A Life without Objectives is More Likely to Achieve Greatness
Transcript from recent book discussion "Why Greatness Cannot Be Planned"
The book I’m going to discuss today is incredibly important. It’s the most mind-opening book I’ve read in recent years, and it has profoundly influenced my values. This book is called “Why Greatness Cannot Be Planned”, and its two authors are core engineers in the development of ChatGPT. Today, people believe that ChatGPT represents a transformation at least as significant as the Industrial Revolution. Hearing insights and inspiration directly from the inventors of such a groundbreaking technology, gained during the development of ChatGPT, will, I believe, make a powerful impact on every one of us.
Throughout human history, all great innovations have often emerged by chance. Every major breakthrough requires stepping stones, yet these stepping stones are hard to recognize. You’d never think that vacuum tubes would be stepping stones for computers, or bicycles would be stepping stones for airplanes. Before the Wright brothers invented the airplane, everyone who tried to fly attempted to do so like birds. Without exception, they all failed. So, what were the Wright brothers doing before? They sold bicycles. In their era, the invention of bicycle chains and automobile engines paved the way for the Wright brothers to invent the airplane. That’s the role of stepping stones.
Some might question: If greatness cannot be planned, then what about the Moon landing program? After all, Kennedy said that we would put a man on the Moon and bring him back—wasn’t that a great plan? But pay attention: the Moon landing program wasn’t a sudden, unprecedented breakthrough. By the time Kennedy made that announcement, Gagarin had already orbited the Earth. All the stepping stones needed to achieve a lunar landing were already in place. In such circumstances, you’ll find that having a plan is indeed useful. Once everyone sees the direction clearly and enough stepping stones have been identified, concentrating efforts in the short term to accomplish a specific goal is perfectly fine. That’s why the authors emphasize repeatedly that they are not trying to overturn the concept of planning or goals. Rather, they want to convey that many things cannot be planned, and that planning itself can have side effects because goals can be deceptive.
Previously, when I talked about leadership, there was a chapter called “Management by Objectives.” We often think, “If a company has no goals, how can it be managed? A company must have goals in order for us to manage it.” I once heard a master lecturer say: “Imagine you’ve arrived in a city and you need to get to your friend’s house. If you have no map—or the wrong map for a different city—how can you possibly get to your friend’s place?” When I first heard this, I thought it proved the importance of having goals. Without goals, you couldn’t reach someone’s home at all. But upon reflection, this is a flawed, mechanistic analogy. Why? Because your friend’s house is already a fixed point. We know it exists and it’s already marked on a map. Under these circumstances, following a goal is, of course, no problem. However, our lives face uncertainty. For example, who can draw you a map for where you want to be in ten years? Therefore, these two situations cannot be compared.
The authors write something in the book along the lines of questioning “ambitious objectives.” For instance, if your child has just graduated from college and sets a goal of “becoming a billionaire,” he may start to feel that everything he is doing now is a waste of time, that every action takes him farther from that billion. That billion-dollar goal may very well end up costing him his entire life.
The stepping stones to great innovation are often very strange things. Take the vacuum tube, for example. It certainly wasn’t invented with future computers in mind; it was intended for use in power systems. Because of this, the authors advocate that in completely unknown domains, we should not deliberately pursue specific ambitions. We must be wary of the “The Myth of the Objective.” This means that when you become too fixated on an impressive, lofty target, you might even discard the stepping stones already in hand simply because they don’t seem to help you reach that goal.
Life’s plans often take shape unintentionally, with countless examples to prove this point. Let me give another example: there was a man named Harland David “Colonel” Sanders—many people may not be familiar with that name. As a child, he was very unfortunate. His father died when he was six. His mother had to work, so at the age of six he started cooking for his family. Over the course of his life, he worked as a driver, farmhand, and insurance salesman, and then, in middle age, he opened a gas station. At the gas station, some drivers needed something to eat, so he made fried chicken for them. At this point, the cooking skills he had honed throughout his life finally found their moment. He then founded a chain called KFC—he was the inventor of Kentucky Fried Chicken. You see, none of the experiences he accumulated throughout his life went to waste. The ancient Chinese had a phrase, “功不唐捐” (no effort is wasted), meaning that all the things you go through may ultimately serve your life’s purpose, contributing to the realization of your goals.
