The Hype is Over, But AI is Just Getting Started

[baoshi.rao]

The AI narrative seems to have stalled, but beneath the frozen surface, subtle currents still flow. By moving beyond fanciful fiction, understanding user needs, and leveraging quality human resources, the golden age of AI has arrived.

People always overestimate short-term (1-2 years) changes while underestimating long-term (10 years) transformations.

From VR to AI, blockchain to 5G, recent technological developments have perfectly followed Gartner's Hype Cycle—overestimated in the short term, underestimated in the long term. Initially, they are hyped with great expectations, attracting swarms of entrepreneurs and investors afraid of missing out. Then the bubble bursts, the market freezes, and enthusiasm evaporates as it hits the trough of disillusionment. It is in this neglected valley that new technologies truly begin their steady climb toward maturity.

AI is currently in this climbing phase.

Not long ago, AlphaGo defeated Lee Sedol, nations rolled out grand AI strategies, and tech entrepreneurs were brimming with ambition, eager to become disruptors across industries.

But within just two years, this narrative seems to have abruptly stalled.

For most AI entrepreneurs, the core challenge is asymmetry: those who understand technology often don’t grasp market needs and pain points, while those who understand markets struggle to find the right technology. This is a tech-demand mismatch.

Tech-focused entrepreneurs often fall into the trap of "showing off"—boasting about their impressive technology but stumbling when asked, "What problem does this solve for users?" This mindset spawns many "cool but impractical" products.

If all you have is a hammer, everything looks like a nail. Entrepreneurs forcing AI into solutions often create superficial needs—applications where AI is cool but not essential.

Even if you identify a real need and pick the right direction, success doesn’t come overnight. Algorithms need time, data, and iteration—patience is key.

Many entrepreneurs love quoting Henry Ford: "If I had asked people what they wanted, they would have said faster horses." But if AI is the car, today’s applications are often clunky, expensive, and prone to breakdowns—hardly an improvement over horses. Why would users pay for that?

Data must accumulate, algorithms must train, computing power must grow—the entire system requires continuous iteration. There’s no overnight success, only gradual progress.

But capital won’t wait. Despite a market flush with promising projects post-shakeout, funding is scarce.

Worse, growth in specialized fields means navigating "deep waters," where entrenched interests pose massive resistance.

Want to disrupt live translation with AI? Translators will resist. Automate financial settlements? Good luck getting past CFOs.

If blocked by incumbents, AI can’t gain traction in professional fields, making data collection, deep learning, and algorithm training impossible.

Tech-demand mismatch, slow maturation, impatient capital, industry resistance—these are the four mountains AI entrepreneurs must climb.

Jack Ma once said: "You must learn to win with your team under pressure—without encouragement, recognition, understanding, tolerance, or retreat." Today’s AI entrepreneurs face exactly that.

But this might be a good thing.

In 2018, a rare African swine fever pandemic ravaged the globe, hitting China hardest—over 100 million pigs died in a year.

Before a vaccine, the only options were cutting infection sources and monitoring outbreaks.

A telltale sign of swine fever? Pigs’ breathing and temperature spike, their exhales turn abnormal, and their movements grow stiff and swollen.

But who can monitor pigs 24/7? Who can distinguish a fever from a cold? And what if farmers delay reporting to avoid losses?

In Luoyang, Henan, an idea emerged: What if AI could create a smart monitoring system using sound and image recognition to detect outbreaks instantly?

If successful, AI cameras and audio devices would outperform human oversight.

Following this, an AI-powered smart farm launched locally, with a renowned tech company’s voice cloud providing the tech. In six months, the first 30 pigs were market-ready.

AI, once a lofty concept, is now in pigsties—seemingly "unserious," like Ding Lei’s pig farming. But this might be the right start.

In 2019, 8.34 million graduates entered China’s job market, up 130,000 from 2018. For coveted roles at top firms, the first hurdle is the interview—a one-shot, high-stakes challenge.

