Understanding AI: Intelligence versus Wisdom

Summary: This new series is a kind of travelogue. It’s a way for me to document my experiences with AI. Today’s question, can an AI tool be wise?

Missing point of view.

I think it’s important to find ways to differentiate how conversational AI works compared to human beings. Because there are times when it seems frightfully similar.

One area is perspective.

Right now pretty much all the leading conversational AI tools are built to have a kind of generic point of view. They’re all trying to be helpful and guide their users through their own discovery process. This isn’t the way people work. Not only do we have knowledge, but we also have a point of view. That point of view is developed over time through experience and socialization.

These different perspectives are part of make diversity so important. Different people with different perspectives will take the same information and come up with different interpretations.

You can actually mimic this through a chat with your favorite AI tool. Ask the tool to take on the persona (or roleplay) as someone. It could be a general persona, like an experienced plumber or a high school social studies teacher. Or it could be a well known person, like an author or public figure. Then ask it a series of questions.

The answers to those questions will change based on the role you give the AI tool, at least to a point. AI is just roleplaying, and sometimes it goes a little overboard. For example here is a prompt where I asked Claude to respond to a basic question as if it were Richard Feynman.

Prompt: Can you give me a definition of evaluation but do so as if you were Richard Feynman?

Claude Sonnet 4:
adjusts in chair and grins

Alright, so you want to know what “evaluation” means? Let me tell you something – most people make this sound terribly complicated when it’s really quite simple.

Evaluation is just figuring out if something is any good or not. That’s it! You’ve got some thing – could be a student’s paper, could be a new theory about atoms, could be whether your car mechanic did a decent job – and you want to know: is this thing worth a damn?

This is just an example from a conversation with Claude.ai, the AI kept going for a bit.

Wisdom is more than a collection of knowledge.

So, most AI tools have access to a lot of information and can generally handle a lot more info than a human can before getting overwhelmed (at least in some respects since they don’t see, or feel, or think about other stuff at the same time, like a human would).

But wisdom is much more than just access to a collection of knowledge. To make it easy to discuss, let’s simplify with a fake equation. Wisdom = Point of View + Knowledge + Time

For an AI tool, Knowledge is generated through a specific training database and a specific training approach.

The default Point of View is the result of the information it was trained on, and the design choices made by/selected by the model development team.

Time is mostly just a snapshot. At least by the time we experience the tool.

A human point of view is going to change considerably over time, ultimately shifting how information is viewed.

Could an AI model be built with wisdom? Perhaps, at least it could imitate wisdom. But I think it’s going to be one of the areas where humans will continue to have an edge.

The race towards the best average.

Right now there are several tools locked up in a very weird race. They are all trying to be more useful than all the other tools at doing all the things. Basically, it’s a race towards being the best average.

I can’t see it staying like this.

There is this story about cockpit design in airplanes. It goes like this, early on in plane design, the air force tried to design a jet cockpit around the idea of an average pilot. The problem is that there is no average pilot. So by creating a cockpit for the average person, you are essentially creating a cockpit for nobody. The better way forward is to build an adjustable cockpit, so you can meet the needs of different people.

This is starting to happen in AI, with different conversational AI models being offered at the same time each trained towards different uses. My guess is that over time we’ll see an increase of specialized AI, built from specific training libraries and trained to serve different needs.

That’s all I got for today. What have you discovered about AI lately?