AI vs Skin in the Game

Recently, I went to an event where we were supposed to talk about issues we were having in business, and then what kind of advice would help move us forward.

One big issue I have is that I tend to build many features and instead of reaching out to customers, I simply keep building more features, or I will move on to build another product entirely.

Instead of giving me advice, some members asked about if I tried asking AI about that. Since I’m a heavy AI user, I obviously already did that. But then I told them, I actually wanted to hear from people who had “real experience” with similar issues. Basically, I wanted to hear from people who experienced my issues, and then solved them.

At that point, they cut me off, and ask which model I’ve been using. I said that I was using Claude Sonnet. Then they exclaimed “Aha!”. They claimed that I should be using Claude Opus instead, which is devoid of logic, because Claude Opus simply has more knowledge and “intelligence”, but still no real experience.

But then they argued that yes, Claude Opus has much more experience. That’s when I realized that most people don’t know how AI works because they were confusing knowledge with experience.

For example, a business studies teacher has a lot of knowledge in topics related to business, but its possible that this person has never owned and operated a business themselves. The experience component is missing, and thus, any real-world advice from that person should be taken with a grain of salt.

But why is that? Isn’t the knowledge enough to validate all claims? I’d argue that no, because the knowledge was simply gathered over time from things that other people shared. They may or may not have actually happened.

That’s exactly how AI works. It pulls information from its training data, and matches keywords from your prompt to that information, and summarizes everything for easy digestion.

For example, if you’re having an issue in your personal relationships, AI might have some advice for you. But AI has never been in a relationship, so it has no experience, and thus, the advice can’t be taken seriously. If AI tells you to break up with your partner, it’s simply saying what other people have said. These people may or may not have been in the same situation you’re in right now, so the advice might not actually apply to you.

But what about programming and knowledge work? AI excels at those, as we’ve seen. Those are different because it’s easy to gain experience in knowledge work. AI can simply write a piece of code, and verify that it works right away. It builds experience immediately, with no consequences.

While discussing issues with AI is a good thing for uncovering hidden knowledge in your own brain, there’s currently no substitute for getting advice from humans with real-world experience in your exact domain.