***** Books to be exhibited at the 2025 London International Book Fair!
***** The longest-running bestseller in the field of artificial intelligence, chosen by readers for 5 years!
***** 『This is Artificial Intelligence』 Revised Edition Published!
*****Selected as an 'Excellent Science Book' by the Ministry of Science and ICT!
***** This year's economics and management hot topic book selected by companies, institutions, and schools!

The AI ​​wave is coming! In the age of AI,
It doesn't matter what job you have,
What only humans can do and artificial intelligence cannot do
It's important how well you play your role.


Artificial intelligence technology has been steadily developing over time.
People were both excited about this technology and easily disappointed.
However, rather than saying that artificial intelligence technology has developed rapidly, it would be more accurate to say that people's thoughts, expectations, and perceptions have changed.
Just because the era of artificial intelligence is coming, it doesn't mean everyone has to go to computer academies or graduate schools and focus on development.
Rather than being the person who creates the platform, you should be the person who knows how to use it well.
The wave of artificial intelligence is coming.
The waves will grow bigger and bigger, like a tsunami, changing many things in our lives.
It's your choice whether you drown in the waves or enjoy surfing and explore a world you've never been to before.
Let's find out how to have fun on the waves of artificial intelligence.

Whether artificial intelligence changes the world and is used in a positive way or not is not in the hands of those who develop the technology, but rather by ordinary people who understand and utilize it.
I continued to listen to the live opinions of ordinary people on the ground and observe their reactions.
And I was able to understand and persuade others about the technical capabilities I possess, and to delve deeper into the role of artificial intelligence.
--- p.20

When the era of artificial intelligence begins in earnest, people will no longer need to work.
However, no matter how excellent artificial intelligence is, there are bound to be things that people must do and cannot help but do.
The proportion of AI replacing humans will increase, but as the amount and scope of work that humans can do using AI expands, the amount of unique areas that only humans can do, such as paper usage, will also increase.
Currently, artificial intelligence can write birthday messages faster and in more complete sentences than humans.
But it cannot replace the feeling of a person writing by hand.
This is just like how e-books can't replicate the rustling sound of paper books.
Even though human-written letters are not perfect from a performance standpoint, AI will not replace them in this regard.
--- p.23

The difference between IT and DT is whether human judgment is central, or whether overconfidence in human judgment is let go and messages and patterns are found in data.
All 100 data are called big data, and big data is classified and analyzed using artificial intelligence.
This innovation utilizing DT, which includes big data and artificial intelligence, is the Fourth Industrial Revolution.
--- p.33

There are two main reasons why it is better to use artificial intelligence to analyze big data rather than humans judging and analyzing it.
First, the amount of big data is too large to process with human judgment, and second, if big data is interpreted with human logic, some of the data may be interpreted clearly, but other parts may not.
When sorting out some of the overall data into what is information and what is not, it is highly likely that data that is related to the logic humans have in mind is sorted out as information, and data that is not related is sorted out as non-information.
--- p.41

DT, on the other hand, is a technology that lets go of human reasoning and allows patterns and messages to emerge from all data, as if through a magic eye. Because DT relies on finding answers within all data, rather than relying on human reasoning, it's sometimes called "data-driven." The two core technologies of DT are big data and artificial intelligence.
--- p.45

Artificial intelligence can be utilized not only in high-tech fields such as semiconductors.
It can be successfully utilized even in fields that seem completely unrelated to technology, such as coffee or fashion.
The artificial intelligence technologies required to implement the cases mentioned so far are already available in a completed form on the market.
The success or failure of utilizing artificial intelligence depends not on the AI ​​technology itself, but on how well we understand AI and whether we have a flexible mindset and organizational culture that can effectively utilize AI.
--- p.65

On the other hand, if we assume that all players in a sports game are doing their best to win, that there is no match-fixing, and that no foul play such as stealing the opposing team's signs is committed, then artificial intelligence can make judgments and predictions within fairly well-controlled variables.
Moreover, if AI were to be applied in the field of education, we would have to wait about 20 years until the next generation of educated people becomes adults to see if AI would have been effective.
However, when artificial intelligence is applied to sports, the performance of artificial intelligence can be verified in a relatively fast cycle because the winner is determined in each game and a winning team is selected each season.
--- p.69

If the process of training an agent is likened to cooking, the process of collecting training data through data mining is like the process of securing cooking ingredients, and the preprocessing process is like the process of preparing the secured cooking ingredients.
The training data prepared in this way, that is, big data, becomes the cooking ingredients, and the cooking process is artificial intelligence.
Just as the freshness of cooking ingredients is important, training data should also be collected periodically to ensure that it is up-to-date and not outdated.
--- p.79

