The problem lies in how you ask the question. Use the following structure to help you come up with an accurate reasoning:
Claire has 6 brothers and each of her brothers have two sisters. How many sisters does Claire have?
The logical connective "and" without a comma establishes the relation that both premises must be given.
yes, prompt engineering is an entire field on it's own right now. A big part of working with AI is figuring out the best prompts, and this varies model to model.
As a human, I don't have to play around with how I ask questions like that.
The real problem is that the common user does NOT know how to ask the question. Logic, mathematics, ..., have an exact way how to word it. A misplaced as, an accent and even a grammatical error, can confuse any human.
If I were to ask you a question like this: "How many bricks does it take to complete a building made of bricks?", neither you nor the AI can give me the answer and neither can any engineer. To solve this question requires data, specifically dimensions of the bricks and dimensions of the building... The way we formulate the question, the information provided, the logic and the structure of the sentence, count for anyone, human or not.
If I were to ask you a question like this: "How many bricks does it take to complete a building made of bricks?", neither you nor the AI can give me the answer and neither can any engineer. To solve this question requires data, specifically dimensions of the bricks and dimensions of the building... The way we formulate the question, the information provided, the logic and the structure of the sentence, count for anyone, human or not.
This though, is a perfect question Google would ask on an interview to test how you solve problems. Questions like how many golf balls can you fit in a limo are good examples of these type of questions.
The question I asked though is easily resolved with a human, but with AI they just miss the point.