The Story Behind Hinton

Foreword

Some say education is to help you cross class boundaries. Others say education merely makes you recognize class. The truth, perhaps, lies in the vast expanse between these two.

——Epigraph

As the wave of artificial intelligence sweeps across today’s world, let’s turn our gaze to the figures at the crest of this wave, exploring the invisible forces that shaped them and profoundly influenced our era. Geoffrey Hinton’s life is an excellent example. With AI currently a hot topic, I’ve examined the life experiences of several individuals, gaining insight into the essence of social class change.

Illustrious Family: The Starting Point of Life

Imagine what it would be like to be born into a family that has produced geniuses for 200 years. Hinton was such a child. His ancestor, George Boole, invented Boolean algebra, laying the foundation for modern computers. His grandparents, aunts, and uncles were all scientists, mathematicians, and professors.

But such a family background, on the contrary, put immense pressure on young Hinton. His parents were atheists, yet they had to send him to a church school because, at the time, the best schools in England were run by churches. This contradiction caused young Hinton great pain: on one hand, his parents told him “religion is nonsense,” while on the other, he had to pretend to be pious at school.

Finally, in a religious class, a teenage Hinton erupted. When the teacher said, “All good things come from God,” he stood up and retorted, “Then who do you say gives good things?” “Russia!” Saying this during the Cold War in 1960 was simply scandalous. But thanks to his family’s influence, he wasn’t expelled.

In high school, Hinton became the “problem student” of the school; no teacher dared to manage him. He had a good friend named Inman Harvey, the kind of person who would skip classes and play games with you every day, but whose grades were always better than yours. Harvey once told him, “You know what? Our brain is like a holographic storage device; even if you scoop out a spoon of brain, it won’t affect memory and thinking.” This statement profoundly shocked Hinton, and he resolved to study the brain from then on.

Thanks to his family background, Hinton was easily admitted to Cambridge University. But his university life was practically a “history of dropping out”:

First time: Studied Physical Chemistry, dropped out after a month. The reason was straightforward: “It was my first time away from home at 18, studying was too tiring, no girls, I was depressed.”
Second time: Switched to Architecture, only attended one day before dropping out again. He found that the teachers didn’t talk about design ideals but only discussed “the budget is overspent, need to replace with cheaper doorknobs.”
Third time: Transferred to Physiology, intending to leave after a few classes, but heard the teacher would talk about “how the brain works,” so he stayed for two semesters. As a result, the teacher merely dryly stated: “The brain has trillions of neurons and works through neural signals.” Hinton immediately lost his temper.
Fourth time: Went to study Philosophy, feeling that understanding the brain required philosophy. But he soon fell out with his supervisor.
Finally: Transferred to Psychology, where he finally persevered until graduation.

Each time he dropped out, Hinton would go to London to work as a carpenter, working while reading Russian literature, thick books like “Resurrection” and “The Brothers Karamazov.” Perhaps unable to bear his parents’ nagging, he would always go back to school.

The True Academic Path: Progressing Through Doubt

In 1972, at 25, Hinton finally found his direction. He heard that a professor at the University of Edinburgh was researching “neural networks,” so he immediately went to apprentice. But as soon as he arrived at the lab, he encountered an “academic funeral.”

That day, his supervisor, Professor Higgins, gathered all his students, who formed a circle. The senior students began to “confess”:

“Our current hardware simply cannot simulate the brain.”
“If this method worked, Turing would have done it already.”
“Stop imitating; neural networks have been proven to be a dead end.”

It turned out that in 1969, a book called “Perceptrons” had practically pronounced the “death sentence” for neural networks. Computers at the time were too weak; 8-inch floppy disks had less than 1MB of capacity, making it impossible to process the large amounts of data required to train neural networks.

Facing this atmosphere of despair, everyone was ready to switch to “symbolism,” an easier path for publishing papers. But the rebellious Hinton refused.

