Over the Spring Festival, I found myself in Vietnam, a country that holds fond memories from my middle school years, now more than two decades ago. The transformation it has undergone is remarkable. During this time, significant events unfolded globally, particularly within the generative AI industry. It's encouraging to see DeepSeek pushing OpenAI to accelerate its pace, hastily releasing the o3-mini large model (for free) and introducing an agent for in-depth research tasks. As an AI layperson, I refrain from commenting on technical intricacies, preferring to wait and see. I envy those who can effortlessly pen detailed analysis pieces, undoubtedly grounded in a profound understanding of generative AI's underlying technology; for those without such a foundation, silence is golden.

Subsequently, the trade war reignited, not merely a bilateral but a multilateral affair, far removed from its iteration six or seven years ago. This issue is in constant flux, with daily reversals, yet the United States retains overall dominance. As an international affairs novice (with no aspirations to become one), I observe this matter with an open mind, offering little commentary.

This morning, another major event dominated social media trends: the Golden State Warriors orchestrated a multi-team player trade, sending out three players, including Wiggins, in exchange for Butler. Butler subsequently declined his player option for next season, signing a lucrative two-year, $121 million contract with the Warriors. This deal makes him the second NBA player to earn an annual salary exceeding $60 million, following his new teammate Curry.

Regarding this trade, many NBA fans initially thought, 'The Warriors are crazy,' paying such a steep price for a 35-year-old star who will command two $60 million contracts next season. Optimistic fans believe the Warriors aim to capitalize on Curry's twilight years, making a final championship push. While not an NBA expert, even as a casual fan, I recognize Butler's contract value is roughly equivalent to 10 times the NBA's mid-level exception or 40 times the minimum contract (depending on seniority). In the NBA, where millionaires abound, this is astonishing! For ordinary Americans and others, over 99.99% will never earn this much in their lifetime, and 99% may never have even seen such an amount.

So, within the two weeks surrounding the 2025 Spring Festival, generative AI technology witnessed significant progress, the global trade war entered a new phase, and Butler inked one of the NBA's most lucrative contracts. How are these events interconnected? From a developmental logic standpoint, the first two share a connection, but adding the third seemingly disrupts this link.

However, they all affirm a crucial trend in contemporary human society: the intensification of the Matthew Effect and the ensuing societal restructuring.

Let's start with Butler. Two to three decades ago, sports stars were wealthier than average individuals, but not to this extent. Michael Jordan's $30 million annual salary contract in 1996 shocked the world, as no one envisioned an athlete earning such a high salary. Adjusting for inflation, $30 million in 1996 is equivalent to $60 million in 2025, meaning Curry and Butler's salaries have matched or will soon surpass Jordan's era-defining level. More players will eclipse this benchmark in the future, though whether their athletic prowess surpasses Jordan's is a moot point.

Since the NBA introduced the salary cap system in 1984, the salary cap has increased by 35 times in just four decades; the salary gap between top stars and average players has more than doubled. Combined, top stars' absolute salaries have skyrocketed nearly 100 times, excluding income from endorsements, investments, etc. LeBron James signed a Nike sponsorship deal worth $90 million before joining the NBA; Iguodala, O'Neal, and others have amassed significant wealth through venture capital investments leveraging their Silicon Valley connections. An NBA star with some acumen and a prolonged career can lead an extremely affluent life. NBA players who go bankrupt often succumb to vice and poor financial management, not income issues.

Meanwhile, how have ordinary people's salaries fared? Taking non-farm workers in the United States as an example, their average hourly earnings have increased by just over three times since 1984. This growth is respectable, yet their income levels continue to widen with NBA players, falling further behind top stars. A prime example of the 'Matthew Effect'!

(Note: When calculating NBA player salaries and ordinary people's salaries above, inflation factors were not excluded to maximize comparability.)

Why are NBA stars earning more? Their talent and skills, amplified by mass media and the internet, create a tremendous 'leverage effect.' In the 1980s, cable media was underdeveloped, and streaming media hadn't emerged. Today, all television and online platforms globally vie for NBA content, with broadcast contracts skyrocketing every 3-5 years. Social media's popularity has significantly increased stars' exposure and fan loyalty, bolstering their bargaining power with advertisers and boosting their traffic value. Additionally, advancements in sports science and medicine have extended stars' careers, enabling them to earn for more years. James at 40 and Curry at 37 are still active, unimaginable 20-30 years ago.

The aforementioned advancements, including mass media, streaming media, social media, and sports science, undeniably benefit everyone. Society's lower limit has risen, but the upper limit has soared even higher, intensifying the 'Matthew Effect.' Possessing talent far surpassing ordinary individuals—a sufficiently long 'long board'—enables leveraging contemporary advancements to monetize talent efficiently. For the vast majority without exceptional talent, life remains livable as long as they don't compare themselves to superstars, fostering a sense of disparity.

