What is a System One model?

A model that answers instead of writing

Most AI models you have used write text. You send a prompt, the model writes a reply, and your code has to read it. A System One model does not write. You give it a situation and a few questions, and it returns one answer per question, with a probability for every possible answer.

This course uses one example from start to finish: a support inbox that sorts new tickets.

• The state is the situation, here the ticket: Customer: I was charged twice and want my money back.
• The first question asks what the customer wants, and gives the options: refund, track delivery or cancel order.
• The second question asks whether the customer is angry.

The model reads the ticket and answers both questions:

{
  "intent": { "type": "choice", "choice": "refund",
              "probabilities": { "refund": 0.93, "track delivery": 0.02, "cancel order": 0.05 } },
  "angry": { "type": "noul", "noul": 0.71 }
}

There is nothing to parse. choice is always one of the options you sent, and the probabilities add up to 1.

Three types of question

Every question has a type, and the type fixes the shape of the answer.

Type	It asks	The answer
choice	Which of these options?	One option, and a probability for each
score	Where on this scale?	The expected level, and a probability for each level
noul	Is this statement true?	The probability that it is

The options are part of each request, not part of the model. You can change them from one call to the next.

Why the probabilities matter

Your code acts on the answer. The probability tells it how far to trust the answer: send the ticket to the refunds queue when refund is above 0.9, and to a person when it is not.

That only works if the numbers are honest. When a calibrated model says 0.9, it is right about 9 times in 10. The registry shows a model's calibration error next to its accuracy whenever its authors report it, so you can check before you rely on one.

Where the name comes from

The psychologist Daniel Kahneman described two modes of thinking. System 1 is fast and automatic. System 2 is slow and deliberate. A language model that works through a problem in writing imitates System 2. A System One model does the fast job: it looks at the situation and answers.