Professor Higgins was clever; he told Hinton: “Of course, you can also use your research to prove that symbolism is wrong and neural networks are viable. Especially you, Hinton, I know you want to do this.”

Hinton replied: “Give me half a year, and I will prove that symbolism is wrong!”

And so, Professor Higgins “hooked” Hinton with a dare. But half a year passed, and nothing was achieved. Hinton proactively approached the professor: “Give me another half year!” Another half year passed, still no results: “Give me another half year!”… This conversation repeated 12 times, a full six years passed.

In 1978, perhaps Professor Higgins couldn’t stand it any longer, forcing him to graduate, earning his Ph.D. in Artificial Intelligence. It was like being swept out the door.

Years in America: Unexpected Smoothness and Painful Choices

Hinton heard that many universities in the United States were researching neural networks, so he decided to develop his career there. The first few years were difficult; he changed two university positions without establishing himself.

Until 1982, he joined Carnegie Mellon University. Suddenly, everything became incredibly smooth: money was not an issue, people were not an issue, and he couldn’t find an authority to contend with. Because almost no one in the university was researching neural networks, his colleagues all believed it was a dead end.

In 1983, at 36, Hinton achieved both love and career success. He found his girlfriend, Rosa, in the US, a staunch socialist and a molecular biologist at the University of California. They married, had a happy family, and his career flourished.

In 1986, after 20 years of persistence, Hinton finally had a breakthrough. He and two colleagues published a paper on “backpropagation of errors,” which completely ignited the field of artificial intelligence. This algorithm proved that neural networks could have multiple layers and achieve autonomous deep learning. Although computational power was insufficient at the time, everyone knew, thanks to Moore’s Law, that such computational power would soon be realized.

Overnight, the academic world was shaken. Everyone wondered: Did we misunderstand neural networks? Was Hinton right?

This was the greatest validation for a rebel. But the terrible reality reignited his rebellious spirit.

It turned out that his smooth research in the US all these years had been funded by the US military. When the Pentagon learned of Professor Hinton’s successful neural network, they immediately expressed their desire to increase investment and asked him to develop weapons for the military.

In 2023, 76-year-old Hinton recalled in an interview: “Most of my AI research funding at the time came from the Pentagon. This conflicted with my values against using AI for military purposes, so I chose to leave.” [2, 3]

Behind this seemingly calm response lay Hinton’s most valuable choice in life. In 1987, he suddenly left Carnegie Mellon University, moved his entire family to Canada, and resumed a reclusive life. Like Russian revolutionaries, he gave up everything for his beliefs.

Long Wait: 25 Years of Persistence in Canada

You cannot imagine how difficult it was to conduct deep learning research in 1987 without national, military, or major financial support.

Hinton’s Ph.D. student Yann LeCun (later head of Meta AI) calculated for him: if Moore’s Law holds, you’ll need to wait another 50 years, roughly until 2037, for computational power to train neural networks. By then you’ll be 90; can you wait that long?

LeCun, of course, couldn’t wait. The following year, he went to Bell Labs, embracing large banks with an algorithm that could recognize handwritten characters. But Hinton was unmoved; he knew he wanted true artificial intelligence, and he would wait for a miracle to be born.

This miracle was called NVIDIA.

One Sunday in 2002, Hinton was coding in his office when he suddenly heard an urgent knock at the door. Opening it, he saw a young student who said he had been frying French fries all summer and was interested in machine learning, wanting to work in the lab.

Hinton originally intended to politely decline, telling him to read the “Backpropagation” paper first and come back next week to discuss it. But when the agreed time came, the student said: “I don’t understand why the paper’s author didn’t include the gradient in the optimizer; it could clearly be more efficient.”

This remark shocked Hinton: “Who are you?”

“My name is Ilya Sutskever, from Israel, and I studied at a community college for two years.”

