AI doesn't think. It calculates.

1. LLMs read everything. They understand nothing.

Apple ran language models on a maths classic: Tower of Hanoi. An old puzzle that hinges on a single logical rule. The models failed, over and over. Not because the puzzle's hard - because they've no idea why the right answer is right. No world model. No common sense. Just probabilities. A person who's never seen the puzzle needs thirty minutes and one lightbulb moment. The model doesn't do lightbulbs. It needs training examples. Without them, it guesses. Confidently.

2. The real problem isn't hallucination. It's missing curiosity.

Hallucinations grab the headlines: invented answers, fabricated quotes, sources that don't exist. But there's a deeper issue. An LLM can't stop and say "I don't know this - let me think." It keeps generating. Confidently. Without doubt. A person who doesn't know something knows they don't know it. A language model doesn't. It doesn't even notice.

3. For 80% of business tasks, none of that matters.

Checking invoice text. Summarising emails. Drafting proposal templates. Cleaning up meeting notes. None of these need a thinking system - they need fast, decent writing. LLMs deliver that. Well. Provided you know what to expect and review the output quickly. Most businesses I work with haven't got tasks that need deep thought. They've got tasks that eat time - and that's where AI does real work.

4. The danger zone: handing over decisions.

I see it again and again: an AI system launches, delivers good results - and three months on, nobody questions it. The model decides, the person nods along. That's when the missing world model gets expensive. AI should never have the final say. Not because it's reckless or makes mistakes. Because it has no clue what it doesn't know.

5. The right question isn't "Can AI think?" - it's "What should it do for me?"

Use AI to replace intelligence and you'll be disappointed. Use it to speed up routine work and free up capacity, and you'll be surprised what opens up. That isn't a limitation. That's a realistic expectation - and it pays off.