KorvaJoin the waitlist
Tech

You Taste an Apple. AI Only Calculates One.

Your idea of an apple is built from life, while AI builds its answer from math that sounds smooth.

AI image
Business Insider · AI isn't making us smarter — it's training us to think backward, an innovation theorist says

Pick up an apple and you know it at once. You feel its weight and crunch, remember autumn orchards and school lunches, and sense what it means in pies and stories.

A chatbot knows none of that, said John Nosta, who studies innovation and leads NostaLab, a think tank on technology, in January 2026. He called artificial intelligence anti-intelligence, meaning it works in the opposite way from human understanding.

When you think of an apple, Nosta said, you connect it to places and times you have met it, to personal memories and culture, and to everyday use. That web of lived connections is what understanding means for a person.

A large language model, the engine behind today's chatbots, does none of that, in Nosta's description. It turns the word apple into a mathematical object, a long list of numbers called a vector, that fixes its position in an enormous space with far more dimensions than the three we can picture.

"An apple doesn't exist as an apple," he said. "It exists as a vector in a hyperdimensional space." The system, he said, looks for numerical patterns that line up.

That distinction mattered, Nosta said, because AI writing is built to sound smooth and consistent rather than to truly grasp things. In his framing, coherence means the wording fits smoothly as language, while comprehension means grasping what the thing actually is in the world.

Questions people ask

Why do AI answers sound so convincing without understanding?

Because it is trained on vast amounts of writing to predict which words usually follow which other words. That training makes its sentences smooth and plausible, even when there is no experience behind them.

Can AI get closer by looking at photos of apples?

Newer systems can also study pictures, sounds, and videos, so they link the word apple to round red shapes and crunch sounds. That adds useful context, but it still is not tasting, remembering, or living with an apple.

If it does not comprehend, is it still useful?

Yes, when used as a helper rather than an oracle. It is strong at summarizing text, drafting language, and spotting statistical patterns, as long as a person checks the result against facts and real experience.

More in Tech