In 1889, a small company was founded in Japan with the goal of selling Japanese traditional playing cards—an entertainment business. The company was called Nintendo. Over time, the playing card industry rose and fell, and in its struggle to survive, the company operated love hotels, sold rice, and produced toys. Eventually, it designed a game called “Super Mario.” That’s how Nintendo developed into the company we know today. The important thing is not that you foresaw, decades ago, that you would one day be a giant in the AI industry, or in the gaming industry, or in running a chain business. The “interestingness algorithm” is what makes life exciting.
The authors say that you should heed the call of chance rather than rely solely on logical goals. Logical goals cannot guide you toward great revolutions or great innovations. How did they arrive at this conclusion? That brings us to the invention of ChatGPT. Before ChatGPT, these young people worked together as entrepreneurs, experimenting with many directions in artificial intelligence without making much breakthrough progress. They had a very interesting idea: build a website called Picbreeder.
What’s a Picbreeder? Imagine you have two images, one like a “father” and one like a “mother.” For example, one is a black grid and the other is a white grid. They would “breed” a next generation of images, automatically generated by a program. How do two images “breed” a next generation? Scientists have a method: they assign “genes” such as shape and color to each image—this is similar to what we now call applying certain parameters. By combining these genes, dozens of new images are produced. From these dozens, you select 10 or 20 as new “parent” images, then recombine them to generate even more images.
During this process, they found that relying on themselves alone to sift through images was too slow. So, they opened the platform to internet users, allowing them to pick the images they liked and combine them. As a result, the number of images increased dramatically. This research took place in 2006. As people randomly combined images based on their “genes,” after hundreds of generations of evolution, one could look back and see what kinds of images people had “bred.”
A selection of compelling images discovered on Picbreeder. The lineage of every image in this gallery traces back to a randomly-generated blob.
Among the images that emerged, there were shapes resembling skulls, butterflies, kettles, polar bears, birds, and cameras—many interesting and meaningful pictures. What’s amusing is that the original “ancestors” of these images hundreds of generations ago were just variously shaped patterns. After hundreds of generations of combination, all sorts of intriguing images appeared.
Then they wondered: what if we set some goals for these images? For example, suppose our goal is to produce a picture of the Eiffel Tower. While selecting “parent” images, we would choose those that looked more like the Eiffel Tower to see what would happen. Unexpectedly, due to the interference of the goal (they added a host of judgments about which image looked more like the Eiffel Tower), they failed to produce an Eiffel Tower image. Do you understand? When you deliberately believe that these images will eventually form an Eiffel Tower, you actually end up moving further and further away from that target.
This realization gave them a tremendous insight: the objective function is not perfect. What is an objective function? Those who’ve studied mathematics know that an objective function is a way to measure progress toward a goal.
“So if we say that the objective function is improving, it means that our measure of progress suggests we’re moving closer to the objective. But here’s the problem: The idea that an improving score guarantees that you’re approaching the objective is wrong. It’s perfectly possible that moving closer to the goal actually doesn’t increase the value of the objective function, even if the move brings us closer to the objective.”
