It depends on what you mean by smarter. AI already wins at chess. But it still can't make a cup of coffee.

The question sounds simple. The answer isn't. Let's break down what 'smarter' actually means — and what the experts really think.

↓

THE 6 DIMENSIONS

What 'Smarter' Actually Means

AI isn't simply 'smart' or 'dumb.' It's brilliant at some things and clueless at others. Tap each card to flip it.

🎯

Narrow vs General

Tap to flip

AI already destroys humans at specific tasks — chess, Go, protein folding, math. But each AI only does ONE thing. A human who plays chess can also drive a car, cook dinner, and comfort a friend. General intelligence means transferring skills across domains. No AI has this yet. ^[1]

⚡

Speed vs Understanding

Tap to flip

AI processes millions of words per second. But speed is not understanding. A calculator is faster than you at arithmetic — that doesn't make it smarter. AI predicts patterns without grasping meaning. It generates text that sounds right without knowing why it's right. ^[1]

📊

Benchmarks vs Reality

Tap to flip

GPT-4 outperforms 99% of humans on the Torrance Tests of Creative Thinking (2023 study). GPT-4.5 was judged as human in 73% of Turing test conversations (2025, Jones & Bergen). But acing a test isn't the same as being intelligent. Tests measure specific measurable outputs, not the full spectrum of cognition. ^[1]

🧩

Patterns vs Reasoning

Tap to flip

AI finds statistical patterns in massive datasets. It does not reason about cause and effect. It can predict what word comes next, but it can't plan a multi-step strategy the way a human does. As of 2023, foundation models still lack advanced reasoning and planning capabilities. ^[1]

🏔️

Moravec's Paradox

Tap to flip

Hard things are easy for AI. Easy things are hard. AI beats grandmasters at chess but struggles to fold laundry, walk up stairs, or recognize a sarcastic joke. Sensorimotor skills that humans evolved over millions of years are far harder to replicate than abstract reasoning. This is why AI can write poetry but can't reliably make coffee. ^[2]

🏐

The Moving Goalposts

Tap to flip

Every time AI achieves something once considered 'intelligent,' humans say: 'That's not REAL intelligence.' This is called the AI Effect. Chess was the benchmark for AI — until AI won. Then it was 'just computation.' Translation, face recognition, creative writing — each dismissed after being conquered. We keep redefining intelligence to exclude what AI can do. ^[3]

QUIZ

Myth or Reality?

Can you separate what AI actually does from what people claim? Test your instincts on these 6 statements.

AI already performs better than 99% of humans on standardized tests of creative thinking.

AI has achieved general intelligence equal to humans across all cognitive tasks.

In the 1960s, AI pioneers predicted human-level AI within 20 years. They were wrong.

Most AI experts believe artificial general intelligence will never happen.

AI can already pass the Turing Test — convincingly fooling humans into thinking it is human.

AI systems today have consciousness and self-awareness.

INTERACTIVE

When Will AI Become Smarter Than Us?

Experts disagree wildly. Drag the slider to explore the range of predictions — from 'already happened' to 'never.'

050100

EXPERT PREDICTIONS FOR AGI

THINK CRITICALLY

7 Ways to Think About AI Intelligence

Don't get swept up in hype or fear. Use these mental tools to evaluate claims about AI. Tap to check off each one.

Distinguish narrow vs general intelligence

When someone says 'AI is smarter than humans,' ask: at what? Chess? Yes. Making coffee? No. General intelligence across all tasks? Not yet.

Watch for the moving goalposts

Every time AI achieves something, we redefine 'real intelligence' to exclude it. Notice this pattern. It means our intuitions about what counts as 'smart' are unreliable.

Benchmarks are not reality

Passing a test — even the Turing Test — doesn't mean intelligence. Tests measure specific outputs. Real intelligence is flexible, contextual, and adaptive in ways tests can't capture.

Confidence is not understanding

AI always sounds confident. But generating fluent text about a topic is not the same as understanding it. Fluency masks the gap between prediction and comprehension.

Don't anthropomorphize AI

AI says 'I think' and 'I feel' — but it doesn't. It predicts what text should follow. As Yann LeCun warns: be careful not to interpret AI intentions as if it were human.

Track who's making predictions (and why)

CEOs of AI companies predict AGI soon. Skeptics say never. A 60-year analysis found predictions consistently cluster at '15-25 years from now' regardless of era. Consider incentives.

Remember: experts have been wrong for 60 years

In 1965, experts said 20 years. In 1980s, 10 years. In 2000s, 20 years. In 2024, 5 years. The pattern of overconfidence is itself a data point. Stay curious, stay skeptical.