And so, Ilya, who could fry French fries, became Hinton’s prized disciple. Ilya told his teacher: “Do you play games? I know a company called NVIDIA that makes GPUs, and everyone uses them for gaming. But I think using GPUs to train neural networks can be dozens of times more efficient than CPUs.”

But gaming graphics cards were expensive, and the lab had no budget. Hinton decided to use his influence to email NVIDIA: “I’ll recommend your graphics cards to 1000 machine learning experts; can I get one?”

NVIDIA didn’t reply at all. Hinton joked: “They completely ignored me. Later, I told Jensen Huang, and he immediately sent me a graphics card.” [2]

In 2006, after a credit card payment failed, the lab scraped together funds to buy a GPU for training. The efficiency immediately increased 30 times! Hinton saw the miracle he had waited 25 years for.

In 2012, when NVIDIA was in trouble, Ilya from Hinton’s lab used CUDA to train a visual recognition AI, and the effect was surprisingly good, winning the visual recognition competition. NVIDIA’s graphics cards were proven to be custom-made for training AI.

Hinton couldn’t let others get ahead; he needed a large number of graphics cards. So he asked his former student Yann LeCun for help. LeCun suggested he start a company, as Andrew Ng of Google would acquire it. Just like that, DNN Company, founded only 4 months prior, was acquired by Google for $44 million.

Hinton used that $44 million to become a major customer of Jensen Huang. AI, powered by deep learning, was finally slowly awakening within NVIDIA graphics cards.

Was all this truly fate’s arrangement?

Pt.I The Chasm

Overturning fate is for the few, otherwise “Nezha 2” wouldn’t have gone viral overseas.

——Epigraph

Such a family background brings not just financial comfort, but also a deeper cultural capital and social network. While children from ordinary families are worrying about tuition fees and struggling to make ends meet, children from such families are already starting from a higher point. The circles they move in from childhood, the resources they acquire, and even their inherent confidence are entirely different from those of ordinary people.

The fundamental reason why ordinary people find it difficult to cross the class divide is not merely a gap in ability or effort. More often, it’s the vast difference in opportunity costs. A child from a working-class family must weigh the risks and rewards of every choice – to pursue further education or to start working early to support the family? To pursue ideals or choose stability? For those with deep family foundations, the cost of trial and error is almost negligible. They can capriciously pursue their interests, start over after failure, and even persist in their views amidst controversy without fear of consequences.

This disparity is often most evident at critical junctures in life. When opportunities arise, many are prepared, but those who can seize them are often those who already possess the corresponding social capital.

Ordinary families cannot afford such risks, because for them, education is not so much about nurturing a child as it is a high-risk investment. Can the child become independent? Will they become a “boomeranger” after graduating from university? When will they marry and have children, establish a family? These questions are unknown gambles for ordinary families.

My classmates and I were watching the comments section of the “VideoForce” YouTube channel’s million-subscriber video when we learned that the creator, Tim Pan, is the son of the president of YTO Express. We were quite shocked when we learned about his family background – this is a manifestation of the differences in family education. Sohu has a detailed report on his growth experience, which is much better written than mine, and I’m too lazy to repost it: https://www.sohu.com/a/802705862_100246658.

Sometimes we have to question the authenticity of films like “The Pursuit of Happyness.” In reality, any business facing competitive pressure and survival challenges is unlikely to risk hiring a job seeker with neither qualifications nor relevant experience – this is simply difficult to understand from a business logic perspective.

And a recent data release on the career destinations of Peking University graduates shows that the overall employment rate for fresh undergraduate graduates reached 93.81%, but the primary destination among them was further education, accounting for 84.23%. One might ask, is this truly the desired path for top talent? If they can’t find ideal jobs, they continue studying. Are they all going to research quantum computers?

Looking at the founder of Telegram, whose father won more than one Olympic Informatics Competition gold medal, one can’t help but look at the power of “genes.”

Regarding the issue of class mobility, some philosophers have simply attributed it to “a matter of perspective.” This seemingly ordinary statement is actually to the point and has profound historical corroboration.

