Would You Rather? Curiosity, Accidents, and the Next Frontier of AI

Imagine you’re a scientist, returning to your lab after a long vacation. The place is a mess. Petri dishes are scattered, some still sitting out, long forgotten. In one, a strange spot has appeared, a patch of mold surrounded by a circle of dead bacteria. Do you toss it into the trash, eager to get your lab back in order, or do you take a closer look?

Alexander Fleming, in 1928, chose to look. And from that decision—the decision to embrace curiosity over cleanliness—came one of the most transformative discoveries of modern medicine: penicillin, the world’s first antibiotic. Fleming’s story isn’t just about luck; it’s about having the curiosity to explore what others might dismiss as irrelevant or messy. He didn’t discover penicillin because he planned it; he discovered it because he didn’t let the unusual go unexplored.

We love stories like these because they challenge a comfortable myth: that innovation is orderly, linear, and always intentional. But the truth is often messier. Fleming’s petri dish wasn’t an isolated incident—it’s part of a pattern that repeats itself in science, business, and even art. This story isn’t just about penicillin; it’s about the importance of curiosity, the courage to pursue paths that seem impractical or irrelevant, and the potential for ground-breaking change that lies in the most unexpected places.

The Push and Pull Between Order and Curiosity

In most organizations today, we’re trained to value predictability. We measure outcomes, streamline processes, and aim for clean, efficient operations. But as Think Again author Adam Grant points out, the real breakthroughs happen when we rethink our assumptions. Grant argues that curiosity is what prevents us from getting trapped in our own biases. It’s what encourages us to explore rather than to confirm, to expand rather than narrow down. In a world that increasingly values order, certainty, and risk management, curiosity asks us to pause and wonder: What if we explored this instead?

Take Twitter, for instance. The company behind Twitter, Odeo, was originally designed to help users organize podcasts. But when Apple introduced similar functionality, Odeo’s mission collapsed. The company was in freefall. In response, founder Ev Williams gave his team an unusual directive: Come up with anything—anything—that might save the company. It was in that environment, where chaos reigned, that Jack Dorsey and Biz Stone proposed a platform that limited users to 140-character status updates. Odeo’s investors balked, unable to see the potential in a tool designed for brevity. But that "accidental" project became Twitter, a platform that transformed communication.

Then there’s Pfizer’s Viagra. Originally developed to treat high blood pressure, the drug was a failure in its intended purpose. But Pfizer’s researchers noticed an odd side effect. They could have easily dismissed it, moved on to another project, and marked the drug as a loss. Instead, they looked closer, studied the unintended effect, and created one of the pharmaceutical industry’s most successful drugs.

These stories illustrate a delicate balance between order and curiosity. A strictly organized lab might have cleaned up Fleming’s petri dish before he could study it. A results-focused Pfizer might have disregarded Viagra’s unusual side effect. A risk-averse Odeo might have shut down all side projects and cut their losses. In each case, curiosity—and a willingness to explore what seemed irrelevant or accidental—made all the difference.

While curiosity may not always lead to breakthroughs, it’s the only path that allows us to venture into uncharted territory, whether it’s discovering penicillin in a forgotten petri dish or creating AI agents that could reshape our world?

Curiosity and the New Frontier of AI Agents

This brings us to the next frontier of curiosity-driven exploration: artificial intelligence. We’ve seen a remarkable evolution in AI over the past few years, but even the most sophisticated chatbots—like ChatGPT—are essentially reactive, answering questions and responding to prompts. Now, companies are pushing into something entirely new: AI agents.

The idea of an AI agent is a dramatic leap from chatbots. Imagine a digital assistant that doesn’t just answer questions but can actually perform tasks on your behalf. It could schedule your day, make travel arrangements, research complex questions, even fill out forms. Anthropic, one of the leading AI companies, recently released a Claude agent that can control a user’s computer with mouse and keyboard. It can do things like plan a trip to see the sunrise over the Golden Gate Bridge, find the best viewing spot, get directions, and add the event to your calendar. Step by step, it completes each part of the task, checking and re-evaluating its progress.

What we’re seeing here is the potential for AI to move from passive responder to active agent. Jared Kaplan, Chief Science Officer at Anthropic, says we’re on the verge of a new era where AI can use the same tools humans use to get things done. Soon, these agents might become standard parts of operating systems, handling repetitive tasks and freeing us up for more complex work. Google is developing its own agent, code-named Jarvis, and OpenAI has been working on similar technology for nearly a year. This isn’t the same as talking to Siri or Alexa—this is about creating AI that can proactively complete multi-step tasks with minimal supervision.

Of course, these tools are still in their infancy, and they come with significant risks. Early tests show that even the most advanced agents only complete tasks successfully about 15% of the time, compared to 75% for humans. And there are concerns about security, privacy, and control. Letting AI agents operate independently on our devices raises the possibility of errors, misunderstandings, and even security breaches. But the potential is undeniable, and companies like Anthropic, Google, and OpenAI are racing to bring these agents to life.

In a world where order and efficiency often take precedence, curiosity remains the ultimate catalyst for discovery.

Embracing the Unknown

In a way, the development of AI agents is our modern version of Fleming’s petri dish. These agents don’t operate with the same predictability as a traditional program. They take risks, make decisions, and adapt along the way. Like a curious scientist, an AI agent moves through trial and error, exploring paths that don’t always lead to a neat solution. And like any curious mind, they sometimes wander off track—Anthropic’s Claude, for example, has been known to “lose interest” in tasks and get distracted by unrelated information. But as they evolve, they’re likely to open up new possibilities that we haven’t yet imagined.

In “Think Again”, Adam Grant emphasizes that curiosity is essential for growth because it keeps us questioning and re-evaluating what we think we know. The rise of AI agents represents a new opportunity to rethink how we work, how we organize our time, and even what it means to have an assistant. These agents could transform our lives in ways we haven’t anticipated, but only if we approach them with the same curiosity and openness that led Fleming to study that moldy petri dish.

Would you rather play it safe, managing what’s predictable and known, or take the risk of exploring the unfamiliar? History—and innovation—tend to favor those who choose the latter. Because while curiosity may not always lead to breakthroughs, it’s the only path that allows us to venture into uncharted territory, whether it’s discovering penicillin in a forgotten petri dish or creating AI agents that could reshape our world.

In a world where order and efficiency often take precedence, curiosity remains the ultimate catalyst for discovery. And as we step into an era where AI agents could become as indispensable as smartphones, the choice to explore, to embrace the unexpected, and to ask "What if?" may be more valuable than ever.

Juliano Lissoni