Curiosity - Stay Ahead at the Forefront of AI

A Reminder from Fei-Fei Li

When we think of groundbreaking AI advancements, our minds often jump to complex algorithms, powerful hardware, or massive datasets.

But what if I told you that the real driving force behind some of the most significant leaps in artificial intelligence is something far more fundamental and human?

Dr. Fei-Fei Li, a name synonymous with pioneering AI research and the person behind ImageNet, a dataset that revolutionized computer vision and catalyzed the deep learning boom. In a recent conversation with Brad Smith on the "Tools and Weapons" podcast, Li revealed a surprising truth about the core of innovation:

"Curiosity is not just moments of delight, or sense of wonder. Curiosity is a state of being, and that state of being makes me happy."

This statement isn't just a feel-good platitude. It's a powerful insight into the engine that drives scientific progress. But how does one cultivate and maintain this curiosity in a field as complex and rapidly evolving as AI? Let's dive deeper into Li's journey and uncover the lessons that could reshape how we approach innovation.

Could the next big AI breakthrough be inspired by an observation at a local flea market or a casual conversation with a stranger?

Picture this: A teenage girl, newly immigrated to the United States, spending her weekends at yard sales with her father. It doesn't sound like the origin story of a world-renowned AI scientist, does it? But for Fei-Fei Li, these seemingly mundane experiences were formative in shaping her approach to science.

Li recounts how her father approached each yard sale with childlike wonder:

"He would approach every little like stand, or table of whatever items with that kind of pure child-like, 'Oh my God, Fei-Fei, come and look at this cup. It has a owl on it', or 'Look at this garden tool. I've never thought of you can design it this way so that it's easy to uproot a plant.'"

This anecdote is quite charming, and to me it sounds so silly, and yet it is a profound reminder. Li draws a direct parallel between her father's approach to yard sales and the scientific mindset:

"Now that I've been a career scientist for decades, I realize that delight in seeing things for the first time, and keeping that curiosity is the same as what we do as scientists the first time I open a paper, look at the results someone gives us, or heard about a algorithm."

The lesson here is clear:

Curiosity isn't just for labs and research papers. It's a way of engaging with the world that can be practiced and honed in everyday life.

This perspective challenges us to reconsider how we approach our daily experiences.

Could the next big AI breakthrough be inspired by an observation at a local flea market or a casual conversation with a stranger?

As a mom with small kids, this is a reminder to me how important it is to encourage them to explore and be curious, rather than just memorizing what’s taught in the text books.

Especially with the current development of AI, perhaps education in the future will not be bounded by pedigree alone. Comparing to aiming for Ivy Leagues, I would rather my kids build their confidence to pursue what they are curious and passion about in life, and not chasing “the rat race”.🤓

Do you have the courage to say “No” to a Six-Figure Salary?

Imagine standing at a crossroads in your career. On one path, a prestigious consulting firm offers you a six-figure salary – financial security and societal approval. On the other, the uncertain world of academic research, with its long hours, grant applications, and no guarantee of success.

Which would you choose?

Tbh, I probably would chose the consulting offer.

But for Fei-Fei Li, this wasn't a hypothetical scenario. It was a real decision she faced early in her career. The choice was made more difficult by her family's financial struggles and her mother's health issues. Yet, when Li wavered, her mother's advice was clear:

"Go for what you love."

Li reflects on this moment:

"Curiosity is a state of being, and that state of being makes me happy, and being a scientist makes me happy, because I'm entitled to be curious all the time, and my mom knew that, and that's what she encouraged me to do."

This decision point highlights a crucial aspect of innovation that's often overlooked:

the courage to choose curiosity over comfort.

It's a reminder that breakthrough ideas often come from those willing to take risks and follow their intellectual passions, even when the immediate rewards aren't apparent.

But Li's story isn't just about individual choice. It underscores a broader question:

How can we, as a society, create environments that encourage more people to make this choice?

How can we build systems that value and reward curiosity-driven pursuits?

ImageNet: You can change the world if you are willing to take the risk.

In the early 2000s, when AI was still in its "winter" phase, Fei-Fei Li embarked on a project that many of her colleagues considered misguided at best and career-ending at worst. This project was ImageNet, a dataset of 15 million images across 22,000 categories.

Li describes the skepticism she faced:

"Some of them will say, 'Well, that's kind of a bad idea.' Some of them would even go as far as saying, 'That might be an idea that hurts your career.'"

Yet, Li persisted. Why? Because she saw something others didn't:

"We observed, before we made ImageNet that that's just a wrong way of thinking. Mathematically, these machine learning models need to be able to learn and generalize, and the way to drive that learning and generalization in addition to the architecture of the algorithm, is actually through data, and diverse data, a large amount of data."

This insight – that big, diverse datasets were crucial for machine learning – seems obvious now. But at the time, it was revolutionary. ImageNet became the catalyst for the deep learning revolution, fundamentally changing the landscape of AI research and applications.

The story of ImageNet teaches us several crucial lessons:

1. The power of contrarian thinking: Sometimes, the most impactful ideas are those that go against the current consensus.

2. The importance of persistence: Breakthrough innovations often require pushing through skepticism and setbacks.

3. The value of interdisciplinary insights: Li's background in neuroscience informed her approach to AI, highlighting the importance of diverse perspectives in research.

But perhaps most importantly, it underscores the unpredictable nature of scientific progress.

Who could have foreseen that a massive image dataset would be the key to unlocking the potential of deep learning?

Why academic freedom matters more than ever?

In an era where tech giants dominate headlines with their AI breakthroughs, it's easy to overlook the critical role of academia in driving innovation. Fei-Fei Li's career offers a compelling argument for why academic research remains indispensable.

Li passionately advocates for curiosity-driven research in academia:

"Curiosity based research fundamentally is linked to the word freedom, is that, when you are a researcher or professor in a university in an academic setting, you don't have a manager, or a director who tells you, 'Fei-Fei, come here to Princeton or Stanford, and these are the three things you should work on.' No, you are just given a desk, an office, and an opportunity to apply for funding, and then do whatever you want."

This freedom allows researchers to pursue ideas that might seem impractical or even foolish in the short term but can lead to groundbreaking discoveries. ImageNet is a prime example of this. Had Li been constrained by short-term profit motives or narrow research directives, this project – which fundamentally changed the course of AI – might never have happened.

But academic freedom isn't just about allowing researchers to follow their whims. It's about creating an environment where:

1. Long-term thinking is encouraged: Unlike industry research, which often focuses on quarterly results, academic research can tackle problems with longer time horizons.

2. Failure is acceptable: The freedom to fail is crucial for innovation. In academia, a "failed" experiment can still yield valuable insights.

3. Cross-pollination of ideas occurs: Universities bring together experts from diverse fields, fostering interdisciplinary collaboration that can lead to unexpected breakthroughs.

4. Ethical considerations are prioritized: Academic research can focus on the broader implications of technology without being driven solely by market demands.

However, this academic freedom doesn't exist in a vacuum. It requires support, particularly in the form of funding. Li emphasizes the crucial role of government funding in her career:

"Overall, it's critical. I would say, of course you have to take it down and analyze which project got funding, but as you said, for my early years, absolutely, The National Science Foundation is one of the major funding source."

This brings us to a critical point: The health of our innovation ecosystem depends on continued investment in academic research. As we navigate the rapidly evolving landscape of AI, maintaining and strengthening this support becomes more important than ever.

The Ecosystem of Innovation: Why we need a symphony, NOT a solo

One of the most striking insights from Li's conversation is the importance of what she calls the "ecosystem of innovation."