Now, when we look back at that sentence that became viral again and again in the Chinese-speaking community on X:

You were always meant to be a worker, you just happened to study for a few years. Studying is not to make you cross class boundaries, but to make you recognize class.

It seems to become understandable.

The core idea of this sentence is: the true role of education has been misinterpreted.

Simple interpretation:

“You were always meant to be a worker” - points out a harsh reality: most people’s social status is largely determined at birth, and family background largely dictates your life trajectory.

“You just happened to study for a few years” - implies that education is more of a process than a magic spell for changing destiny. Studying is just part of your life experience, but it cannot fundamentally alter your social attributes.

“Studying is not to make you cross class boundaries” - directly denies the traditional notion that “knowledge changes destiny.” It suggests that achieving class mobility simply through education is an illusion.

“But to make you recognize class” - this is the key. The true value of education lies in allowing you to:

Understand how the social structure truly operates.
Recognize the objective existence of class differences.
Understand your true position within this system.
Learn how to survive better within the given framework.

This statement reflects a rather pessimistic but perhaps very realistic social observation: education is more about making people “accept their fate” than “change their fate.” It reminds us not to hold overly romanticized expectations for education, but to view the practical difficulties of social mobility more rationally.

Pt.II Sifting Through the Sands

When we finally recover from the high pressure of education and look back at the “sifting through the sands” of ancient times, it seems that opportunities for advancement were also sought through education. During the Warring States period, Shang Yang’s reforms in the State of Qin began to reward military merit, and the aristocratic hereditary system instantly collapsed. Shang Yang’s measure indeed broke the solidification of social classes for a certain period.

By the Sui Dynasty, the establishment of the imperial examination system provided a completely new channel for talent selection. Compared to the flawed “recommendation system,” the seemingly foolproof examination became the optimal solution for selecting talent. From Confucius’s “three thousand disciples” to Zhou Jin’s despair in “The Scholars,” we can see what a significant class leap passing the imperial examination once represented.

Behind Fan Jin’s madness upon passing the imperial examination lay a sea of blood. This was not merely the ideological confinement imposed by the imperial examination system, but also a microcosm of the brutal oppression of the old society. Hu Tu’s transformation from haughty to obsequious offers a glimpse:

Before Fan Jin passed the examination, he sarcastically ridiculed his son-in-law; after Fan Jin passed, he eagerly straightened his son-in-law’s clothes, “lowering his head and tugging at them dozens of times along the way.” This dramatic transformation reflects society’s almost pathological worship of power and status.

“Fan Jin Passes the Imperial Examination” is famous for its satire because Wu Jingzi, in creating one Fan Jin, actually criticized a group of “Fan Jins.”

Systems can change, but the human desire for and reverence for class seem never to have changed.

Pt. III The Mirror

If Leo Tolstoy was the mirror of the Russian Revolution, then “puppy love” can be said to be a mirror of modern Chinese society — a complex reflection of its education system, family structures, cultural values, and even social anxieties.

The term “早恋” (zǎo liàn), meaning “early romance” or “puppy love,” is a typical Chinese mainland concept, appearing in the 1960s. At that time, the country was in a highly politicized phase, with personal life integrated into ideological management, and individual emotional expression was no exception. Before this, in the first half of the 20th century, especially in rural society, dating at 16 was not considered remarkable and was even seen as a natural and normal part of growing up. However, from the 1960s to the 1980s, against the backdrop of the state emphasizing “serving socialist construction,” students dating were generally seen as having “unstable political positions” or “not cherishing valuable learning opportunities.” By the 1990s, although society gradually marketized and individual space expanded, in secondary and family education, romance was still viewed as a symbol of “premature maturity” or even “going astray,” and parents and teachers maintained a high degree of vigilance towards it, which continues to this day.