“This predicament sounds strange. Why would the objective function not register that we’re moving in the right direction? But of course the objective function (or any measure of progress) is going to be imperfect because imperfect human beings are the ones who have to come up with the measure. A very common type of objective function does no more than compare our current state directly to the objective. The more it resembles the objective, the higher the score we give it. This common approach reflects the rule of thumb that when you reach a fork in the road it’s always better to take the road that heads in the direction of your desired destination. ”
In other words, if you want to cultivate a good student, you actually have no idea what a great student of the future would look like. You can only try to make the student resemble great students of the past. This is how we determine the objective function. Since you don’t know what incredible images might be produced, you just pick those that look more like the Eiffel Tower. The final outcome, however, ends up being farther from the Eiffel Tower than ever, which shows the misleading nature of the objective function. “When the objective function serves as a faulty compass, this situation is called ‘deception’”—the deception caused by setting goals.
But you might think: in the industrial era, didn’t we do everything with an objective function and then continually move closer to it? We produce cars, rockets—all like this, right? Notice that in all these cases where the objective function works, it’s either short-term, very concrete, or all the stepping stones are already in place. This book doesn’t aim to overthrow the idea of having objective functions, but rather wants to tell us that in unknown territories, it’s very difficult to rely on a simple, human-designed objective function to achieve success.
Here’s a simple example: human evolution. Our ancestors were once fish in the sea. Humans now live on land; fish live in the water. It seems impossible for fish to turn into humans. Yet, humans did originate from fish that gradually crawled ashore, became amphibians, then reptiles, switched from laying eggs to bearing live young, and slowly evolved into our human ancestors. We can’t judge which stepping stone will be effective. The evolution of humanity is a process of collecting stepping stones—picking up one after another. Whenever something seems interesting, useful, or offers a new capability, we retain it. Eventually, through these cumulative changes, we emerged as human beings.
Another thought experiment: imagine you travel back more than two thousand years and meet Mozi (a great ancient Chinese thinker and inventor). You tell him, “In the future, we’ll use computers,” and you show him a modern laptop, saying, “This is the greatest invention of the future; you should invent it now.” You’d be dooming Mozi to a lifetime of misery because he cannot find any components that fit inside it. Then suppose someone else travels back and gives him a vacuum tube, saying, “This is a critical stepping stone to inventing computers.” He would think it’s useless. This is what we mean by the goal’s deceptiveness. Many young people set a goal to become millionaires, but this goal often deceives them, causing them to aim too high and neglect the small, interesting things they could be doing now—things that might actually serve as stepping stones to becoming a millionaire later.
Next, let’s talk about “interestingness” which is the essence of this book. The past provides us with countless clues. Our world is made up of a multitude of clues inherited from the past. “The past doesn’t tell us about the objective but it does offer a clue to something equally if not more important—the past is a guide to novelty. But unlike the future, there’s no ambiguity and no deception”
This is critical: we rely more on the past, more on what we’ve already mastered, because it’s certain and not misleading. “because we actually know where we were in the past, so we know how it compares to where we are today. Instead of judging our progress towards a goal, the past allows us to judge our liberation from the outdated. Interestingly, the question then changes from what we’re approaching to what we’re escaping. And the exciting thing about escaping the past is that it opens new possibilities.The point is that novelty can often act as a stepping stone detector because anything novel is a potential stepping stone to something even more novel. ”
Therefore, we need to pay more attention to new, interesting things that appear in our lives and maintain a persistent curiosity about them. This curiosity is the most important “algorithm.” Humans have a keen nose for interestingness.
How do computers achieve this? There’s a particularly intriguing experiment in the book: imagine there’s a room with a door, and we want a robot to learn how to walk out through that door. How should we design the algorithm? There are two completely different approaches. The first is the traditional approach: we give it a program with an objective function that rewards it for getting closer to the door. The closer it gets to the door, the more we reinforce it, so it quickly learns to walk out that door.
Based on the insights from the image-generation platform, these ChatGPT scientists tried another method: they only gave the robot one requirement—to do interesting things. What are interesting things? Things that aren’t repetitive. At first, the robot will wander randomly, bumping into walls again and again. After hitting walls repeatedly, it realizes that constantly bumping into walls is not interesting. It needs to find a non-wall-bumping method. Miraculously, it learns how to escape the maze. People never gave it instructions to leave the maze, yet it learned to do so because exiting the maze turned out to be more interesting.
