When it comes to hot terms in recent years, "artificial intelligence" must be on the list. As ChatGPT went viral last year, "Artificial Intelligence" (AI) has repeatedly dominated the top searches on the screen and was named word of the Year for 2023 by British dictionary publisher Collins.

In addition to "artificial intelligence", we often hear about "machine learning", "deep learning"... What do these terms mean? What is the relationship between them? Follow the document to understand it ~~

Artificial Intelligence - Artificial Intelligence

When it comes to artificial intelligence, people's first reaction may be those robots with human intelligence in science fiction movies, but in fact, artificial intelligence is more than robots.

Artificial intelligence was coined by John McCarthy in 1956, when it was defined as "the science and engineering of building intelligent machines." The current artificial intelligence refers to "the research and development of theories, methods, technologies and application systems used to simulate, extend and expand human intelligence." It sounds a little loopy is it not, the document Jun to sum up, artificial intelligence is to let the machine can simulate the human thinking ability, so that the machine can feel, think and even make decisions like people. Today, artificial intelligence is no longer a simple subject, but involves the intersection of computer, psychology, linguistics, logic, philosophy and other disciplines. Artificial intelligence seems to be a profound technology, but it is actually a concept that covers a wide range. Around us, there have long been various artificial intelligence, such as: automatic driving, face recognition, intelligent robots, machine translation and so on.

In the face of a variety of artificial intelligence, we can divide artificial intelligence into three categories according to its strength:

Artificial Narrow Intelligence (ANI)

An AI that is good at something can only perform certain tasks. For example, the face recognition system can only recognize images, if you ask it what the weather will be tomorrow, it will not know how to answer.

Artificial General Intelligence (AGI)

Human-level AI is capable of exhibiting human-like intelligence in multiple domains, understanding, learning, and performing a variety of tasks. At present, strong artificial intelligence has not yet been realized and is still the long-term goal of artificial intelligence research.

Artificial Superintelligence (ASI)

Artificial intelligence, which surpasses human intelligence, is smarter than humans in every field, can perform any intellectual task and surpass humans in many ways. Although superartificial intelligence often appears in science fiction works, it is only a theoretical concept in practice and there is no possibility of realization at present.

Speaking of which, the document would like to ask you, what artificial intelligence does the AlphaGo that defeated the Go world champion belong to?

Machine Learning - Machine Learning

As mentioned earlier, the purpose of artificial intelligence is to enable machines to think and make decisions like humans, how to achieve it?

If you think about it, we are basically illiterate when we are born, and after decades of study, we have acquired all kinds of knowledge and skills. The machine is the same, to let it think, it is necessary to let it learn first, summarize the law from the experience, and then have a certain decision-making and discrimination ability, which is the core of artificial intelligence - machine learning.

Machine learning focuses on how computers simulate or realize human learning behaviors, acquire new knowledge and skills through learning, so as to reorganize the existing knowledge structure and constantly improve their own performance.

Machine learning is a multidisciplinary discipline, involving probability theory, statistics, approximation theory, algorithm complexity theory and other disciplines.

How do machines learn? Let's take a look at the learning process:

1. Attend class: Learn theoretical knowledge and input knowledge

2. Summary review: Through review, strengthen understanding

3. Sort out the knowledge framework: sort out knowledge and form a system

4. Homework: Practice to further deepen your understanding

5. Weekly quiz: Check your mastery

6. Check for gaps: Improve learning methods

7. Final exam: Check the final learning results

The machine learning process is similar and consists of the following 7 steps:

1. Data acquisition: Collect relevant data

2. Data processing: Transform data and unify data format

3. Model selection: Select the appropriate algorithm

4. Model training: use data to train models and optimize algorithms

5. Model evaluation: Evaluate the model performance according to the predicted results

6. Model adjustment: Adjust model parameters to optimize model performance

7. Model prediction: prediction of unknown result data

In short, machine learning is the automatic induction of logic or rules from data through algorithms, and based on the induction results and new data to make predictions.

For example, if we want the computer to know that a dog is a dog when it sees one, we need to show the computer a lot of pictures of dogs and tell it that this is a dog. After a lot of training, the computer will sum up certain rules, the next time it sees a dog, it will capture the corresponding features, and come to the conclusion "this is a dog." If the algorithm is not perfect enough, it may mistake a cat for a dog, which requires the computer to automatically improve the algorithm through empirical data, thus enhancing the predictive power.

According to the learning style, machine learning can be divided into the following four categories:

Supervised learning

Learning from labeled data, which contains independent and dependent variables, makes predictions by learning from known input and output data, such as classification tasks and regression tasks.

Classification tasks: Predict the categories to which the data belongs, such as spam detection, identifying animal and plant categories, etc.

Regression task: Predict data based on previously observed data, such as house price prediction, height weight prediction, etc.

Unsupervised learning

Analyze the data without labels, that is, there are only independent variables in the data and no dependent variables, and find the rules of the data, such as clustering, dimensionality reduction, etc.

Clustering: Bringing similar things together without focusing on what they are, such as customer groupings.

Dimensionality reduction: By extracting features, the high-dimensional data is compressed into a low-dimensional representation, such as combining the mileage and service life of a car into wear values.

Semi-supervised learning

Only part of the training data is labeled. Unsupervised learning is used to process the data, and then supervised learning is used to train and predict the model. For example, the mobile phone can recognize the same person's photo (unsupervised learning), when the same person's photo is tagged, the new photo of the person will be automatically tagged (supervised learning).

Reinforcement learning

By interacting with the environment, the algorithm is optimized based on rewards or penalties until the maximum reward is obtained, resulting in the optimal strategy. For example, after the sweeping robot hits an obstacle, it will optimize the cleaning path.


Deep Learning - Deep Learning

Through the above understanding, I believe that everyone is no stranger to machine learning. So what is deep learning? What does machine learning have to do with it?

Deep learning is a new research direction in the field of machine learning, which is an algorithm that learns and understands complex data through multi-layer neural networks. Machines learn complex tasks by learning deep representations of sample data, and can eventually have analytical learning capabilities like humans, able to recognize words, images, and sounds, among other things. Unlike traditional machine learning, deep learning uses a neural network structure, and the length of the neural network is called the "depth" of the model, so neural network-based learning is called "deep learning." The neural network simulates the neural network of the human brain, and the neuronal nodes can process and transform data. Through multi-layer neural networks, the features of the data can be continuously extracted and abstracted, so that the machine can better solve various problems.

There are four typical types of deep learning algorithms:

Convolutional Neural networks (CNNS) are commonly used for image recognition and classification tasks.

Recurrent Neural Network (RNN) : suitable for processing sequential data, such as natural language processing.

Long Short-Term Memory (LSTM) : A special RNN structure that is better able to handle long sequences of data.

Generative Adversarial Network (GAN) : Used to generate new data, such as images, audio, or text.

With the blessing of deep learning, artificial intelligence has developed rapidly, and I believe that in the near future, we will have a new era of AI.

Closing remarks

Useful knowledge has increased, the document to summarize it briefly:

"Artificial intelligence" is a broad concept that aims to make machines think and perform tasks like humans.

"Machine learning" is a method of implementing artificial intelligence that aims to learn patterns from data, whereas traditional machine learning requires human determination of data characteristics.

"Deep learning" is a specific branch of machine learning, based on neural networks, capable of automatically learning data features.

I believe that through today's study, you will no longer be silly and confused


* Special notice: The content is reproduced from other websites for the purpose of conveying more information than for profit, and does not endorse their views or confirm their descriptions, the content is for reference only. Copyright belongs to the original author, if there is infringement, please contact us to delete.