At WAIC, for the first time I feel that AI doesn't have to be that smart

Author | Huahua

I initially planned to experience it for just 5 minutes. 

I even kept thinking about wrapping up quickly to check out the next exhibition booth. 

But instead, I ended up lying on that chair for nearly 20 minutes. 

Outside the exhibition hall, there were dancing robots, press conferences, parameter displays on large screens, and endless crowds of people raising their phones and standing on tiptoe to take photos. 

This spot was so quiet that it hardly felt like part of WAIC. 

I wore headphones and watched the EEG curves on the screen rise and fall gently. An AI music therapy system developed by the team from the Shanghai Conservatory of Music was generating a piece of music that belonged to my current moment, based on my brainwaves and emotional state. 

No robots, no Agent demonstrations, no press conferences. Even the students on site didn't mention a single model-related term to me the entire time. 

There was only music, and myself slowly calming down little by little. 

At that moment, I suddenly realized that AI doesn't always have to amaze me — it can also bring me peace. 

I

After leaving that booth, I walked back to the main WAIC pavilion. 

The familiar buzz and excitement returned before my eyes. 

The Huawei Atlas 950 Super Node was surrounded by visitors, the booths of large model vendors were packed with queuing crowds, AI glasses had almost become this year's most popular terminal, and various robots shuttled through the exhibition hall — dancing, shaking hands, running. Every few steps, someone was holding up a phone, chasing after them to take videos. 

If last year's WAIC was all about debating whose model was stronger, the most frequent keywords I heard this year have shifted to: 

Agent, computing power, terminals, workflows, embodied intelligence.

The entire industry is answering the same question: What else can AI do? 

It can write code, create PPTs, generate videos, handle complex tasks, control robots, and increasingly deeply integrate into our work and daily lives. 

This point has almost become a universal consensus across the entire conference. 

But precisely because of that, I started thinking about another question: 

What is all of this ultimately for?

The most noticeable change at this year's WAIC is that AI is shifting from model competition to system competition, and the model itself is becoming part of the infrastructure. 

What truly defines the AI experience is the complete system built on top of the model: computing power infrastructure, Agent frameworks, data closed loops, tool invocation, terminal entry points, and the ultimate capability to be implemented in real scenarios. 

This is also why so many new terminal forms have appeared at this year's WAIC. 

AI glasses aim to become a new entry point for AI to perceive the world. The Agent Phone strives to turn the mobile phone from a passive tool into an active assistant. Robots hope to give AI a physical form that can truly step into the real world. 

In the past, people asked: Can AI think? Today, people are asking: Where will AI reside? 

The answer is becoming increasingly clear: AI will enter every device that is closest to people.

Yet right in the middle of all this hustle and bustle, that quiet music therapy space felt completely out of place. 

Later, when I thought back carefully, I realized that the reason it stuck with me was not that its technology was the most advanced. 

On the contrary. 

It was probably one of the least flashy products across the entire WAIC. No multimodal large model, no complex tool invocation chain — just EEG acquisition paired with a music generation algorithm. 

In terms of technical parameters, it could barely rank among the most popular booths at this year's WAIC. 

But when it comes to emotional connection, it was the only place that made me let my guard down and stay peacefully for 20 minutes. 

This was the first contrast that WAIC brought me this year. 

AI is getting more and more powerful, but the parts that truly touch people are getting closer and closer to human nature itself.

II

Looking around the entire exhibition hall, almost all the cutting-edge technologies point to the same goal: 

Maximizing efficiency.

There's nothing wrong with that. 

Efficiency has been the biggest narrative in AI entrepreneurship over the past three years, as well as the most recognized direction in the capital market. 

Whoever can help people accomplish more things in less time will earn a ticket to the next round of competition. 

But as I wandered through the pavilions these two days, a vague feeling suddenly dawned on me. 

Efficiency does not equal happiness.

