Placing humanity at AI's core

At Beijing's Zhongguancun National Innovation Demonstration Zone Exhibition Center, discussions about artificial intelligence (AI) moved in an unusual direction. Instead of the familiar focus on speed and scale, attention turned to a more basic question: how AI relates to human subjectivity.

That perspective shaped the U8 World Innovation Summit (U8-WIS) 2025 Annual Summit, held at the venue on Dec 13.

Founded by students from eight leading universities worldwide, the forum blended academic inquiry with youthful initiative.

Under the theme "Xiaoyao You" — a Taoist concept often translated as "free and easy wandering" — more than 40 scientists, entrepreneurs, and professionals gathered alongside nearly 1,000 young participants to explore how technology might contribute to human flourishing and a deeper sense of inner fullness.

One of the speakers who took up that question was Joseph Sifakis, a Turing Award laureate and academician of the Chinese Academy of Sciences, who joined remotely as a guest lecturer.

Sifakis challenged a common assumption in today's AI discourse: that progress means building machines that can outperform humans at every task.

"AI may soon pass medical exams as well as students," he said. "But does that mean AI should be allowed to practice as a doctor?"

For Sifakis, the point was that performance benchmarks can be a misleading measure of success. While current AI systems may excel at certain defined tasks, he noted, they still lack situational awareness, adaptability, and creative reasoning — capacities required once machines are expected to carry human-level responsibility.

AI, he suggested, should be designed to align with human life rather than replace it. "The success of AI will depend on its ability to understand and contribute to systems of human attention — the ultimate stage of this kind of technological development," he said.

From this perspective, he argued that China is well-positioned to articulate a human-centered vision of AI — one that places societal needs on equal footing with technological advancement.

Gu Shi, a well-known science fiction writer who also delivered a keynote, approached this vision by drawing on imagined futures to examine how emerging technologies are reshaping the human condition.

In one conversation with a robotics practitioner at the forum, she realized that a device she had once invented for a story was already moving into real-world development.

It wasn't the first time her fiction had arrived early. In The MagiMirror Algorithm, a short story published in 2022, the "MagiMirror" is a micro-expression analysis app embedded in contact lenses — an always-on system that reads emotion as seamlessly as sight itself. Now, just three years later, Gu believes that today's AI is beginning to resemble the MagiMirror, functioning like an external organ growing onto the human body.

"As technologies increasingly integrate into everyday life, such extensions will only multiply," she said.

Yet in the story, the system is ultimately suspended: reading others without their consent is deemed an ethical violation — a decision that suggests technological capability may need to give way to human agency and choice.

For Gu, this is precisely where science fiction matters. "Science fiction cannot halt technological progress, but by imagining future scenarios, it helps us better understand the implications of our actions today," she said.

That emphasis on agency also shapes her view of AI in her creative process. Gu does use AI tools during the research phase to learn unfamiliar disciplines and map technical possibilities, but when it comes to writing itself, she draws a clear line. "For me, creation is joy — even a painful one — so I don't want anyone else to play the game for me," she said, adding that letting AI write in her place would reduce the creator to a mere spectator.

Gu remains optimistic about the younger generation of creators. She believes they will choose to hold onto authorship rather than hand it over to machines — learning to take full ownership of imagination and language as their craft.

Who bears responsibility

That expectation found a more immediate echo at U8's youth-focused forum, where young participants from different fields shared how they navigate authorship, responsibility, and choice in an age of intelligent systems.

Among them was Pan Zhoudan, a PhD candidate in engineering science at the University of Oxford. He reframed a common anxiety about AI with a simple question: when people say they "trust AI", who are they really trusting?

Pan noted that we are naturally drawn to what resembles us, and that human-like AI can create a compelling illusion of being understood.

"But resemblance should not be mistaken for accountability," he argued, because in the foreseeable future, AI systems cannot bear legal or moral responsibility. "When something goes wrong, responsibility will still lie with the humans who design, train, and deploy them."

For Pan, trust should ultimately be placed in people, not machines. He cautioned that trust formed through limited human-AI interaction should never substitute for independent judgment.

He doesn't use AI as a shortcut, but as a testing ground. When entering unfamiliar fields, he first builds his own understanding through reading and reflection, then asks AI to challenge his reasoning. Sometimes, he poses the same question to multiple AI systems and compares their responses to sharpen his judgment.

"I treat AI as an audience. If I'm not satisfied with its answer, I'll work out my own and show it back to the system — to see how it responds," Pan said.

In the end, he added, the goal isn't to outsource thinking but to strengthen it — so that in a world of increasingly capable machines, the final responsibility still carries a human name.

mengshuyan@i21st.cn