Then, the experimenters made the maze more complicated. Once the maze became more complex, you’d find that the robots guided by objective functions failed almost every time. However, the “interestingness” robots, guided only by novelty, maintained a very high success rate.
Robot Maze (The large circle indicates the robot’s starting position, and the small circle indicates the target position.)
“If we keep this kind of process going, would it eventually discover a behavior that solves a whole maze (in other words, a behavior that drives the robot all the way from start to finish) even though solving the maze is not its objective? It turns out that the answer is yes—if we run novelty search for a while, the computer will consistently produce behaviors that drive the robot through the entire maze. That’s an interesting result because no one programmed the robot to drive through the maze and also, more importantly, driving to the goal was never an objective. The program never even knew there was a goal.”
“novelty search was much more reliable at finding behaviors that solve the maze. To be specific, we repeated the experiment with novelty search 40 times and in 39 of them a robot behavior was discovered that solves the maze. The result with objective-based search: three times out of 40.”
In other words, when you give the robot an objective function telling it to head for the maze’s exit, it only succeeds three times out of forty attempts. But when the maze gets complicated and the robot just pursues what’s interesting, refusing to keep banging into walls, it succeeds thirty-nine times out of forty. This is remarkable—a transformative merging of the algorithm behind human evolution with the algorithm behind machine evolution.
What does this have to do with ordinary people? It represents a philosophical shift. The future of your life, or that of your children, is unknown. Your plans often become obstacles. Short-term goals, like doing well on an exam, can be planned out more concretely. But who you will become in the future cannot be planned out; in fact, planning might hinder you. We need to be treasure hunters in life—adopting a treasure-hunter mentality and always being sensitive to interesting, passionate pursuits. By doing this, we can keep collecting one stepping stone after another, allowing our lives to keep moving upward. Life’s development is not a process of always getting what you want, but rather one of constantly discovering more possibilities.
Later, they applied this algorithm to training bipedal robots—robots that mimic human walking. If you try to teach it to walk just like a human, you’ll find it can’t learn properly. It wobbles, and when it encounters an obstacle or a step, it can’t get past it. So, how did they train the bipedal robot? They said, “As long as what you do is interesting, just try all kinds of interesting movements.” Thus, at first, the bipedal robot kept falling—falling to the left, right, forward, backward. After it had tried falling in every possible direction, it realized that falling was no longer interesting. Then it stood up. Standing up offered even more possibilities, so it started to walk. The bipedal robot came about not because humans programmed it step-by-step to mimic human movements, but rather because they gave it only one requirement: be more interesting. This prompted it to produce a wide variety of strange actions.
Therefore, the authors say, “embrace the reality of the much more powerful treasure hunter”. This is the treasure-hunter’s mindset. Whether we’re doing research or making life plans, we need this treasure-hunter mindset. However, it’s also possible to get lost. If a person has no goals at all, they can become aimless; robots sometimes fail too. It’s common for humans to reach dead ends in their thinking.
Some might think, “I get it. The book is saying we should combine goals with diversity. We should have objectives and then add diversity, always staying ready for it. Is that what it means?” The author says no, absolutely not. Why? Because objectives are deceptive. It’s not as simple as having objectives and then adding diversity. The objective itself introduces deception, which leads to bias in our choices. Especially in the maze scenario, you’ll find that no maze is really simple. In a complex maze, the exit itself is deceptive. Your objective function is to approach the exit, but the objective of “approaching the exit” is incorrect. In life, we have many such possibilities. We hope for safety and stability; we hope to be quickly respected by others. But you’ll discover that your desire for safety, stability, and respect can actually take you farther away from truly achieving them. This is the deceptive nature of objectives.
