Is it difficult to create AI?
What do you think of when you hear the term AI (artificial intelligence)?
Reversi (Othello), chess, shogi, Tetris, card games and fighting games, interactive AI such as chat, translation tools, etc. AI is now active everywhere in our lives
Finally, it has become possible to easily create AI without specialized knowledge of programming and the workings of artificial intelligence.
Advantages of creating AI
AI has become a familiar part of our work and lives, but rather than unknowingly receiving the benefits of AI, people will be able to live better lives if they are actively involved with AI.
By trying to operate and experience what current AI can do and to what extent, you will be able to see the appearance of AI that you only vaguely knew.
What you need before creating AI
Anyone with an internet connection and a device that can operate a computer can make a simple AI experience.
The key is curiosity and needs and ideas about what you want AI to do. Of course, in order to move from the AI ​​experience side to the AI ​​provider side, more specialized knowledge is required.
But first, let’s get a rough idea of ​​how the current AI is made.
5 ways to make AI
AI is good at repeating similar processes, but it has evolved to be able to make complex decisions through learning.Now, let me briefly introduce how such an AI is created and how it is created.
How to make AI 1: Data collection
AI cannot make any judgments about the input data in a blank state.
In order to know the judgment criteria or to learn the judgment criteria by themselves, a huge amount of data is required.Currently, the data itself is provided free of charge for research and training, and there are also services that automatically collect data.
How to make AI 2: Model making
It is no exaggeration to say that the stage of designing a model is the heart of artificial intelligence.Think of an AI model as something like an AI’s “way of thinking”.
We set “rules of thinking” in the computer, such as how the computer processes and labels the collected data.
There are various models designed by AI experts, and tools have been released so that even non-experts can easily fine-tune the models.
How to make AI 3: Embedded
It is necessary to incorporate it into the system according to the scene where AI is used.This step requires some programming knowledge, even at a beginner level.
AI can be incorporated into software for each OS and smartphone apps so that AI responds to user input, or it can be incorporated into web apps for universal access.
How to make AI 4: Operation check
Check model accuracy.We will check the operation of AI to see if it processes one of the prepared data or data similar to it and produces results as expected.
The system as a whole also checks whether bugs or errors occur.
How to make AI 5: Implementation
Once the AI ​​is complete, it’s finally time to implement and operate it.
Custom-made AI can be completed by fitting and implementing it for each situation, such as publishing it on the Internet in the form of a web application as described above, providing it in combination with other services, or using it in internal business. increase.
4 recommended tools for creating AI
Today, in order to create AI models, there are services that support the creation of AI models without having to prepare an environment that can run a programming language on your own or to introduce a powerful machine to process a large amount of data. increase.
Here, we would like to introduce four recommended tools, from tools that allow you to easily experience AI to full-fledged tools that can be applied to your business.
Recommended tool for AI creation 1: AI Maker
It is a very casual tool that allows you to create AI from your smartphone just by linking with your Twitter account.By entering tags as labels to be recognized, the tool automatically collects image data to be used for learning, making it easy to perform machine learning for image classification.
In addition, by uploading various data, services that perform text analysis, transcription from speech, etc. from trained models are provided free of charge.
Recommended tool for AI creation 2: Neural Network Console (SONY)
Neural network models can be designed and optimized mainly by drag-and-drop operations. You can learn about deep learning using sample projects. Examples include image recognition, image generation, and image segmentation.
In addition to image recognition, models such as numerical estimation based on vector data and classification of sensor data are applied to actual products.You can try it for free for 10 hours.
Recommended tool for AI creation 3: Cloud AutoML Vision (Google)
Train machine learning models on images, videos, text, or CSV (tabular data). An example is the classification of thousands of images.
AutoML Vision detects and extracts not only information about what is in the image, but also the text contained in the image.A free trial is available for a usage fee of US$300.
Recommended tool for AI creation 4: IBM SPSS Modeler (IBM)
SPSS Modeler is good at extracting concepts, themes, emotions, trends, etc. from CSV data and text data such as various customer information, customer feedback, emails, blogs, various SNS, etc., and making it useful for customer analysis and demand forecasting. It looks like
It also suggests analyzing geographic data, such as postal codes and addresses.
You can try it for free for 30 days after registration, but university officials can use the academic version for free by registering an email address issued by the university.
8 recommended books for making AI
You can easily create AI by using external tools, but you will want to deepen your knowledge about AI.
If you want to create AI from scratch, want to learn more about the unknown mechanism of AI, or want to explore the possibilities of AI, here are 8 books that will meet your needs.
Recommended book for creating AI 1: Machine learning starting with Python — Learn the basics of feature engineering and machine learning with scikit-learn
This is a text published by O’Reilly, which is famous for its animal covers. It can be said that it is a must-have for students. It is a textbook book recommended first for beginners who want to learn about AI in earnest. scikit-learn is one of the main libraries used to practice machine learning in Python.
What is the concept of machine learning? This is a must-read book for comprehensively learning learning models from the basics.
Recommended book for creating AI 2: Deep learning from scratch—Theory and implementation of deep learning in Python
Another O’Reilly publication, this is a must-have textbook for students that explains the theory and foundations of DeepLearning. Deep learning is often contrasted with machine learning.
