Want to be great with AI? Start with imagining that you had an opportunity to interview Albert Einstein.

What would you ask?

There are all sorts of intriguing questions about the role his first wife Mileva (an accomplished physicist in her own right) had in the development of his theories. Or about roiling antisemitism in the 1930s.

Or, you could put your baseball hat on backwards and say “uh yeah, so Albert, tell me about E=MC² again?”

I think of that regularly as I’m learning to work with AI.

It turns out that, when faced with one of the most rise-of-the-machines technologies we’ve ever created, the actual power is wielded by us, the humans. AI challenges us to respond to its existence with perhaps the single most powerful tool we’ve possessed throughout history: great questions.

That blinking cursor has never felt so freighted.

My pitch to you today is that instead of allowing AI to gnaw at our sense of self, it is an invitation to double down on 100% human curiosity.

Ask bad questions, and AI will spit mid-grade pablum at you all day long. Just as in life: ask bad questions and you’ll struggle to negotiate great deals with your business partners. You also won’t discover what really motivates your customer base. Your friendships and romantic prospects might not fare that great either.

I started my career as a photojournalist and journalist. I like to say that that taught me some of the most valuable skills I’ve ever learned, like quickly gathering and vetting information, putting it in a hierarchy, figuring out why it matters, and then retelling that in words or pictures.

But really, the driving force of all that is curiosity. It won’t be replaced by AI. It can’t be, since your answer to “Why?” is yours alone. And that personal answer holds the secret to the other great human power: a point of view.

An inquisitive point of view leads to great startup messaging, which is my work and the purpose of this newsletter. But it’s a fundamental skill, and it will benefit every aspect of your business and personal life.

I recently read an article about an Adobe employee working with people testing Photoshop’s new AI function. The biggest predictor that someone would come up with great images? How much art training they had had.

AI systems are great with coming up with stuff. We have to stop being so impressed. Instead, we need to decide if it’s any good, and then ask thoughtful follow-ups with a vision of making it better. And by better, I mean better in our own exalted, precious, and very human opinion.

A few caveats. I’m still a baby at AI. I’ve tiptoed into learning about what’s known as “prompt engineering,” which is geek-speak for “asking better questions.” I’m still learning about how LLMs work and what multiple-agent systems are.

Second, I’m talking about human-readable output. The magic of the current generation of LLMs is to find hidden patterns in oceans of training data, and then apply those learnings to output. That makes these systems game-changers for things like materials research, protein folding, and drug discovery. But none of those challenge our humanity—in those fields AI is simply a new tool to discover things for which human faculties simply aren’t powerful enough.

But creative output like text and images? That’s our turf. And from what I see, AI simply heightens the premium on curiosity and point of view. If you arrive at your dinner with Einstein bearing great questions, and then know how to follow up, you’ll thrive in the AI era.

A friend of mine related a story to me about a college professor distraught at the state of teaching today. She is convinced that at least 80% of her students are using an AI to do their assignments. But what if we flipped it on its head? What if we required students to use AI to answer the questions, and then graded them on the quality of the place to which they guide the conversation?

Because if you don’t know where you’re going, any road will take you there.

Know more about AI than I do? Please tell me. Want to take the next step? Open a browser window and take an online class about prompt engineering (Andrew Ng’s Coursera course is a good place to start).