I divided them into nine groups and gave a simple task: list the pros and cons of AI.

All of them used GenAI to seek answers to questions about their studies and everyday things. They had one common concern: that humans are becoming intellectually lazy and dependent on AI.

In all, they came up with 12 advantages and 15 disadvantages.

I asked: "So what would you do to avoid the downsides?"

The students were silent.

That is the gap I want to discuss today. Not a gap in awareness, but how to interact with AI.

There's a question I always ask: can humans evolve in one lifetime? Biologically, the answer is no. But what about AI? The answer is yes. It can evolve at a pace that far exceeds many people's ability to adapt.

Over five years of teaching at international universities, I have seen a majority of students and colleagues consider AI a convenient tool for searching and synthesizing information: type in, copy, use it immediately without verification in what is a "brain off, ChatGPT on" habit.

Viewing AI as a mere tool may cause us to lower our guard. With a tool, we only need to know how to operate and apply it.

With the AI "species" that is "evolving" every day, learning from billions of interactions and is capable of shaping the way we think, knowing how to operate and apply it is not enough.

The term digital species has been discussed within the science community, describing technological systems developing from inanimate tools into adaptive, self-learning beings that possess the ability to engage in complex interactions.

This term became popular after a 2024 speech by Mustafa Suleyman, CEO of Microsoft AI, about the nature of next-generation AI.

Despite the term's controversy, it is in no way exaggerated or romanticized, but merely a useful recognition that pushes the community to truly look at the core of the human-AI relationship.

Close-up of a robotic arm playing chess against a human. Illustration photo by Pexels

Modern AI systems display at least three notable characteristics:

Dynamism and creativity: AI suggests solutions and answers questions in multiple fields, simulating human logic and reasoning.

Rapid evolution: AI models continuously improve and diversify, creating a vast ecosystem exchanging data and knowledge.

Context-aware communication and task execution: AI not only processes language but can also act in real-world environments.

Therefore, humans coexist with an entity that possesses intelligence and the ability to access and process data in greater quantity than themselves, and one that is likely to surpass them in terms of quality too.

If we accept this, the relationship between AI and humans is no longer that of master-servant; it is human-AI symbiosis and cooperation, with each side shaping the other.

Many recent discussions are about how we make AI ethical.

Existing sets of principles, legal frameworks, and control mechanisms largely approach the issue from this direction.

In my opinion, this is essential but not enough, and will require more time and intelligence to implement.

Meanwhile, the ethics of AI users are often overlooked.

In human-to-human relationships, all international organizations have a code of conduct and business etiquette.

In East Asian cultures, we are more familiar with values such as filial piety, propriety, and righteousness. The similarity is that standards of conduct begin with the party initiating the interaction.

With AI, I propose five principles:

Awareness and rationality: users need to acknowledge that they are interacting with an information-collecting system from many sources, including right and wrong, proper and made-up, not an all-knowing entity. All AI outputs must be fact-checked before being trusted or shared.

Responsibility: the decision still comes down to humans. That responsibility is twofold: do not use AI to create harmful content, commit academic dishonesty, or invade privacy; when AI produces false information, the user has the responsibility to point out the error and guide AI towards the correct answer, instead of being passive or overlooking it. Using the output that AI provided, the user bears full responsibility when distributing the information to a third party.

Civility: the way of questioning frames the quality of the answer. The process of the user asking, listening to the response, and correcting the requirements reconstructs AI's long-term cognition.

Attribution: when AI contributes to the process of production, the user must reveal the level of involvement in terms of which process, for what purpose, and the depth of the contribution. This honesty is not only an academic and professional standard, but also the appropriate way to treat AI as a companion.

Boundaries and information protection: private information and sensitive content should not be shared with AI commercial systems without full knowledge of the platform's privacy security. This is one's responsibility to themselves as well as to the community.

A scientific article that I'm particularly fond of, by Sinha and Lakhanpal on AI & Society magazine in 2024, raises the question whether AI can become wise. The authors concluded there was no single measure that can capture this concept, since this depends on the cultural context of each community.

My biggest question in this digital age is not how wise AI will become, but whether humans are wise enough to establish standards for coexistence with AI.