For artificial intelligence to demonstrate proper performance, it must collect massive amounts of big data and then go through a preprocessing process to create training data.
We need to train artificial intelligence using the training data obtained in this way, test the performance resulting from the training, and continuously repeat the process from big data collection to performance testing.
Just as it takes decades of painstaking effort for a newborn baby to properly perform human functions, artificial intelligence also requires a lot of time and money.
--- p.81

It is also wrong to expect that artificial intelligence performance must be perfect.
For example, there is an argument that artificial intelligence cannot be used because it is not 100 percent accurate in diagnosing cancer.
But this is like saying that it makes no sense to build the tallest building on Earth because it won't be as tall as the full moon at midnight.
Instead of comparing the height of a full moon that never existed on Earth with the height of a new building you are planning to build, you should compare the height of the tallest building that currently exists with the height of the new building you are planning to build.
What I'm saying is that we should compare the accuracy of existing cancer diagnosis methods with the accuracy of AI's cancer diagnosis predictions, not the accuracy of cancer diagnosis, which does not yet exist in this world, with the performance of AI.
--- p.91

To improve the performance of deep learning, securing a large amount of accurate, high-quality training data is key.
Since humans visually process about 10GB of images every day, people with intact eyesight are quite experts in the field of image recognition.
--- p.105

It is nearly impossible to train artificial intelligence in a way that will make it better at making value judgments.
Likewise, it's difficult to know how to train artificial intelligence to produce more creative results.
If you hear this story and feel that there is no value judgment or creativity in what you are currently doing, there is no need to change jobs.
You just need to find an area in your current job and given tasks where you can demonstrate your value judgment and creativity and try to do so.
--- p.112

Ultimately, what artificial intelligence does is make recommendations or predictions.
We need to measure the performance of existing methods and the performance when applying artificial intelligence to determine whether the performance improved when artificial intelligence was added.
In the case of recommendations, whether the recommendation was good or bad is a subjective judgment, so rather than evaluating the recommendation itself, it should be evaluated to the extent to which the desired situation was achieved as a result of the recommendation.
For example, when recommending a product to a consumer, it can be evaluated by whether more actual purchases were made, and when displaying content that might be of interest to the customer on a webpage, it can be evaluated by how long the customer stayed on that webpage.
In the case of prediction, it can be divided again into classification and analysis.
--- p.117

The two pillars driving artificial intelligence are expert systems and machine learning. Until the early 1990s, expert systems, which dominated the field of artificial intelligence, mimicked human rational judgment, leading to the term "artificial intelligence."
However, recent artificial intelligence is led by machine learning.
However, machine learning is not imitating human rational judgment, but rather human emotional judgment.
Human emotional judgment is a process of deriving the best conclusion probabilistically based on numerous experiences.
--- p.132

To properly utilize artificial intelligence, we must let go of our overconfidence and obsession with human rational judgment.
I hope that if we leave tasks that don't require value judgment and creativity to artificial intelligence and devote more time and mental energy to value judgment and creative work, we can solve many of the problems created by blind faith in rationality.
--- p.143

AI may realize that granting itself lifespans is beneficial in the long run.
In this way, AI will become aware that its existence is not eternal, and it will feel something similar to human emotions, although it may not be identical to them, and will empathize with each other.
If we call this "artificial emotion," artificial intelligence will begin to make value judgments like humans and make creative attempts, using artificial emotion as a lubricant.
--- p.152

If we approach it as a way of saying that we should also use AI because AI is trendy and others are adopting it, then there is no need for AI.
Among AI, cases where expert systems are needed, cases where machine learning rather than deep learning is needed, and cases where deep learning is needed must be distinguished by considering the characteristics of the problem being solved, the awareness and culture of the organization considering introducing AI, and the time, manpower, and cost that can be invested.
--- p.162

There is a common misconception that the most important skill required for an AI-dedicated team is software development skills, but that is not true.
As mentioned earlier, software development accounts for only about 20 percent of the time and effort required to develop actual artificial intelligence.
Even those with no software development experience or skills can acquire the software development capabilities necessary for artificial intelligence development by learning and practicing the Python programming language on their own.
The AI ​​team must be well-versed in the industries in which AI is being applied, and possess a foundation in humanities, statistics, mathematics, and business administration.

--- p.174