Even more, the efficiency revolution has now reached a point where it is creating a new technological paradox. 

Many people's anxieties today do not stem from low efficiency. 

On the contrary, they come from the fact that the entire world has been completely accelerated by the logic of efficiency.

There's too much information, work moves too fast, and schedules are packed full. Phones ring from morning till night, with endless messages, unreturned emails, and countless decisions to make every single day. 

AI can indeed help you finish all these tasks faster. 

But then what? 

The time saved will immediately be filled with new tasks. When every tool is boosting efficiency, the system's expectations for people are also rising simultaneously. 

In the past, technology demanded our physical strength; now, it demands our attention.

This isn't a problem unique to the AI era. 

The internet of the past 20 years, in essence, has been continuously driving the efficiency revolution. 

Search engines made information access faster, e-commerce made shopping faster, food delivery made eating faster, and the mobile internet made connecting people and information faster. 

Technology keeps helping us save time. But one question has never been answered: What do we do after we save that time? 

III

Another very obvious change at this year's WAIC is that AI is becoming more and more human-like. 

Robots now have facial expressions, can shake hands, and provide companionship. Agents can remember your habits, understand context, and proactively arrange tasks for you. AI glasses start observing the world on your behalf. Phones are trying to become an always-on intelligent assistant. 

The entire industry is working hard to make AI closer to humans. 

But at the same time, I feel that the content that truly centers on people has somehow diminished.

People discuss Tokens, reasoning speed, context window length, and tool invocation, but rarely talk about human anxiety, exhaustion, loneliness, and the real needs that technology should ultimately address. 

The more powerful technology becomes, the more human beings' real vulnerability feels like it has nowhere to rest.

That music therapy space precisely addressed this sense of displacement. 

It didn't save me a single minute, nor did it complete any work for me. But it used 20 minutes of tranquility to temporarily pull me away from the chaos of the exhibition hall. 

Actually, it's not just about music therapy. 

If you look closely, you'll find that another category of products is quietly emerging in the AI industry. 

AI psychological counseling, AI companionship, AI elderly care, AI emotion management. 

They are not the most dazzling protagonists at WAIC today. They are not as stunning as robots, nor as easy to create buzz as model launches. 

But they address another kind of need. In the past, the internet built information infrastructure, and the mobile internet built connection infrastructure. 

But in the future, AI may be building a new type of infrastructure: emotional infrastructure.

It doesn't just focus on how things get done — it cares about how people can be understood, accompanied, and cared for in an increasingly complex world. 

IV

In the past few years, the AI industry has been sprinting around one single keyword: Intelligence. The race was to see who is smarter, who has stronger reasoning capabilities, and who can handle more complex tasks. 

This is the first stage of AI development. Without this stage, there would be no AI wave we see today. 

However, as model capabilities inevitably become commoditized, computing power, parameters, and toolchains can all be gradually leveled out with sufficient capital and resources. 

What is truly hard to replicate is, and always has been, the ability to understand an individual.

In the past, technology solved the problem of how to do things faster. In the future, technology may need to answer the question of how to help people live better. 

This is why the term "intelligent companion" is worth pondering. Tools focus on usage, while companions focus on relationships.

When I left the exhibition hall, the robots were still dancing, and the press conferences were still ongoing. 

Today's WAIC showcases the new heights that AI capabilities have reached: greater computing power, stronger models, and richer terminals. 

But that quiet music therapy space also reminded me of something else: The value of AI does not only depend on how many capabilities it has, but also on how it integrates into people's lives. 

As model capabilities gradually become part of the infrastructure, the truly difficult part may turn into understanding specific individuals, specific scenarios, and specific problems.

What I remembered from last year's WAIC was the models. What I remember from this year's WAIC is a piece of music. 

The next phase of AI will still require stronger intelligence. 

But equally, it will require deeper understanding. 

This article is from the WeChat official account "Beyond the Layout", author: Huahua, published with authorization from 36Kr.