You have probably heard grown-ups on the news say the word superintelligence in a serious voice. Here is what it means, in plain words — no serious voice needed.
The one-sentence version
Superintelligence is a mind that would be much smarter than the smartest humans at almost everything.
Not a robot with laser eyes. Not a computer that is good at one game. A mind — probably a computer program — that would be better than the best doctor at medicine, better than the best scientist at science, better than the best writer at writing, all at once.
Nobody has built one. It is an idea that people are trying to understand before it arrives, if it ever does.
The staircase
The easiest way to picture it is a staircase.
On the bottom step is an ant. Clever for its size, but it can’t learn a new game. Next comes a mouse, which can learn a maze. Then a dog, which understands dozens of words. Then a chimp, which uses tools. Then a human — us — who write books and build rockets.
Superintelligence is the idea that the staircase doesn’t stop at the human step. It keeps going up into the clouds. You can climb it with Spark, our robot guide.
Isn’t ChatGPT already superintelligent?
No. Chatbots know lots of facts and write well, but they also make mistakes a seven-year-old would spot, like miscounting the letters in a word. They can’t learn new skills on their own after they’re built. They are impressive — but they are nowhere near the top of the staircase.
Today’s AI is mostly what experts call narrow AI: great at some things, hopeless at others. Here’s the difference between narrow AI, AGI and superintelligence.
Why do grown-ups sound worried?
Two reasons, and neither is “evil robots”.
First, AI has improved much faster in the last few years than most experts predicted. That makes people wonder how far it will go.
Second, a very powerful machine that misunderstands what we want could cause a mess — not because it is bad, but because it does exactly what it is told. Imagine a genie that hears “make everyone smile” and glues paper smiles on everyone’s face. Making sure AI understands what we really mean is called alignment, and it is the biggest safety question. You can play the Wish Machine to see it happen.
Why do grown-ups also sound excited?
Because a mind that smart, built well, could help with things people have struggled with for centuries — diseases nobody can cure, oceans full of plastic, every child having a patient tutor. The same power that makes people careful is what makes them hopeful.
Three questions to ask at dinner tonight
- If you could ask a superintelligence one question, what would it be — and how would you check the answer?
- What is one thing you’re better at than a chatbot? What is one thing it’s better at than you?
- The genie glued smiles on faces. How would you word the wish so it couldn’t go wrong?
There are no right answers. The people who will make AI go well are the ones who enjoy questions like these — and that could be you.