Is there a way to easily create your own AI?

To answer these questions and requests, we will explain the steps to create AI and introduce APIs and services that can be used free of charge.

Specifically, we are talking in the following order:

  • 3 steps to create your own AI (artificial intelligence)
  • What you need to create your own AI (artificial intelligence)
  • APIs and services for easily creating your own AI (artificial intelligence)

This article is aimed at beginners who want to create their own AI (artificial intelligence), so please read it.

Table of Contents

  • 3 steps to create your own AI (artificial intelligence)
    • Collect data
    • Create trained model
    • Incorporate the created model into the service
  • What you need to create your own AI (artificial intelligence)
    • Large amounts of data (big data)
    • Programming skills
    • Knowledge of mathematics
  • Free APIs and services for easily creating AI (artificial intelligence)
    • Watson APIs
    • A3RT
    • AI maker
  • Kredo if you want to develop AI (artificial intelligence)
  • Summary: Creating your own AI (artificial intelligence) from scratch is difficult

3 steps to create your own AI (artificial intelligence)

There are three major steps to creating your own AI.

Collect data

AI finds regularity and makes judgments based on large amounts of data. A large amount of data is required for AI to learn.

Companies often make AI learn using big data, but what should individuals do?

We recommend using the provided service (dataset) depending on the type of AI you want to create.

For image recognition, you can also use Google image search. For chatbots, etc., Japanese datasets are provided in various places.

For example, there are datasets from the National Institute of Informatics.

Collecting data is very important in creating AI, but it is difficult to collect a sufficient amount of data.

It’s relatively easy to collect using various services, but be aware that it does require some effort.

Create trained model

After preparing the data, let AI learn using machine learning. Machine learning includes “supervised learning” and “unsupervised learning”.

Supervised learning is learning in which humans act as teachers and give input data and correct data to AI, and learn based on that data.

Unsupervised learning is learning in which only input data is given to AI, and AI classifies it into clusters with common terms and finds frequent patterns.

These learning methods should be used to improve the accuracy of the output results for the input data. As for the learning model, in addition to the method of programming directly by you, there is also the method of using the model provided by each company.

In addition, there are cases where pre-trained models are provided, so search for available learning models depending on the type of AI you want to create.

Later in this article, I will introduce some free learning model APIs and services.

Incorporate the created model into the service

AI cannot be used by itself.

The trained model automatically outputs the input information, and it must be incorporated into the service for it to be input by the user.

When integrating into a service, you must be clear on which platform you want it to run.

For example, if you want to publish it on the web, embed it in a web service (web application). If you provide it as a smartphone application, it will be incorporated into the application.

When incorporating a model created by an individual into a service, it is recommended to incorporate it into a Web service.

This is because web services are easier to implement than other platforms.

Normally, when publishing as a web service, it is necessary to prepare a server, but recently it is also possible to publish serverless using public clouds.

What you need to create your own AI (artificial intelligence)

What you need to create your own AI (artificial intelligence) We will introduce the things and skills necessary to create your own AI.

What you need to introduce here is the bare minimum.

Large amounts of data (big data)

As I explained earlier, a large amount of data is required to create AI.

To collect data, it is recommended to use public datasets.

I will introduce some sites that publish datasets, but most of them are overseas sites.

  • National Institute of Informatics

We provide one of the few Japanese datasets. Although some datasets are available to the general public, they are mainly datasets for researchers.

  • Amazon AWS Public Dataset

It is a public dataset that anyone can use for free. Crawl data for over 5 billion websites and more.

  • Kaggle

Various companies and research institutes publish data. It is a platform where data scientists from all over the world compete.

  • MOST

It is a dataset for image recognition and is so famous that it is said that everyone uses it for machine learning. The size of the data is also small, making it a relatively easy-to-handle dataset.

  • Megaforce

A dataset for face recognition algorithms. There is data on the face images of over 670,000 people.

If you look for other things, there are various things, so please look for them.

Programming skills

Programming skills are essential for creating AI learning models.

