of AI, including:

• Reactive machines
• Limited memory
• Theory of mind
• Self-awareness

Reactive machines. The most common AI we interact with every day is reactive machines, which is where AI can react to your actions and adapt to work better with you. You’ll notice after you’ve been on apps like Netflix, social media, and YouTube for more than a few minutes, you start to see recommendations or similar content to what you’ve interacted with pop up. Interactions can be everything from a like, comment, watching a video for an extended period of time, or even something as small as hesitating as you scroll. This is reactive AI. It reacts to interactions on your side, then pushes content that’s similar to what you’ve been interacting with. The goal of the algorithm is to feed you content you like, so you spend more time on the app.

Limited memory. As a part of machine learning, AI can maintain and recall certain types of data in limited memory. This works by retaining an image of an object or even a situation. Self-driving cars are a perfect example of limited memory because the algorithm holds onto images of stop signs, lane markings, and even brake lights and then learns how to react in each situation. What’s more, limited memory AI can learn how you drive and mimic your techniques. The more data input to the AI algorithm, the more it can improve. So, a self-driving car might be pretty good when you first get it, but the more you drive it, the better driver it will be over time.

Theory of mind. As we mentioned, some AI is still in the sci-fi phase. Anyone who’s seen “Captain America: The Winter Soldier” is probably familiar with Zola’s Algorithm. Zola’s Algorithm is an early form of AI written by a character, Arnim Zola, that can predict human behavior. Hydra (the bad guys) uses the algorithm to identify and neutralize human (or alien, in Marvel’s case) threats before they happen based on behavioral analysis. We can see the ethical problems here. In contrast, theory of mind AI could be used to better understand mental health disorders and give health providers better insight into treatment. We’re still quite a way off from the theory of the mind AI breaking into the mainstream.

Self-awareness. Of all the types of AI we see in sci-fi movies, self-awareness AI has been used as the villain the most. “The Terminator” franchise was built entirely off this premise and its ability to make decisions based on helping itself to thrive. Self-awareness AI would be able to tell a user that it’s tired and needs to shut down for a period. We see also self-awareness AI in other franchises like “Star Wars” with the different types of droids that serve the diverse beings in their galaxy far, far away. We haven’t reached a point where self-awareness AI is part of our everyday lives, and (maybe fortunately) we’re not even close to getting there.

Interacting with AI is getting to be a more common part of our lives every day. While movies and other entertainment media have done a great job showing us what can go wrong with AI, there’s plenty of good that can and will be done with the future of AI.