AI Learning: How Machines Learn Like Humans

 Today I’m going to write about how the ways AI learns are similar to how I learn. Sounds interesting, right?

let me clarify what I mean by AI. AI stands for artificial intelligence, which is a broad term that covers any system that can perform tasks that normally require human intelligence, such as understanding language, recognizing images, playing games, making decisions, etc.

AI can be divided into four types:

Strong AI: this is the ultimate form of artificial intelligence, it can reason on its own and act human like, it has not yet been made.

Weak AI: it cannot truly reason and solve problems, but can act as if it were intelligent. Weak AI holds that suitably programmed machines can simulate human cognition.

Applied AI: aims to produce commercially viable “smart” systems such as, for example, a security system that is able to recognise the faces of people who are permitted to enter a particular building. Applied AI has already enjoyed considerable success.

Cognitive AI: computers are used to test theories about how the human mind works — for example, theories about how we recognise faces and other objects, or about how we solve abstract problems.

Now, how does AI learn? Well, there are many ways, but one of the most common and powerful ones is called machine learning.

Machine learning is a branch of AI that allows systems to learn from data and experience, without being explicitly programmed.

Machine learning can be further divided into three types: supervised learning, unsupervised learning and reinforcement learning.

Supervised learning is when the system learns from labeled data, which means that the data has the correct answers or outputs attached to it. For example, if you want to teach an AI to recognize cats and dogs in images, you need to provide it with a lot of images that are labeled as either cat or dog. The system will then try to find patterns and features that distinguish cats from dogs, and use them to make predictions on new images.

Unsupervised learning is when the system learns from unlabeled data, which means that the data has no correct answers or outputs attached to it. For example, if you want to teach an AI to cluster similar images together, you don’t need to tell it what the clusters are or how many there are. The system will try to find patterns and features that make some images more similar than others, and group them accordingly.

Reinforcement learning is when the system learns from its own actions and feedback, which means that the system interacts with an environment and receives rewards or penalties based on its actions. For example, if you want to teach an AI to play a video game, you don’t need to provide it with any data or rules. The system will try different actions and see what happens, and learn from the consequences.

So far so good? Now let’s see how these ways of learning are similar to how I learn.

Supervised learning is similar to how I learn from teachers or books. When I want to learn something new, I often rely on someone or something that already knows the answer and can guide me through it. For example, when I learned how to drive a car, I had an instructor who taught me the rules of the road and gave me feedback on my driving skills. Or when I learned how to play guitar, I had a book that showed me the chords and songs I could play. But for computers supervised learning comes with making the computers model using neural networks (brains), then training that brain by feeding it, the goal state, and the initial state of the problem to be solved, it’s like getting the answer to a complex math problem and being told to map out the solution to the problem.

Unsupervised learning is similar to how I learn from exploration or curiosity. When I want to learn something new, I sometimes just try things out and see what happens, without having a clear goal or direction. For example, when I learned how to cook, I experimented with different ingredients and recipes, without following any instructions or measurements. Or when I learned how to draw, I doodled with different shapes and colors, without copying any reference or style. But for computers unsupervised learning also comes with making the computers model using neural networks (brains) and training the brain by giving the computer data and telling it to figure out what the answer is supposed to be.

Reinforcement learning is similar to how I learn from trial and error or feedback. When I want to learn something new, I sometimes just do it and see if it works or not, and adjust my behavior accordingly. For example, when I learned how to ride a bike, I fell down a lot and got bruises and scrapes, but I also got better at balancing and steering. Or when I learned how to speak a foreign language, I made a lot of mistakes and got corrected or laughed at, but I also improved my pronunciation and vocabulary.

How does the AI come to the goal state while learning?

The AI uses things like search to create patterns or to find answers to the goal state.

We have two types of searches:

Informed search and uninformed search

Informed search: this means that you know what you’re looking for.

Uninformed search: this means you have no clue about what you’re looking for.

Under uniformed search we have different search algorithms, like breadth first search, depth first search.

Breadth first search: this is an algorithm the search all possible nodes before moving down the path to the goal state.

Depth first search: this algorithm picks the furthest path and keeps going down until it reaches a dead end, then it switches to the next path.

I will give basic definitions for these algorithms, if you want to know more about them, follow and drop a comment. I will make a separate blogpost for them.

Components of AI

Agent:

An agent reduces the work the artificial intelligence is supposed to carry out, they are trained to carry out specific tasks. An Intelligent Agent must sense, must act, must be autonomous (to some extent),. It also must be rational.

Rationality:

Rationality assumes that the rational agent knows all and will take the action that maximizes her utility. Human beings do not satisfy this definition of rationality. Rational Action is the action that maximizes the expected value of the performance measure given the percept sequence to date. However, a rational agent is not omniscient. It does not know the actual outcome of its actions, and it may not know certain aspects of its environment. Therefore rationality must take into account the limitations of the agent. The agent has too select the best action to the best of its knowledge depending on its percept sequence, its background knowledge and its feasible actions. An agent also has to deal with the expected outcome of the actions where the action effects are not deterministic.

As you can see, the ways AI learns are not so different from the ways I learn. Of course, there are also some differences and limitations between human and machine learning, but that’s a topic for another blog post. For now, I hope you enjoyed this one and learned something new. If you did, please leave a comment below and share it with your friends. And don’t forget to subscribe to my blog for more fun and funny content about AI and other topics. Thanks for reading!