For AI programming, I recommend Python.

Python is easy to use even for beginners, and there are plenty of libraries and frameworks for AI programming. It can be said that it is an easy language to learn because there are many commentary books and commentary sites.

AI frameworks that can be used in Python include “Keras” and “TensorFlow”.

If you want to know about AI frameworks and libraries, please refer here.

Knowledge of mathematics

To understand AI algorithms such as machine learning and deep learning, high school to college graduate-level mathematics knowledge is required.

Specifically, the following mathematics knowledge is required.

Mathematics knowledge necessary for self-made AI
  1. Vector
  2. Queue
  3. Differential
  4. Linear algebra
  5. Probability
  6. Statistics


By using a Python framework, etc., it is possible to build an AI without mathematical knowledge, but it is essential knowledge for a fundamental understanding.

Books that focus on the mathematical knowledge required for AI programming are also on sale, so it would be a good idea to study there.

Free APIs and services for easily creating AI (artificial intelligence)

It is difficult to create an AI learning model from scratch. The content of programming is also advanced, and it can be expected that many people will be frustrated.

Therefore, we will introduce APIs and services that can use already created learning models for free.

Please try it once before you start AI programming. is a service that allows you to create your own AI for language and voice input based on natural language processing.

It can be used in areas such as chatbots and home automation. can create AI learning models with almost no programming.

Account registration with a GitHub account or Facebook account is required.

We will implement it by registering a new application, creating a conversation story to be developed in the application, and creating a conversation.

If you are thinking of creating your language/speech processing AI, why not give it a try?

Watson APIs

Watson API is a collection of multiple API services available from IBM Cloud.

There are “Watson Assistant” for chatbot (query response system) creation, “Visual Recognition” for image recognition, “Speech to Text” for voice recognition, etc.

In addition, we provide 16 APIs and services. (As of October 7, 2019)

IBM Cloud is a public cloud provided by IBM and is a pay-as-you-go service.

The Watson API is available with a free lite account, so give it a try.


A3RT is a group of API services released free of charge by Recruit Technologies Co., Ltd.

There are “Talk API” for chatbot creation, “Proofreading API” for grammar proofreading, and “Image Search API” for mutually searching images and text.

As of October 7, 2019, we provide 11 APIs.

The API key for using the API can be obtained by registering your email address. Demos are available for some APIs, so you might want to get an API key and try them out.

AI maker

AI Maker is a service that allows anyone to easily create AI.

As of October 7, 2019, AI for image recognition and transcription can be created.

Image recognition automatically learns from image data that is automatically collected through labeling.

You can create simple AI without programming, so it’s good to use when you want to learn an overview of how to create your own AI.

Kredo if you want to develop AI (artificial intelligence)

Kredo if you really want to develop AI (artificial intelligence) By using the APIs and services introduced in this article, you can create your own AI relatively easily.

However, if you want to fine-tune the details or develop AI as a job, you have to create your own from scratch.

To create AI from scratch, programming language skills such as Python are indispensable.

In addition, although the technology in the AI ​​field continues to develop in Japan, even now, much of the information is centered on overseas sources.

From now on, if you want to become an AI engineer or AI programmer, you can say that English skills are essential along with programming skills.

Would you like to learn together with Kredo, where you can learn English for working abroad, along with programming skills to become an AI engineer or AI programmer from inexperienced?

Summary: Creating your own AI (artificial intelligence) from scratch is difficult

Summary: Creating your own AI (artificial intelligence) from scratch is difficult To create your own AI, you need three major steps: “collect data”, “create a trained model”, and “incorporate it into a service”.

The things you need to create your own AI are “large amounts of data”, “programming skills”, and “mathematics knowledge”.

Even if you don’t have programming skills, you can create your own AI by using APIs and services that are open to the public for free.

However, programming skills are indispensable if you want to create original AI.

First of all, it is a good idea to get an overview of AI through free and open APIs and services and acquire programming skills to create original AI.