How should AI systems behave, and who should decide? While most people fight to learn "in-demand" skills, smart people are learning rare skills instead. These two seemingly unrelated topics have more in common than meets the eye. Both highlight the importance of thinking differently, being open-minded, and redefining value.

How should AI systems behave, and who should decide? While most people fight to learn "in-demand" skills, smart people are learning rare skills instead. These two seemingly unrelated topics have more in common than meets the eye. Both highlight the importance of thinking differently, being open-minded, and redefining value.

When it comes to AI systems, the question of behavior and decision-making is crucial. AGI, or highly autonomous systems that outperform humans at economically valuable work, have the potential to revolutionize various industries. However, the process of fine-tuning these systems is imperfect. Sometimes, the intended outcome falls short, and the user's expectations are not met.

To address this, companies like OpenAI have implemented a collaborative approach. They work closely with human reviewers to review and rate model outputs based on a range of example inputs. This ongoing relationship allows them to learn from the expertise of their reviewers and continuously improve the default behavior of AI systems.

Improving default behavior is the first building block towards achieving AI system behavior that aligns with users' values. The goal is for users to find these systems useful "out of the box" and feel that their values are understood and respected. By focusing on providing a positive user experience from the start, AI systems can become valuable tools for individuals.

However, customization is also important. AI should be a tool that can be personalized by each user within certain bounds defined by society. This allows users to define their AI's values and tailor it to their specific needs. By empowering individuals to have control over the behavior of AI systems, we can ensure that these systems serve their intended purpose.

But who gets to define these bounds? OpenAI believes that public input is crucial in avoiding the concentration of power. Giving people who use or are affected by AI systems the ability to influence the rules and defaults of these systems is a way to ensure accountability and fairness. This approach allows for diverse perspectives to be considered and prevents any single entity from having undue control over AI technology.

Now, let's shift our focus to the concept of learning rare skills. In a world where everyone is chasing after the same "in-demand" skills, those who think differently and acquire rare skills are the ones who stand out. Being a contrarian and betting against the consensus requires humility and open-mindedness. It's about being willing to question the status quo and explore new paths.

If your skills align with everyone else's, you become a commodity. To be in the top 1%, you need to be different and right. You can't expect to follow the herd and outperform the herd. Learning rare and valuable skills is akin to following the Hero's Journey, a universal archetype that has existed throughout human history. It is the path less traveled, but one that leads to greatness.

History has shown that many scientific ideas and technological breakthroughs were initially met with skepticism. The earth revolving around the sun and the concept of time not being constant were once considered heretical. The internet was initially seen as a mere toy. These examples remind us that innovation often comes from those who dare to challenge conventional wisdom.

To learn valuable skills with hidden benefits, we must look beyond the obvious. Humans have blind spots when it comes to recognizing the value of certain skills. We tend to underestimate the importance of abstract, long-term payoffs, micro-skills, skills from other disciplines, and classic ideas that have been forgotten. By broadening our perspective and embracing these overlooked areas, we can gain a competitive edge.

In conclusion, both AI system behavior and acquiring rare skills require a mindset that goes against the grain. Incorporating public input, improving default behavior, and allowing customization are essential for responsible AI systems. On the other hand, being a contrarian, seeking out rare skills, and redefining value are key to standing out in a crowded world. By embracing these principles, we can shape a future where AI systems align with our values and where individuals thrive by embracing their uniqueness.