So, the pursuit of a prematurely determined objective can become a kind of curse. “Search is at its most awesome when it has no unified objective. Just look at natural evolution, at human innovation, at Picbreeder, or at novelty search. ”
When I read this passage, I suddenly thought of Laozi’s words, “Heaven and Earth are unkind and treat all things as straw dogs; the sage is unkind and treats the common people as straw dogs.” This is because the development of the natural world is a purposeless process. It gives every species, every individual, the same opportunity. You can grow outward, and once you’ve grown, we just need to let the interesting things remain. This process might not even involve human intervention.
After understanding the principles of novelty exploration and the image incubator, let’s consider their impact on the real world—such as progress in education. Education is a painful issue for every country right now, since almost no country is fully satisfied with its educational system. In sociology, there is a well-known principle called Campbell’s Law. Campbell’s Law states: “The more a quantitative social indicator is used for social decision-making, the more subject it will be to corruption pressures and the more apt it will be to distort and corrupt the social processes it is intended to monitor.”
“In other words, social indicators like academic achievement tests are least effective exactly when the objective is to bring them higher. ”
As soon as you impose such a simple index on a complex system, all manner of “malicious” behaviors will arise, pushing the target further and further away from what we originally intended.
We discussed a book called “Messy” before, in which the author presents a very insightful view on how to assess complex systems. How should we evaluate them? We can’t just ignore them—if we do, we have no sense of security and can’t tell what’s good or bad. The most effective way is random checks; the least effective way is deterministic indicators.
For example, during British colonial rule in India, there was a problem of venomous snakes running rampant. The government said, “For every venomous snake a citizen turns in, we’ll give a reward.” As a result, people started breeding venomous snakes to hand them in, because it was profitable. Later, the government realized it couldn’t go on and stopped accepting snakes. People then released their snakes onto the streets, causing even more snakes to appear. This is a classic example reflecting Campbell’s Law. Another case: In a place where dinosaur bones were discovered, scientists had villagers collect bone fragments. They said, “Bring us dinosaur bone fragments, and we’ll pay per piece.” On the surface, it seemed fine, but when villagers found a large piece of bone, they would smash it into many small pieces to maximize their earnings. So when we use a single, simplistic indicator, it leads to worse and completely unforeseen consequences. The world is complex, and these indicators are extremely simple. Education faces this exact problem.
This book is not about opposing goals and evaluations, because in many cases, goals and evaluations are effective. For example, in everyday scenarios, or when following in others’ footsteps, or dealing with short-term tasks, goals and evaluations work well. But in unknown realms, in complex systems, we must be cautious.
Many adults, seeing young people who seem to be doing nothing, lament from the bottom of their hearts: “What’s going to happen to this generation? What about those born in the ’90s or after 2000?” Yet Steve Jobs was exactly such a young person who seemed to do nothing. This book cites many of Jobs’s recollections. Jobs said that when he had nothing to do, he went to learn English calligraphy and typography. If he hadn’t studied calligraphy and typography back then, the later Apple computers wouldn’t have had their aesthetically pleasing typography. He felt this area needed improvement because older computer typography was too ugly. So we can’t know if a seemingly idle young person is in the process of gestating a great stepping stone. If you judge him by a simple indicator, you’ll end up discarding many potential stepping stones. The more complex the problem, the weaker the goal-oriented approach becomes, leading to ever more evaluation and distortion.
So what is “non-objective thinking” in education? The most important aspect of non-goal thinking in education is diversity among teachers and schools. Schools should give teachers a certain degree of freedom, allowing them to try different ideas and believe that some of them will become great educators. That’s a non-goal-oriented approach to education. Finland’s education system is acknowledged as one of the best in the world, partly because Finnish teachers have high status and a lot of autonomy. A teacher can differ from other teachers or even from the principal. Each teacher takes responsibility for their own class, creating educational diversity.
