Generative AI, game AI… Is evolving AI a friend to humans or a threat?

In the era of interactive AI, the boundary between trust and utility has become a key issue

Recently, the performance of AI has improved dramatically, bringing about changes in the way we gather information. Previously, we would use search engines like Google for some information, but now, more and more people are turning to generative AI like ChatGPT.

But it is a mistake to assume what AI says is always right. Large language models (LLMs) such as ChatGPT are not omniscient. They are merely adept at reconstructing average discourse in almost all fields. That makes them unsuitable for fact-checking, complex reasoning, or creating novel ideas. Currently, LLMs have difficulty checking specific facts or solving problems that require multi-step reasoning owing to their structure.

The evolution of generative AI is also having a major impact on education. The trouble for us educators is that the generated sentences are too natural and fluent. It used to be easy to recognize when students copied from online sources, but with AI tools, it is becoming harder to tell whether their work is original. We often struggle to know whether our students genuinely understand what they study.

In the U.S., fake news generated by AI has gone viral, becoming a social problem. There was a case where a public agency issued a report, carelessly citing a reference that did not exist.

Generative AI constructs responses in the form of returning the most typical answer to a given question. It cannot judge whether what it says is true or false. In other words, it merely follows the general flow of conversation in the world and does not speak with belief as humans would.

We humans seem to be susceptible to natural and persuasive speech, and this may be one reason why we uncritically accept the answers generative AI provides fluently. We need to be careful not to confuse fluency with reliability.

Of course, the benefits of AI are huge. For example, AI can now give fairly accurate advice on what is generally considered the standard or conventional way to do something. AI has great potential to help us make small decisions in our daily lives and at work.

Recently, a new form of application called AI agent is also attracting attention. This is a system that can perform advanced tasks, such as arranging flights and hotels based on a user’s travel plans or finding and summarizing materials according to a user’s requests. AI is now able to break down users’ work into smaller tasks, carry out each one on their behalf, and support the overall workflow.

AI is no longer just a search and assistance tool; it is deeply penetrating into various aspects of human society. At the same time, rather than easily trusting the answers AI provides, we increasingly need to understand its mechanisms and limitations and use it appropriately.

Behind the evolution of game AI lies a dramatic increase in exploration capabilities

The evolution of AI is also bringing big changes to the gaming sector. Video games used to be structured so that a story or scenario was prepared in advance, and players would follow it. Recently, however, there has been an increasing number of what is called open-world games, in which players are free to explore a virtual world and weave a story of their own.

AI is pervasive in competitive games, too. In live TV broadcasts of Shogi (Japanese chess), AI judges the state of a match and points out mistakes in real time, and viewers can understand the flow of the match easily no matter when they start watching. This has lowered the barrier to watching Shogi, expanding a new fan base known as mirusho (Shogi fans who only watch matches).

When discussing the development of game AI, we should mention AlphaGo, the game AI playing Go that emerged in 2016. When it first appeared, it became a hot topic on the internet with people saying, “There is an unidentified strong player.” The powerful AI boasted an undefeated record against dozens of strong Go players, including professionals, and when it turned out to be AI, many people in the world were astonished.

In 1997, in chess, AI shocked the world, defeating a human champion who could be regarded as a symbol of human intelligence. In Go, however, the number of board positions that appear in the game is so large that it was thought impossible to defeat humans by simply applying the method used in chess. The name AlphaGo is widely remembered with great surprise as one of the milestones that brought neural networks, a breakthrough in AI technology I will discuss later, back into the limelight.

In Shogi, AI also functions as a learning partner, as seen in the case of professional Shogi player Fujii Sota, who hones his skills by playing against AI. In the past, basically, humans programmed Shogi AI with rules and conditions and gradually made it stronger over time. Then, we began to have AI learn strong players’ ways of making moves, using statistical and machine learning techniques. Now that technology has progressed even further, we can use a method known as reinforcement learning, in which AI learns winning strategies by playing repeatedly on its own, evolving without humans providing any knowledge.

Behind this evolution lies the ability to explore a vast number of positions in a Shogi match. For example, AI can now analyze Shogi positions ahead at a rate of one million positions per second, evaluate the outcomes of each possible move, and reason, “Based on what I learned from analyzing a million positions before, this move is likely to be better.” It has learned to choose the optimal move without having to look ahead. While human intervention used to be required for judging the state of a match, AI itself is now able to automatically combine factors (features) to understand the situation, such as the types of pieces in hand or the defensive formations around the Ōshō (King). To achieve this, deep learning with a neural network is often used.

All these games have reached their current level not only through technological advancements but also through improvements in hardware performance. The processing speed of computers is said to be increasing at a rate of about 1,000 times per decade, and this growth shows no sign of slowing down. These days, AI can allegedly reach the top human level in any game, given a certain amount of time. In familiar games such as mahjong, too, AI is steadily approaching humans.

Co-creation of AI and humans will create the future of games and content

Game AI is no longer just strong as it used to be. Focusing solely on AI’s strength is not adequate; we need to ask how AI should be used in relation to humans and how it will co-create experiences.

In Shogi, for example, AI’s strength is sometimes useful for level-based learning and entertainment. One professional Shogi player says that, when playing with very young children, he intentionally loses to motivate them. It is a so-called entertainment play. But the children would be unsatisfied if they always won by a large margin. It is essential to let them win by a narrow margin. This kind of approach is important not only for children but also in general practice for skill improvement, and it is necessary to develop games that help players gain insights to win.

Only with overwhelming competence can you “lose in just the right way” or “win in a way that allows your opponent to learn”. One of the goals of recent research is to enable AI to skillfully present winning strategies that make humans feel happy and satisfied. This leads to new possibilities for AI in education and motivation design.

Furthermore, the experience in open-world games is becoming more personalized, and players are finding value in stories tailored to them or in events that can only be experienced at a particular moment. AI has begun to analyze players’ preferences and behavior to optimize the story and difficulty.

Meanwhile, we still want to share the same experience with other people. Between the desire to talk and compete with others through the same game and the trend toward individually optimized experiences, future content design will require a flexible sense of balance.

The content delivery mechanisms of AI are also evolving. In e-sports, where streaming such as YouTube is the mainstream instead of traditional TV broadcasting, users can choose the player and scene they want to watch. This viewing experience is also personalized by AI, increasing fan enthusiasm.

AI is now an indispensable part of content creation. AI is becoming a creative partner in diverse fields such as games, music, video, and even dance. In particular, the value of immersive content and customized experiences is growing, changing the very nature of what content is.

Today, content creators cannot escape the influence of AI. Based on the materials and expressions created by AI, the essence of creativity will be to come up with what to communicate and what kind of experience to deliver.

As we move forward with AI, we should ask ourselves what AI can do and how far we should allow AI to go. This question may be ethical and social, or it can be about human pride in creating real value. In the business world, too, the key to the future lies not in pursuing efficiency and automation alone, but in how to build a relationship in which AI and humans create value together.

* The information contained herein is current as of February 2025.
* The contents of articles on Meiji.net are based on the personal ideas and opinions of the author and do not indicate the official opinion of Meiji University.
* I work to achieve SDGs related to the educational and research themes that I am currently engaged in.

Information noted in the articles and videos, such as positions and affiliations, are current at the time of production.