Judging AI by its worst version

Most people's perception of AI is limited by exposure to its most basic forms, leading to a narrow and often inaccurate understanding.

The two products the average person encounters most are ChatGPT's free tier and Google's AI Overviews. Neither is a fair representation of what AI can actually do. The Overviews, especially, appeared suddenly between the search bar and results, often providing incorrect information. That's a difficult first impression to shake.

This limited exposure leads many to view AI as an unreliable, time-wasting tool. This is a reasonable but incomplete conclusion that misses AI's broader capabilities.

Think about forming an opinion of any tool or craft entirely through the most basic, entry-level version of it. You're not wrong about what you experienced. But when someone starts describing what the thing can really do, it sounds like they're talking about something else entirely. Because, in a sense, they are.

The gap here isn't just about quality. It's about imagination. If your only reference point for AI is a free chatbot that sometimes hallucinates and can't remember what you said three messages prior, it's hard to picture what these tools can do when used with care, with a capable model, in the right context. People who spend real time with more capable systems are operating with a different picture of what's possible. And that gap is growing.

This creates an odd tension within organizations. An executive may want to use AI after reading about its productivity benefits. Meanwhile, employees try the free version at home, see it fumble simple requests, and conclude it's overhyped. Both experiences are valid, but they describe different products.

I'm not suggesting that everyone needs to spend hundreds of dollars a month on AI tools. That is neither realistic nor necessary for most people. But if your only exposure has been through something free and stripped-down, consider broadening your experience. Try a month-long paid mid tier subscription to a model such as Claude, ChatGPT or Gemini. Use it to solve a problem, write something you've been putting off or explore something you've struggled to understand. The goal is not to become an AI enthusiast, but to form your own informed opinion based on a wider and fairer sample.

This gap matters because perception shapes adoption, and adoption shapes what gets built next. When most people only interact with the least capable version of a technology, they are more likely to be disengaged from future adoption. This can be reinforcing and threaten the broad adoption of the tools.

Take care,

Emanuel