How do we measure this? How do we measure whether a school is good or not, especially when there is a lot of funding to allocate? The authors recommend anonymous evaluation and using randomness. Teachers don’t need to meet metrics set by the education board. Instead, teachers themselves summarize their teaching features, their methods, and noteworthy examples—without rigid constraints. After that, they submit these summaries to a committee or the education department, which assigns scores anonymously and at random. The benefit here is that it forces you to remain a genuinely good person, to genuinely do your job well.
I’ve read many books on education in the Republican era of China. There were wars and upheavals, and teachers and students at Southwest Associated University (a renowned makeshift university in wartime China) had to constantly move around. Yet this environment produced many great masters. Why? Because the teachers themselves were also masters who had no imposed metrics. They taught however they pleased. If you look at these masters’ teaching methods, you might find them strange. For example, the famous geographer Hou Renzhi would give exams that baffled students. When it was time for a geography exam, students would be anxious, not knowing what to expect. Hou Renzhi would walk in and say, “Write down where you’re from and list the local specialties there.” That was the entire exam. The students found it baffling, but he had his own ideas and interests. If we placed this in today’s environment, you’d think these masters were unqualified. Their educational methods wouldn’t pass muster. But it aligns with diversity, randomness, passion, and the algorithm of interestingness. This is the suggestion for education: anonymous assessment with random checks. Note that it’s not about having no evaluation at all, but changing the approach to incorporate randomness.
When it comes to innovation, there’s also the issue of research funding. Research funding allocation is goal-oriented: we have a certain objective to achieve, and then we allocate funds accordingly. But the author argues that this is a superficially rational process. On the surface, it must appear fair, with expert scoring and anonymous voting to maintain fairness. Yet what lies behind this seemingly rational procedure? It demands the formation of consensus. Academic consensus, however, is established on a foundation that often stifles interesting stepping stones. This is because intriguing stepping stones typically have nothing to do with the current consensus. When these like-minded experts come together to evaluate and vote, they end up strangling the directions that could be more promising.
A classic example from the book “The Cancer Code” illustrates this point. It describes how humanity’s major breakthroughs against cancer came after we realized that doctors had reached a dead end in their cancer research methods. They were stuck in a rut with no new progress. So, people brought in a group of physicists to study cancer from a fundamental physical perspective. Only then were the deeper secrets of cancer revealed. In other words, you must bring in different thinkers to uncover new stepping stones. So consensus is less useful than following individual interests.
“importance is just another broken objective compass.” In the scientific community, what is considered “important” is nearly always connected to an objective. Yet some things that seem unimportant at the time can have huge impacts later. The book cites an example from the history of mathematics.
“many researchers in pure mathematics have no intention of ever impacting the real world, and their crowning theories often sit for years as purely intellectual achievements. The famous mathematician G.H. Hardy once called practical applications “the dull and elementary part of mathematics [91],” in contrast to the poetry of pure mathematics (which seeks truth without regard to application). Yet, despite the best efforts of pure mathematicians, these “useless” results often later end up supporting developments in physics or enable practical computer algorithms. Although created for purely mathematical purposes, a particular branch of abstract algebra called group theory nonetheless has practical applications in both chemistry and physics . ”
In other words, if you ask a mathematician what they enjoy researching, they might find the applied part boring and not worth their time. However, even though pure mathematicians strive to keep mathematics purely theoretical, these seemingly “useless” theories often later end up underpinning advances in physics or enabling practical computer algorithms. Even group theory, a special branch of abstract algebra originally developed for pure mathematical purposes, later found real-world applications in both chemistry and physics.
This means that even the greatest mathematicians don’t know what practical uses their research might have. And that’s the most fascinating part. Years later, it might turn out to be extremely valuable. Take imaginary numbers, for example. Many of us in college don’t understand why they are useful, but in quantum mechanics, imaginary numbers have very important applications.