Just as the above “Machine Learning with Python” is structured so that you can learn about machine learning by practicing programming, this book also explains what kind of mechanism and idea it consists of by repeating the execution of sample code. Designed for learning by doing.
This book is intended to help readers “feel” the principles of deep learning.
Recommended book for making AI 3: Details! Python 3 introductory notes
This is a reliable book for beginners who want to use Python3 by actually building an environment where the program runs on their own machine.
In this one book, we will follow the knowledge that will be a stepping stone from beginner to intermediate level of Python. O’Reilly also has a book called “Introduction” to Python, but this book carefully explains the places where programming beginners tend to stumble.
This is a recommended book for those who don’t know much about programming, but want to write Python code in earnest.
Recommended book for creating AI 4: Machine learning super introductory with Excel
Excel is a spreadsheet program that is installed on most computers.If you have Excel at hand, you can learn about how machine learning works by following this book.
Machine learning is designed based on various mathematical thinking and statistical methods, but it is very painstaking to learn the principles of machine learning based on the formula with only paper and pen.
However, this book makes it easier to think about the mathematical aspects of machine learning by using the functions built into Excel.
It can be said that it is a good book to get an overview of AI without learning a programming language or preparing a development environment.
Recommended book for building AI 5: Let Python do the boring stuff
This is a supplementary reading book published by O’Reilly, highly recommended for programming beginners in society.
For people who use Microsoft Office in their work but have never touched programming, they devote a large number of pages to carefully and politely suggest the introduction of Python into their daily work. increase.
Ideas for automating daily computer work are explained in detail in a tutorial format so that even those who are not programmers can easily understand them.
This book takes you from zero prior knowledge of programming to being able to run Python and apply automation to tasks similar to the many examples in this book.
Recommended book for making AI 6: Ready to use! You can practice in business! How to create AI, machine learning, and deep learning applications with Python
This book is recommended for those who are familiar with Python programming to some extent and who want to use the machine learning and deep learning they have just started to learn in the real world.
I’ve tried a little bit of Python, but I want to increase the drawer of how to make AI, and I want to know how it can be applied to real life, and for those who want to know examples, I will show you the libraries and sample code to use. , It is one book that lists abundantly.
Recommended book for creating AI 7: Learning by playing with Python and real data Data analysis course
This book is also suitable for those who have some familiarity with Python programming.Recommended for intermediate and advanced learners who want to learn more about AI and machine learning, and want to learn more about how to create AI using real-world data.
The feature of this book is that it uses data obtained in the real world (actual data), such as “data provided by various companies that can be freely obtained by API” and “statistical data provided by the government”. The point is that
As a result, data analysis operations, which can be complicated and troublesome, can be learned while feeling more familiar and realistic.
Since this book focuses on data analysis, which is the basis of AI, it mentions not only Python but also the R language used in statistical software.
Recommended book for making AI 8: How to make artificial intelligence
This book is a little different from the books we have introduced so far.
With Game AI as the axis, various AI technologies applied to games and a wide variety of knowledge related to game AI, it is a book that gives you a bird’s eye view of how to create a virtual and real world view. I’m here.
It can be said that game AI technology, which builds AI that gives a sense of personality, such as game characters, is one of the fields that the general public can easily associate with the word artificial intelligence.
4 learning methods for AI creation
There are AI that can be easily interacted with through play, and there are also full-fledged AI tools that can be expected to be effective immediately in business, and technical terms such as Python and machine learning have appeared in the introduction of the book.
For those who want to learn more about how to create such a deep AI, I will introduce how to learn AI.
AI-making learning method 1: Learning from books
These days, there are a lot of teaching materials available online, but nothing beats the teaching materials that are organized as books.If you have the motivation, anyone can self-study how to make AI and programming using the various reference books and supplementary books introduced above as texts.
Also, even if you become a professional, you will keep the reference at hand and read it again and again.
AI-making learning method 2: Make use of the free trial
Let’s learn through the free trial of the public tools introduced above, and participate in free trial courses such as programming schools.
There is no charge, so you can easily start learning.If you are wondering whether to start learning how to make AI, please take advantage of the opportunity to experience it for free and take a peek into the world of AI.New encounters await.
AI-making learning method 3: Attend a school
Although there is a fee, studying at a school that guarantees a certain level can still be expected in terms of learning achievement and learning speed. It will be easy to maintain the pace and motivation of learning by being able to ask questions to experts immediately and taking constant courses.
The school has a schedule and curriculum for learning about how to make AI, so it can be said that it is a notable advantage that you can learn efficiently without hesitation and with the instructor taking the lead.
AI-making learning method 4: Using blogs as a reference
If you come across technical terms that you don’t understand in the learning process of how to make AI, search the internet. Related articles will be hit. If you explain from the standpoint of other learners or experts, you may gain a deeper understanding or come across techniques you didn’t know.
It is a good idea to follow blogs that regularly post easy-to-read articles on how to make AI.
Learn and practice how to create AI
The possibilities of AI will continue to develop in the future. So far, I have introduced various things such as what AI can do and what it can do for you, but sometimes it is difficult to realize it unless you actually make it yourself and use it.
