MaHui Book Giveaway | Why Greatness Cannot Be Planned

The world's first computer was built with vacuum tubes.

In 1918, two British physicists, William Eccles and F.W. Jordan, described a method for generating pulse sequences using cross-coupled vacuum tube amplifiers. This circuit became the foundation of the flip-flop, which has two states and thus became the basic element of electronic binary digital computers — humanity had its first generation of computers, the vacuum tube computer.

But imagine if you lived in the 18th century with the goal of building some kind of computer. You would never think to invent the vacuum tube first. And for more than 100 years after the vacuum tube was invented, no one realized it would prove so useful in computing. It wasn't until 1945, when ENIAC (Electronic Numerical Integrator and Computer) was invented — the first programmable electronic general-purpose digital computer, built with over 17,000 vacuum tubes forming its logic circuits — that the connection became clear.

If there were a space storing all historically possible inventions, the vacuum tube would certainly sit right next to the computer. Once the vacuum tube was invented, the computer wasn't far behind. The vacuum tube was a "stepping stone" to the computer; you only needed to discover the link between them.

But who could have foreseen this?

In our cultural mindset, setting goals, working hard to achieve them, and measuring progress along the way has become the dominant path to success. For concrete, "small wishes," goal-setting works extremely well — if a company decides to increase production capacity by 5%, setting a target is highly effective. But what if that same company wants to leap across the development curve and pioneer the next generation of a national-level application? For these more ambitious goals, goal-oriented thinking seems to fall short — because achieving such objectives is a complete unknown.

If your goal were to invent the microwave oven, you certainly wouldn't think to research radar.

If you wanted to invent an airplane, you wouldn't spend decades developing the engine.

If you aimed to build a computer, why would you devote the vast majority of your time to studying vacuum tubes?

...

Yet in cases like these, what you would never do is precisely what you should do. Only those who didn't treat the microwave, the airplane, or the computer as their ultimate end goal could find the critical stepping stones that led to these inventions.

This perhaps suggests that in our society, which increasingly emphasizes "goal orientation" and measures everything by quantifiable metrics, our standards for success — the very way we judge whether we're moving in the right direction — may be deceptive.

No one knows for certain which stepping stones will give rise to the greatest discoveries. And the key to success lies in maintaining openness to an ever-changing environment, allowing possibilities to grow freely, and avoiding getting locked into our initial grand ambitions. When genuine opportunity arrives, seize it boldly.

Kenneth Stanley and Joel Lehman are both world-renowned AI scientists and researchers at OpenAI. More than a decade ago, they began a rather unique scientific experiment — creating the "image breeding" website Picbreeder.org. The AI scientists found a way to create an artificial DNA for images stored inside computers, enabling people to combine image genes through hereditary integration, much like animals.

Just as humans breed animals for desirable traits — faster horses, better-proportioned purebred dogs — the evolution of domesticated animals reflects human preferences. On Picbreeder.org, users could select their preferred images from 10-20 original pictures, prompting the AI to generate new image sets, ultimately creating images they found appealing or interesting.

In breeding technology, having a breeding goal helps people cultivate their desired breeds. But in AI image generation, the opposite proved true: the images users ended up liking most were often not what they had initially set out to create. To verify this result, the AI researchers designed a program that used a specific image as the final generation target, with the computer automatically selecting the image most similar to the target each time. Yet every choice that appeared to bring them closer to the target ultimately produced an image completely different from it. In other words, this approach could not produce the target image — the experiment failed.

Starting from this small, counterintuitive AI experiment, Kenneth Stanley and Joel Lehman began rethinking the value of "goal supremacy," eventually writing the book Why Greatness Cannot Be Planned. Beginning with the AI experiment, they expanded their inquiry to natural selection, cross-era technological progress, and other fields, ultimately telling us: If you want to achieve great goals — exploring the heavens, crossing the furthest horizons — then in certain cases, you need to abandon goals.

At the dawn of the AGI era, as new technologies bubble up and transform the times, Source Code Capital recommends this book to you. Don't judge whether each project is worth pursuing by its potential for standalone success, but by its potential to spawn other projects. The only important function of a stepping stone is to bring more stepping stones.

You never know where the path beneath your feet will lead.