The development of generative AI will undoubtedly exacerbate the Matthew Effect. Everyone acknowledges generative AI will significantly reduce the demand for human repetitive labor but fails to realize (or is unwilling to admit) that most human labor is repetitive. This applies to both physical and mental labor, as anyone who has spent time in an office knows well! Lawyers, accountants, investment bankers, analysts, consultants—these are among the most lucrative 'ordinary' jobs, with much energy spent on document formatting, modifications per client needs, and document sending/collection. Only a small fraction of documents require 'complex logical thinking,' and those needing significant decision-making abilities are rarer still, often not made by workers.

Until late 2024, generative AI's primary role seemed to 'assist' rather than 'replace' workers, determined by its working mode: most large model applications are instant dialogue-based, where you ask a question, and it strives to answer promptly. For complex, multifaceted questions like 'an industry's investment value,' 'a project's feasibility report,' or even 'how to write music player software,' workers still need to break them down, feed parts to the model, and organize responses. In this process, the model handles most work, but the worker's central coordination remains essential.

However, since mid-2024, more large models have introduced 'deep thinking' capabilities: feed a detailed, complex question with reference materials, and wait for a considered response. OpenAI's recently launched 'Deep Research Agent' boasts completing tasks taking humans hours in minutes. Technically, this isn't difficult to implement and even optimizes large model vendors' computing power allocation.

After adjustment, human tasks may only include: accurately and comprehensively describing tasks, making critical decisions, and engaging in highly creative labor.

This also heralds the 'white-collar society's' collapse, not just in the West but anywhere needing white-collar workers. Large models save costs, improve efficiency, and reduce organizational friction. This reminds me of the Western world's 'blue-collar collapse' from the 1980s to 2010s—a massive manufacturing job transfer to China-led developing countries. While expanding the economic pie, it significantly widened the wealth gap, causing indelible social and cultural impacts. The emergence of the 'Rust Belt' and 'Shantytown Song' are footnotes to this monumental change. In a sense, globalization and the Western world's 'blue-collar collapse' are synonymous.

Addressing globalization's problems elicits diverse responses: some embrace globalization, believing 'gone times cannot be relived'; others advocate trade wars to 'revive the past.' This is more a position than intellectual or logical stance, as globalization enriches Silicon Valley and Wall Street while hurting Rust Belt interests. Silent in the media, harmed blue-collar workers never stopped seeking their 'spokesperson'; now, they've found one.

From a broader perspective, we curiously discover a 'double Matthew Effect' globally (especially in the West) over the past few decades:

A small elite at the pyramid's peak—super sports and entertainment stars, large company CEOs, successful tech entrepreneurs—have grown richer relative to society's average, with gaps widening by tens to hundreds of times. This is primarily the internet's 'Matthew Effect.' Upper-middle-tier individuals—Silicon Valley coders, Wall Street investment bankers, senior white-collar workers in multinational corporations—have grown richer relative to lower-middle and blue-collar classes, with gaps widening several times. This is mainly globalization's 'Matthew Effect.'

Now, generative AI's in-depth development will create a 'third Matthew Effect': the white-collar class's collapse, at least for those neither at the decision-making level nor creative. The result is twofold—the white-collar class will increasingly converge with the blue-collar class (already happening in many places), and diverge from the pyramid's peak. Most foresaw this day but couldn't emotionally accept it. For instance, on Zhihu, so-called 'internet practitioners' confidently declared, 'Large models only impact liberal arts, not science and engineering, especially not coders.'

I hope they sleep soundly after witnessing GPT-4o1 and DeepSeek R1's programming capabilities.

How should we respond to this technological progress-induced social change? I don't know, and probably no one does right now. Historically, such drastic changes evoke opposition, conservatism, and adaptation. Over hundreds or thousands of years, you might argue 'progressive forces always win.' The question is, who are the 'progressive forces'—those promoting the Matthew Effect or striving for equality? Also, whose 'victory' is this—compared to the elusive 'human society,' we care most about our vital interests.

What I know is, at historic junctures, we must sit tight, observe actively, and decide cautiously. In this technological revolution-dominated era, 'change' is the norm, and 'no change' a pastoral fantasy. The world is unfair and will become increasingly so. We must accept this to survive and grow. If someone says 'problems can definitely be solved,' they're offering meaningless comfort, because most world problems are unsolvable. Humanity's greatest wisdom is acknowledging problems' existence and cleverly coexisting with them until we transcend them.