The term “早恋” itself carries strong contextual and cultural connotations, almost becoming an unavoidable “keyword” in the Chinese narrative of growing up. Unlike mathematical formulas like “odd-even rule, sign depends on quadrant” that are confined to classrooms, or “Peking Opera” which is merely a part of traditional culture, “早恋” permeates the daily conversations of countless families, campus disciplinary systems, and social moral consensus. It is worth noting that in the English context, there is almost no exact equivalent for this term. Western society uses terms such as teenage romance, puppy love, or young love, which do not carry explicit negative connotations; instead, they imply a gentle, naive, and tentative beauty of youth. In contrast, the term “早恋” carries a strong moral judgment in its linguistic structure – “早” (early) implies “inopportune,” “transgressive,” as if love must also adhere to a socially prescribed timetable and cannot “jump the gun.”

Behind this linguistic difference lies a contrast between two distinct philosophies of growth and life perspectives:

The Western model tends to view growth as a process of exploration and trial and error. Romance, as a natural part of adolescence, is an important way to understand oneself, learn about relationships, establish boundaries, and gain emotional experience. Society generally holds a more tolerant, even encouraging, attitude towards adolescent romance, considering it a necessary stage towards maturity.
The Chinese model emphasizes the completion of staged tasks and the achievement of linear goals. From childhood, children are strictly divided into different “stages”: elementary school for laying foundations, junior high for hard work, the high school entrance exam as the first turning point in life, and the college entrance exam as the determinant of fate. In this model, the identity of “student” almost becomes a state of temporarily “freezing other roles” in life, meaning one can only “study hard” and not be “distracted.” Romance is thus seen as a dangerous distraction, an act that could potentially lead one “off track.”

Under this model, romance is not only suppressed but also “demonized.” Puppy love is seen as a disruption to the educational path, a violation of family expectations, and, at a deeper level, a “challenge” to the established social order. Therefore, we can say that the strict regulation of puppy love is not due to the “problem” of romance itself, but because it touches the sensitive nerves of society’s anxiety about a “standard life template.” On the “study hard → get into a good school → find a good job → buy a house and marry” track, any emotional deviation could be seen as a “dangerous derailment.”

The cultural roots of this phenomenon can be traced back to the Confucian cultural tradition deeply embedded in Chinese society. Confucianism emphasizes “self-cultivation, family harmony, governing the state, and bringing peace to the world,” advocating that emotions should be subservient to rationality and social order, proposing “emotions are expressed but stopped by rites and righteousness,” and establishing gender segregation norms like “men and women should not touch each other.” The core of these norms lies in suppressing individual desires to maintain the stable structure of family and society. Individual emotions in this system are viewed as “private desires” that need to be restrained, tamed, and even disciplined.

In modern society, this cultural concept has not disappeared but has been integrated into the modern education system in new forms. The past “self-restraint and restoration of rites” has become “no early romance,” and the past “self-cultivation and family harmony” has transformed into “improving grades, no distractions.” The moralized suppression of emotional expression continues subtly in the “grades first,” “admission first” educational structure, simply donning a modern, institutionalized veneer.

Therefore, in Chinese society, “早恋” has become a highly sensitive issue due to two levels of pressure: one is at the realistic level – intense competition and immense pressure for college admission; the other is at the cultural-psychological level – a deep-seated vigilance and moral suspicion towards emotional desires that has always existed. This cultural gene has been continually internalized in the growth process of generations, becoming a “collective unconscious,” so much so that even the original “parties involved” – including teachers, parents, and students themselves – may not truly realize the source of this suppression.

Precisely because of this, “早恋” is not only a social phenomenon but also a cultural symbol with significant value for reflection. What it reflects is not just the naive impulses of adolescence, but a microcosm of an entire social order and cultural logic. When we discuss “早恋” again, perhaps what is more worth noting is not romance itself, but why we are so afraid of its existence.

Foreword

Pt.I The Chasm

Pt.II Sifting Through the Sands

Pt. III The Mirror

References: