AI is a field that has seen numerous advancements in the past two decades. But what exactly is it? Well, it is in the name. AI stands for Artificial Intelligence, which means machine (artificial) thinking (Intelligence). Artificial means not natural; intelligence is how you react based on past experiences. However, this is a misnomer. AI is just a machine trying to get as close to the best solution to fit the data. They cannot think as we do, or be truly creative (although it can look like they can).
AI allows computers to handle tasks usually done by humans.
AI allows computers to handle tasks usually done by humans. These tasks include understanding language, solving problems, and learning. Many learning platforms like Dualingo (Birdbrain) and Khan Academy (Khanmingo) have already integrated AI chatbots into e-learning. This is a clear example of AI. On YouTube, an AI selects the recommended videos you see. A similar concept is in Netflix and Disney +. In video games, some NPCs (non-player characters) use AI. This helps them decide what to say. The conversion between text and speech usually also uses AI.
First Artificial Intelligence Model
The idea of AI has been around longer than many believe. The first AI model, made in 1951, was a heuristic for playing checkers or draughts. This model used an algorithm to look ahead by several steps and determine the move it should make. Frank Rosenblatt created the perceptron in 1956, which was the first model that could learn as it played. The Perceptron used weights. Each weight was multiplied by an input, after this all the results were added together. This process created the net input function, and this early form of Artificial Intelligence is about learning from data.
After that, this function went through an activation function. The final output was a number. This number showed how strongly it favored choosing a specific option. The perceptron could only determine true or false. Two methods improved this: the nearest neighbor algorithm, which matches a new example to the closest known one, and backpropagation, which helps the system learn from mistakes and better recognize patterns. This led to increased excitement about Artificial Intelligence.
This machine would have the general intelligence of an average human being.
More than 50 years later, we are still nowhere near that point. As a result, this initial excitement eventually led to an AI winter. An AI winter occurs when people lose interest in AI and stop using it for a long time. For instance, in 1980, scientists created expert systems, which were designed to make decisions using a set of rules called if-then logic. You can think of them like decision trees that map out different paths based on specific choices. However, at that time, computers were not very powerful, and AI technology was still underdeveloped. Consequently, this lack of progress led to another AI winter in 1987.
Moore’s Law
By 1994, things had improved. Moore’s Law, which says computing power doubles every two years, helped make this progress. Better algorithms were created as a result. These algorithms used training data to make their decision trees. Neural networks are a type of algorithm. They advance beyond the older Perceptron model by including many layers of nodes. In a neural network, each input node has weights. These weights connect the input nodes to the first hidden layer. Then, they connect the hidden layers to each other. Finally, they connect the last hidden layer to the output. It can be displayed like this:
Each line represents a connection between two nodes. At times, connection weights are shown through their thickness or color. Alternatively, the Greek letter omega is placed above them, as I have done. The training data is used to refine the weights and improve the performance of Artificial Intelligence. At the end, an activation function (which I haven’t displayed) is used to refine the output. Now, let’s look at reinforcement learning. This is rewarding/punishing an agent based on how great it is doing. The agent applies gradient ascent. This strategy aims to find a local maximum point on a curve. Its goal is to achieve the highest reward.
visualize the difference
You may notice that I said a local maximum instead of the maximum. I visualize the difference here:
An AI using reinforcement learning can hit a performance limit. This occurs if we don’t carefully select the right loss function. Every AI uses a loss function. So, we must choose each AI’s loss function with care, as with reinforcement learning. Lastly, I want to highlight the increase in AI patents from 2010 to 2020, illustrating that Artificial Intelligence is about continuous innovation.
Artificial Intelligence is Future
AI is becoming more and more dominant in our lives. For example, back in the 1980s and 90s, people dreamed of robots driving cars. However, nowadays, the robot is the car itself. AI has already become a part of our everyday lives and will continue to evolve in the future. Furthermore, AI will become even easier to access and create.
In Part 2, I will demonstrate how to create Artificial Intelligence with minimal code using Python. Moreover, I will cover how kids and teens can learn AI and ML for free. In particular, they can take advantage of various online resources that do not require coding, making it accessible to everyone.
AI has and will continue to appear in our everyday lives
Additionally, AI will be ever easier to access and create. In part 2, I will show you how to create AI with minimal code using Python. I will cover how kids and teens can learn AI and ML for free. They can use different online resources that don’t require coding.
Additionally, to stay updated with the latest developments in STEM research, visit ENTECH Online. Basically, this is our digital magazine for science, technology, engineering, and mathematics. Furthermore, at ENTECH Online, you’ll find a wealth of information, covering various aspects of AI.
