Most people think about artificial intelligence as a tool.
You ask it a question, and it gives you an answer.
Simple, right?
But this week’s chart should prompt you to start thinking about AI in a completely different way. Because it suggests that today’s leading AI models don’t just possess knowledge…
They possess something that looks surprisingly like a worldview.
Is AI Objective?
It’s easy to assume that artificial intelligence is objective.
After all, a machine doesn’t grow up in a particular country or culture. It doesn’t have political opinions, and it doesn’t practice a religion.
So it’s natural to believe AI sees the world without the biases and assumptions that shape the rest of us.
But AI didn’t appear out of thin air.
It’s been trained on enormous amounts of human writing. What’s more, researchers spend months refining each system’s behavior, rewarding some responses, discouraging others and teaching them how to interact with people.
In other words, AI learns from us.
And that’s what makes this week’s chart so fascinating.
It comes from The Economist, which compared several leading AI models with data from the World Values Survey, one of the largest studies ever conducted on human beliefs and cultural values.
Researchers mapped both countries and AI models across two broad dimensions.
One measures how traditional or secular a society tends to be. The other measures whether people place greater emphasis on survival or self-expression.
Now, no chart can perfectly capture something as complicated as human values. But this one offers a useful way to compare how different societies tend to see the world.
And what surprised me is that most of today’s leading AI models don’t resemble the average person.
Just as interesting, they don’t cluster around the United States, or China or even Europe as a whole.
Instead, they occupy a part of the map that’s distinct from almost every country included in the survey.
Some lean more secular than others. Some remain closer to traditional values.
But nearly all of them sit far toward the self-expression side of the chart.
In the World Values Survey, self-expression isn’t just about speaking your mind. It reflects a broader set of priorities, including individual autonomy, personal freedom, tolerance and quality of life.
Countries tend to move in that direction as they become wealthier and more economically secure.
Yet it appears that AI models emphasize those values even more consistently than most of the countries that created them.
In other words, the world’s leading AI models don’t simply reflect the societies where they were built. They appear to have developed a worldview that’s uniquely their own.
The obvious outlier is a model called Talkie that sits much closer to the center of the chart.
According to The Economist, Talkie was trained exclusively on text published before 1931. As a result, it places much greater importance on religion, patriotism and law and order than today’s frontier AI models.
That’s an extraordinary result. Because it suggests an AI model’s worldview isn’t fixed.
It depends, at least in part, on the information it learns from.
Here’s My Take
This chart doesn’t tell us which AI model is “right.”
It shows us that AI isn’t neutral in the way many people assume.
Every model is shaped by the data it learns from, the people who train it and the rules its creators build around it.
In a weird way, that makes AI more human. Or at least, it makes AI models a reflection of the humans who built them.
And that’s going to matter even more as AI becomes part of daily life.
Because AI is no longer just helping people write emails or summarize articles. It’s helping students learn, patients understand medical information, workers make decisions and businesses interact with customers.
It’s also doing more than just retrieving information.
AI models decide what kind of answer is responsible, what kind of advice is helpful and what lines shouldn’t be crossed.
In many ways, these models are deciding what kinds of behavior should be encouraged.
And that raises a question we’ll dig into tomorrow.
If AI models are learning values, who gets to decide what values they should learn?
Because the answer could shape how this technology develops from here.
Regards,
Ian King
Chief Strategist, Banyan Hill Publishing
Editor’s Note: We’d love to hear from you!
If you want to share your thoughts or suggestions about the Daily Disruptor, or if there are any specific topics you’d like us to cover, just send an email to dailydisruptor@banyanhill.com.
Don’t worry, we won’t reveal your full name in the event we publish a response. So feel free to comment away!
