We live in a world where AI tools can draft essays, summarize books, and even chat with us as if they truly understand. For many, this feels like proof that machines are becoming intelligent. But David Krakauer, President of the Santa Fe Institute and a leading thinker on the evolution of intelligence, offers a very different perspective. He warns that while AI is powerful, mistaking its capability for genuine intelligence could weaken our own ability to think and understand. His message is clear: AI should support human intelligence, not replace it.
AI as Capability, Not Intelligence
Krakauer describes most of today’s AI as “fake intelligence” or “a very quick lookup.” It’s not thinking — it’s retrieving. This matters because when we rely on AI summaries or outputs without doing the work ourselves, what we gain is not true knowledge but what he calls “fake knowledge.” Real intelligence, he insists, comes from the act of understanding and explanation, not from having pre-packaged answers handed to us.
Why Understanding Matters More Than Output
For Krakauer, intelligence means making a hard problem easy. That process involves effort, explanation, and comprehension. This is why he critiques the famous Turing Test: it rewards outputs that appear human-like, but it doesn’t require the AI to actually understand anything. And if we accept those outputs blindly, we risk bypassing the very mental work that builds real intelligence in ourselves.
The Problem-Solving Example
To explain the difference, Krakauer borrows an example from astrophysicist Neil deGrasse Tyson. Imagine two interns asked to determine the height of a church steeple. One works it out using tools and reasoning, arriving at a thoughtful estimate. The other simply memorizes the correct figure from a book. The first shows intelligence; the second only mimics it. AI, Krakauer argues, is like that second intern — highly efficient at recall, but without genuine problem-solving. And if we let AI do all the problem-solving for us, we miss the chance to engage in the thinking that creates understanding.
Tools That Make Us Smarter (and Tools That Don’t)
Krakauer also introduces a useful distinction between tools that complement human intelligence and tools that compete with it.
- A pencil or an abacus, for example, is a complementary cognitive artifact: it enhances our mental powers without replacing them.
- GPS navigation or full reliance on AI summaries, on the other hand, can become competitive cognitive artifacts: they offload skills we should be practicing ourselves, leaving us “dumber” in the long run.
AI, depending on how we use it, can fall into either category.
Using AI the Right Way
The good news is that Krakauer doesn’t reject AI outright. He believes it can be an invaluable assistant when used properly. A tool like NotebookLM, for instance, can help analyze transcripts, highlight key points, or surface relevant details. This kind of support makes it easier for us to focus and navigate complexity. But the crucial step — the actual summarization and synthesis — should still be done by us. That’s where true understanding happens, and that’s how we keep our intelligence sharp.
Conclusion
Krakauer’s warning is timely: don’t outsource your thinking. AI can make us faster and more effective, but only if we remember that the real work of comprehension is ours to do. By using AI to guide us but taking ownership of explanation and summarization ourselves, we ensure that machines remain what they should be — powerful assistants — while intelligence, problem-solving, and true understanding remain deeply human.