AI vs Engineer

I recently found the following simple logic puzzle from Dell Logic Puzzle (1998), and I was wondering if ChatGPT would be able to solve it:

The explanation is not convincing, and the result is wrong. Even pointing out that each child needs to be between two and six years old did not help, as ChatGPT then computed the same age for two different children.

From other posts I know that ChatGPT can be a powerful tool for various Python programming tasks. So I asked the following:

The result of running ChatGPT’s Python program was even less convincing:

Instead of starting to debug the Python code, I decided to write my own Java program using the powerful OR-Tools from Google. I wrote the babysitting puzzle as a simple constraint optimization problem that is solved within milliseconds:

The conclusion: Engineering still beats AI, even for very simple logic problems.

The source code may be found on Github.

Adrian Trapletti
Adrian Trapletti

Quant, software engineer, and consultant mostly investment industry. Long-term contributor and package author R Project for Statistical Computing.