This is all you need to remember about how AI works.

AI is a complex combination of technologies, but if you want to understand how it works, at a minimum, you need to understand the meaning of the following to get a rough overall picture.

1. Data
2. Algorithm (technology)
3. Models

Data (training data)

Data in an AI system is the primary source for making discriminative judgments about data using algorithms.

Training data for AI systems must be of high quality, selected for the purpose, and suitable for the algorithm used, and training data has a significant impact on AI quality and other factors. The amount of training data also has a significant impact on the model to be generated, so as much data as possible should be available.

In this sense, data is one of the most important factors when building an AI system.

Algorithms

An algorithm is a set of rules used by a computer to make predictions and decisions as it learns and trains from data.

Algorithms can be those published in academic papers, those developed independently by companies, those released as open source, or those that are proprietary improvements on publicly available algorithms.

Since algorithms are not suitable for all purposes and for all training data, it is important to select the best algorithm for the purpose and the data, and use it in combination.

Models

A model is a mathematical representation of data learned using an algorithm. When AI actually operates, it generates results from input data based on the learned model.

A model is not something that is created once and never ends; it is a part that is constantly being improved by adding training data to gain new knowledge, training to adjust parameters, and so on.

Brief process to build an AI

To simplify the process of building an AI, the flow is as follows

1. select data suitable for the purpose
2. train the data with an algorithm to generate a model Develop an application or system that incorporates the AI model

Although simple in words, selecting the most appropriate data and algorithm for the purpose is extremely difficult and costly, and is a point of differentiation for each service.

Frequently asked questions about how AI works

What is the difference between "learning" and "training" in AI?

Although the words "learning" and "training" are similar, there is actually a difference.

Learning" in AI refers to the process of using algorithms to extract knowledge and patterns from given data and build models. During the learning process, the parameters of the model are adjusted using the data set, and patterns and relationships from the data are incorporated into the model.

Training" in AI, on the other hand, refers to the process of optimizing the constructed model against specific data. In the training process, a trained model is applied to real data, performing tasks such as prediction and classification while the algorithm adjusts parameters to make predictions and decisions.

What is the difference between an algorithm and a model in AI?

An algorithm in AI is a set of rules about how to learn from data. For example, machine learning algorithms include supervised learning, unsupervised learning, and reinforcement learning, and the results will vary depending on which algorithm is used.

A model in AI, on the other hand, is a set of results or parameters from which an AI is learned using an algorithm. In other words, it could be compared to "knowledge and experience" in humans.