Executive Director Saying, "Let's Make Something with That AI Too" [AI Error Notes]

⑥Easy Words Like "Just Try It" with AI
Various Explanations, Definitions, and Expectations of AI
Practical and Specific Framework: 'Prediction Machine'

Editor's NoteExamining failures is the shortcut to success. 'AI Error Notes' explores failure cases related to AI products, services, companies, and individuals.

Everywhere, it's all about 'AI'. Office workers are struggling too. A meme capturing this situation was created and resonated strongly online. Online community screenshot

These days, you might be feeling AI fatigue. No matter what product or service advertisement you see, it has ‘AI’ attached to it. Washing machines have AI, robot vacuum cleaners have AI, large supermarkets use ‘AI for price optimization,’ and stocks to invest in are recommended by AI, and so on. Encounters with the word AI in daily life are increasing more and more.

However, rather than understanding AI better, many people feel even more confused. Are the world’s top AI companies explaining it more clearly?

Various Explanations of Artificial Intelligence

▶Microsoft: AI is the capability of computer systems to mimic human-like cognitive functions such as learning and problem-solving.

▶Google Cloud: Artificial intelligence is a set of technologies that enable computers to perform various advanced functions, including recognizing, understanding, and translating spoken and written language, analyzing data, and making recommendations.

▶Amazon (AWS): AI is a technology with human-like problem-solving abilities. Actual AI appears to simulate human intelligence by recognizing images, writing poetry, and making data-driven predictions.

▶IBM: AI is a technology that allows computers and machines to simulate human intelligence and problem-solving abilities.

▶Samsung Electronics: Artificial intelligence is a field of computer science that studies how to apply intelligent human behavior to machine systems.

▶SK Hynix: Artificial intelligence is a computer algorithm designed for a specific purpose that automatically processes tasks based on given inputs.

Do you have a sense of what AI is now? Probably not; it’s still difficult. That’s understandable because AI itself was born from the convergence of various academic fields. It is the result of combining philosophy, psychology, linguistics, computer science, and more.

Since experts from each field think about AI from their own perspectives, various definitions inevitably arise. Also, because technology advances rapidly, the definitions and concepts of AI evolve accordingly.

So, should we give up on an intuitive understanding of AI? No. When looking at any object or phenomenon, you should not waste time trying to find the one and only accurate and truthful perspective. Since AI is a developing field, it is more reasonable to see that there is no single unchanging definition.

What matters is how to utilize AI and what benefits to gain from it. Set your own perspective and framework regarding AI, and view AI accordingly. This way, you can avoid getting trapped in abstract concepts, definitions, or philosophical quagmires. The important thing is what we will do with AI and what we can do with it.

The More You Know, the More Puzzling... A Useful Frame to View AI: 'Prediction Machine'

The original cover of the book "The Prediction Machine," co-authored by Ajay Agrawal, Joshua Gans, and Avi Goldfarb (out of print domestically). In this book, the authors argue that "Artificial intelligence is a prediction machine."

In that regard, if I were to pick a useful frame to view AI, it would be the frame that ‘AI is a prediction machine.’ Joshua Gans, Avi Goldfarb, and others from the Rotman School of Management at the University of Toronto wrote this in their 2018 book Prediction Machines:

"What the new wave of AI brings is not intelligence itself but prediction, an important element of intelligence."

If you are wondering what to do with AI or what AI can do, think of ‘prediction’ right away. Ask yourself, ‘What can I predict in my current work?’ or ‘What predictions would be beneficial for our company?’ Let’s look at real examples. The business world runs on countless predictions.

Uniqlo predicts seasonal changes and regional trends to adjust clothing inventory. The more accurate the prediction, the more inventory costs can be reduced.

Amazon predicts the next purchase product based on consumers’ previous purchase data. Through predictions called recommended products, Amazon maximizes sales.

Banks can predict the repayment ability of loan applicants. By analyzing credit scores and past transaction records, they assess credit risk and set appropriate margin rates.

Insurance companies can increase premium margins if they accurately predict subscribers’ disease occurrence rates.

Farmers can predict harvest yields by forecasting weather and soil conditions, thereby reducing logistics burdens.

Entertainment companies like Netflix make accurate recommendations (predictions) based on various viewer data, encouraging consumption.

Media companies can also rely on the power of prediction. If they successfully predict news that readers will like and want to read by analyzing reader data, they can increase traffic.

In other words, prediction forms the basis of decision-making in all industries. How predictions are made affects responses and results. It enables satisfying consumers, reducing costs, and increasing sales. Prediction ability is competitiveness.

The Power of AI: Lowering the Cost of Prediction

But wait. Prediction is not something only AI can do. Is there anyone who lives a day without prediction? Every office worker makes some predictions in their work, and companies already operate based on predictions. No one makes decisions without looking at past data. Prediction is not a feature unique to AI but a universal human ability. So why is AI called a prediction machine? Why can we call AI a prediction machine?

Both humans and AI make predictions. However, AI’s predictive ability overwhelms humans in scale and complexity. Humans predict based on limited data, experience, memory, and intuition. They are weak at handling large amounts of data and recognizing complex patterns. Even when decision variables are three or two, it becomes very complicated.

AI is different. It can process large-scale data with dozens of variables and detect complex correlations and patterns that human intuition cannot even perceive.

Also, humans are accustomed to linear thinking. Even the article you are reading proceeds linearly from the first letter to the last period. We are used to thinking like ‘If A increases, B decreases.’ However, complex reality is nonlinear, and the interrelationships among variables are multidimensional.

AI’s learning methods (deep learning, machine learning, etc.) help discover nonlinear relationships and hidden patterns through algorithms.

If you ask a person to recommend products, they will generally predict by synthesizing understandable variables such as age, gender, and income level.

However, AI mobilizes all accumulated information such as purchase history, click history (even if not purchased), time spent on pages, device types, and country-specific consumption trends to calculate correlations and produce predictions. Moreover, it does this complex process instantly. Automation is also possible. This is where the decisive power lies.

The Power of AI: Improving Decision Quality

An illustration depicting linear thinking and systems thinking. Screenshot from the U.S. National Institute of Standards and Technology (NIST) website.

Automation means reducing the cost of prediction. AI prediction facilities for data analysis and computation do not require investment in physical assets. Although data centers are needed, they are nothing compared to manufacturing plants.

Prediction is an expensive commodity. In the past, making precise predictions required numerous experts, heavy software, and long hours. But AI delivers high-quality predictions with much less time and cost. In other words, AI makes prediction affordable. When prediction becomes affordable, anyone can purchase predictions. When high-quality predictions are supplied cheaply, decision quality improves.

Let’s go back to the beginning and reconsider. ‘Everyone around me keeps chanting AI, AI, AI, but what exactly is AI?’ Yes, that’s right. AI is a prediction machine. Now everything becomes clear. You can treat AI-related information that does not help you and is cited just to promote products as noise.

AI is a prediction machine. What predictions are important in your work? What predictions are important in your business right now?