So why are people reluctant to let go of goal orientation, committee decisions, and voting when it comes to certain matters? Because doing so provokes deep fear. It makes us feel directionless, as if money is being thrown away. To soothe this fear, the author says: “there’s a simple truth that can push us past the fear: We don’t need objectives to find great things. We don’t need to seek top performance or perfect accuracy to discover something amazing. It’s like when we traded objectives for novelty in Chap. 5—we weren’t left without principles, but with different principles that better reflect how discovery really works.”
Our worry about relinquishing control is merely an obsession with appearances. We think that having a committee in control ensures safety. But in unknown fields, almost no one truly has any say. This sense of control is illusory—just self-comfort. If you can’t let go of this obsession, a vast amount of resources will remain within the realms we already understand, while those unfamiliar domains remain unseen and unexplored.
The book includes a brilliant line: “great inventors don’t peer into the distant future.” We used to think great inventors were those who had cracked some cosmic code, foreseeing what the future would be like long ago. They might have said some ambitious things, but the exact shape of the future isn’t what matters. What matters is that they found stepping stone after stepping stone and never discarded them, thereby achieving one new invention after another.
At the same time, the author points out that artificial intelligence dreads excessive ambition. When everyone races toward an AI “end goal,” you’ll find that your objective functions may prevent you from ever training a functioning AI. A truly general AI, one akin to humanity, would similarly pursue novelty and possibility. It’s only under such conditions that something as revolutionary as ChatGPT could be born.
If we cling stubbornly to a prematurely fixed objective, what will happen? Those who refuse to let go of their objectives easily fall into a kind of criticism-driven mindset. What is this mindset of criticism? It’s taking a lofty, god’s-eye view: “You’ve strayed from the goal, this is wrong, that’s a step backward, you can’t do it this way.” People who don’t actually do the work love to stand on the sidelines and criticize, full of endless worries. But we can say: “Maybe we can’t control everything else, but at least we should be able to control our own lives, right?”
If we let our own lives fall into this criticism-driven mindset, we risk internal friction. I’ve met many young people who say they’re always experiencing inner turmoil—this “internal friction” is a form of criticism. It means you’re constantly criticizing yourself. And why criticize yourself? Because you believe there’s some fixed target you’re supposed to reach, and you feel you’re getting farther and farther away from it. But if you understand what this book is conveying—that the most important thing in life is to remain happy, interested, and curious at all times—then you know you don’t need to criticize yourself every day, because no step you take is truly wasted.
In the final section of the main text, the author says: “we don’t claim to have solved all the greatest problems. Education isn’t yet perfect; science funding will never be a science itself; natural evolution still offers mysteries to be solved; and AI remains a distant goal. But while these challenges still stand, we hope you share our excitement in how non-objective thinking can change how we look at all of them. Perhaps this kind of liberated thinking can help us move forward. It may be just the stepping stone we need. ” This is a stepping stone of thought.
After the main text ends, there are two interesting case studies at the back of the book. The first is about natural evolution. By reinterpreting the process of natural evolution, we can better understand the deceptive nature of goals. When we talk about natural evolution, most people bring up “natural selection” and “survival of the fittest.” However, the phrase “survival of the fittest” has caused many misunderstandings. It seems as if our evolution has a goal. What is the goal of evolution? Some say: Isn’t it just to survive and reproduce better? No, survival and reproduction are not goals, they are conditions. They’re the basic conditions that allow life to continue playing the game of evolution. Species that fail to meet these conditions die out. But if survival and reproduction were truly the goals, then the best life-form would be single-celled organisms, since they survive and reproduce extremely well. So what is the goal of evolution? It’s novelty and abundant vitality.
Some people say that competition drives us forward. That’s not correct. It’s actually avoiding competition that drives evolution. Evolution does not come from competition, but from avoiding it. In other words, humanity, this branch of life, has always been the loser in direct competitions. Why can we say that? Because whenever a species succeeds in a certain niche, it no longer needs to explore new possibilities. It occupies that niche. Only those who fail to secure a niche are forced to explore more interesting methods. Step by step, we evolved to where we are today. The drive to avoid competition leads us to constantly search for new possibilities in space. When we were reptiles, we basically lived in two dimensions, crawling on a plane. Now we need to stand up; standing isn’t enough, we need to fly in airplanes; airplanes aren’t enough, we need rockets and spaceships to go into space. Humans have truly reached the heavens. This is the process of exploring all possible spaces.