What if candidates could simulate interviews beforehand, with AI mimicking corporate HRs, critiquing every answer and gesture? Repeated practice would build confidence.

This exists for civil service exams, but corporate interviews are far more complex.

A Shanghai startup proposed an AI interviewer that learns HR preferences, analyzes responses, and gives tailored feedback.

They built an AI interview app where students practice via video and receive personalized advice.

Could such AI eventually replace HRs in some roles?

The same logic applies to education. "Great teachers create great students," but they must tailor their approach. Yet elite teachers are scarce, leaving poorer regions underserved. Even the best teachers lack time to personalize lessons for every student.

Enter AI education: Companies use AI to "scan" students’ levels, then deliver customized content (created by elite teachers), simulating a virtual, omnipresent mentor.

Or take beauty pageants and talent shows, rife with vote-rigging and favoritism. Could an AI system learn diverse aesthetics and simulate public reactions, bypassing judges and votes to ensure fairness? A Hangzhou firm is testing this with Miss Asia.

These projects reveal a pattern for AI’s real-world application.

First, identify areas where expertise is scarce. There’s a Lee Sedol in Go, elite teachers in education, top doctors in hospitals—but they’re rare. This scarcity creates market gaps, and gaps mean opportunity.

1. Experts can be simulated by AI. While expertise involves both talent and accumulated experience, AI can approach expert-level performance through massive data training. Just like AlphaGo, which played millions of Go games against itself overnight after losing its first match.

2. Massive data requires massive manpower, creating demand for personalized solutions. Pig farms can't have professionals monitoring 24/7, HR faces mountains of resumes, and even the best teachers can't customize education for every student. These are precisely the areas where AI shines - you don't need an army of people when you have a high-performance AI system.

Recently, TIME Magazine listed the 100 Best Inventions of 2019, with 'Accessibility' being the most notable category. It featured AI-powered hearing aids, eye-tracking technology, smart canes, intelligent prosthetics, and audio-visual conversion systems.

Similar breakthroughs are happening in China. In Guangdong, a hearing-impaired debate team member became 'Best Debater' despite being completely deaf. How does someone who can't hear - including their own voice - debate?

In China, about 210 million people have hearing impairments, with 72 million having moderate-to-severe hearing loss. Yet only 2.5% have access to effective hearing aids or cochlear implants.

AI offers solutions like:

• Real-time captioning apps that convert speech to text
• Speech rehabilitation training using voice recognition and synthesis

The Guangzhou debater uses such apps to convert opponents' speech to text, while his rehabilitation training enables verbal responses - even if his pronunciation isn't perfect yet.

Another challenge: sign language interpretation. In the U.S., the interpreter ratio is 1:120, while in China it's below 1:10,000. Chinese companies are developing smart bracelets that translate sign language into text and animations by tracking muscle signals.

While disabilities deprive people of certain abilities, AI is breaking down barriers:

1. Text-to-speech and image recognition help the visually impaired 'hear'
2. Speech-to-text and sign language translation enable communication for the deaf
3. Voice-controlled wheelchairs empower those with mobility impairments

Major tech companies are pursuing 'accessibility' goals. The assistive technology market is clear in both product and business models. The AI-assisted rehabilitation market is projected to reach ¥103.3 billion by 2022.

These solutions address daily communication needs, evolving through data accumulation and algorithm improvements. Future integration with IoT, robotics, and VR will bring more smart devices and rehabilitation tools.

As iFlytek's Liu Qingfeng noted, AI's true potential is shown through:

1. Tangible outcomes
2. Scalable applications
3. Measurable efficiency gains

AI thrives where there's:

• Imbalanced resources and scarce expertise
• Need for massive manpower
• Areas beyond human limitations

When these needs meet AI's data-processing capabilities, we find the 'sweet spot' for innovation.

AI may seem distant yet is remarkably close. To all entrepreneurs in this field: though the AI bubble has burst, its best era has just begun.