Speaker A
We use AI, or artificial intelligence, every day, whether we realize it or not. AI has entered every field, including work, entertainment, education, and shopping. In fact, it might even be an AI recommendation that brought this video to you. What exactly is AI? How does it learn things? In this video, I will try to explain the technical aspects of AI in the simplest way possible, with examples that anyone can understand. Simply put, artificial intelligence is the ability of machines to mimic human intellectual capabilities to think and make decisions. However, machines do not gain this ability on their own. It is developed by specifically training them. That is why it is called artificial intelligence. AI needs a lot of data to learn. The more accurate the data, the smarter the AI becomes. A key method for learning in artificial intelligence is machine learning. It is a term used to describe how machines like computers learn things. Machine learning happens based on an algorithm. An algorithm is a set of instructions on how to process incoming data and turn it into an output. These instructions are adjusted slightly over many iterations until an accurate output is achieved. This process is called training. Therefore, before training, an algorithm can be compared to the brain of a newborn baby. It won't have much information, but it has the capacity to learn new information. An algorithm that has learned new information through training is called an AI model. Machine learning can be divided into three types: supervised learning, unsupervised learning, and reinforcement learning. Supervised learning is like teaching a child by showing them an apple and saying this is an apple, or showing an orange and saying this is an orange. Later, even if you show the child an apple they have never seen before, the child will be able to identify it as an apple. In supervised learning, the data we provide to the algorithm includes both the input and the expected output based on that input. Once trained, the resulting AI model will try to provide an output based on the input data, even if it is data the model has never seen before. AI models for detecting spam, diagnosing diseases, and image classification are developed through supervised learning. Unsupervised learning can be compared to how a newborn baby distinguishes between familiar people and strangers. The baby sees many faces. Some faces are seen together many times, for example, family members, while others are seen only rarely, for example, neighbors. In this way, the baby creates groups in their mind like familiar and strangers. They do this without knowing who they really are or what their names are. Similarly, in unsupervised learning, a lot of data is given to the algorithm, but the expected output is not provided. The algorithm's job is to find similarities in the data and categorize them into groups based on those patterns. Once trained, the resulting AI model can take new data and determine which group it belongs to. AI models for customer segmentation, anomaly detection, and topic clustering are developed through unsupervised learning. Reinforcement learning is like a child learning to walk. If they take a wrong step, they fall. Then the child will be careful to avoid such steps in the future. If a step is correct, they move forward. Then the child will try to take such steps again. Gradually, the child learns to walk. In reinforcement learning, instead of raw data, we provide the algorithm with an environment to develop a skill. The algorithm tries to do certain things within that environment. If an action helps it reach the skill it needs to develop, it is given a reward. Then the algorithm will try to do those things again. However, if an action moves it away from the skill it needs to develop, it is given a penalty. Then the algorithm will try to avoid doing those things again. Gradually, it develops that specific skill. AI models for self-driving cars, gameplay, and robotics are developed through reinforcement learning. Artificial intelligence can be divided into several levels based on its capabilities. What we see now is ANI, or artificial narrow intelligence. It can only do what it has been trained to do. For example, image detection or text generation. Its intelligence is lower than that of humans. The next stage is AGI, or artificial general intelligence. It will have intelligence in many areas just like a human. It will also keep learning new things. In terms of intelligence, it will stand at the same level as humans or slightly higher. Coming after AGI is ASI, or artificial super intelligence. This will be superior to all humans in terms of intelligence. In such a scenario, AI will be able to learn without human help and develop even more capable AI by upgrading itself. It is believed that AGI and ASI might become reality in the future. What do you think? Will these be possible?