Life began about three billion years ago, but the Cambrian explosion didn’t occur until about five hundred million years ago. Almost all ancestral forms of today’s modern animal groups appeared during the Cambrian explosion. Why did this explosion occur five hundred million years ago? Because by that time, enough stepping stones had accumulated for life’s great burst of diversity.
The author writes: “Evolution is the ultimate treasure hunter, searching for nothing and finding everything as it spills through the space of all possible organisms. It’s the world’s most prolific inventor. Even so, everything it ever produced was done without thinking about where it might someday lead. That’s why it becomes possible to understand evolution largely without the ideas of competition and fitness, perhaps providing a new perspective that doesn’t appeal to the myth of the objective.” Evolution has no mythical target; it constantly produces new things and allows whatever can survive to remain.
Another topic is artificial intelligence. We won’t go too deep into it since it’s highly technical. Simply put, what is AI? AI is the search for search algorithms. In the past, we said AI was like a bigger, smarter search tool. Actually, it’s not. AI is about searching for better ways of searching—that’s meta-search. For example, the authors of this book, who are a lucky group of OpenAI scientists, discovered a search method that is not guided by a goal function but by interestingness. This is a search for a method of searching, and it touches on the essence of AI.
As leading experts in AI, they summarize: “Good performance isn’t a stepping stone to revolutionary performance. Guarantees aren’t stepping stones to revelations. In case it’s hard to think of alternatives to arguing about performance or guarantees, there are in fact many other important clues we can consider: inspiration, elegance, potential to provoke further creativity, thought-provoking construction, challenge to the status quo, novelty, analogy to nature, beauty, simplicity, and imagination. All of these are possible for a new algorithm or any other kind of new idea. While they may lack objectivity, perhaps that is exactly what can liberate the field of AI, and many other fields at that. Anyone can say that performance should improve, but who has the courage to see the beauty of an idea? We could use a few more brave experts like that.”
Let’s take Nikola Tesla’s research on alternating current (AC) as an example. At the time, Thomas Edison was developing direct current (DC). Early on, DC appeared to have better performance—safer and less prone to accidents, among other advantages. Edison even performed cruel experiments to prove how dangerous AC was. But Tesla could see the greater possibilities hidden behind the apparent instability of AC. That was a great stepping stone.
To sum it up in the author’s words: “when even visionaries grow weary of stale visions, when the ash of unrequited expectation settles on the cloak of the impenetrable future, there is but one principle that may yet pierce the darkness: To achieve our highest goals, we must be willing to abandon them.”
People and organizations can have objectives. Personally, I think if an objective is somewhat abstract and inspiring, it might actually be good. If we aim for some great yet intangible purpose without specific indicators, that can be beneficial. But if these ambitions get turned into concrete metrics, they might mislead us. As Nietzsche advocated, you should live like a superhuman, live out the meaning and mission of your life. Such an objective has no problem because it’s a lofty inspiration. But if you set a specific goal like “I must make at least one million dollars in one year,” that might actually hinder your entire life’s development.
Most people reading this book are not AI or chip research experts; we’re all ordinary people. For us, the greatest meaning of this book is to rethink the design of our lives, especially the design we have in mind for our children’s lives. Too many parents love to set extremely clear and specific goals for their kids. Now you know how terrible and foolish that can be. After reading this book, my feeling is relaxation, a lack of anxiety, learning to enjoy life and the pleasures of exploration, and always maintaining curiosity and a desire to explore new and interesting things that appear in life. Only then can we step onto the next crucial stepping stone. Thank you all, and see you next time with another book.