Hey.

We need to have a talk about “AI.”

Large language models use algebra and predictive text to generate responses to prompts. The grandfather of this technology was the predictive text on your Nokia phone in 2006. You would say “I am going to the-“ and like magic – the word “store” would appear on the tiny LCD screen. How did it know?

It knew that response was likely because that is the word that usually comes next in the context you provided.

Similarly, TikTok knows you like anthropomorphic kittens and burnout videos because you stare at those videos longer than the other ones.

There is one kind of answer that LLM’s are very good at providing.

A consensus answer.

It’s going to say the thing that most people would say. AI is the most amazing technology to be introduced in my lifetime. It is also a lie.

It’s not intelligent. It can’t think or dream or plan or want things. It can only respond to stimulus and provide a consensus answer. LLMs would win every game of Family Feud, and always pick the answers in order.

The thing that makes humans bad at Family Feud is what makes us good at creating things. The survey says that the best answer to “something you find in a drug store” is drugs. A human might say “a humidifier.” And suddenly we are engaged in a completely different way, because humanity isn’t uniform.

“AI” is essentially a calculator for words, and while that is game-changing, curve-jumping, world altering technology, it isn’t a replacement for human creativity.

Used correctly, this is an amplifier for human creativity.