Why are we using LLMs as calculators?

LLMs turn computing on it's head.

Instead of reliable results at low cost, they produce unreliable results at high cost. And they can't do math and logic.

The really scary thing about all this is the number of seemingly intelligent "leaders" who are convinced that an AI apocolypse of some sort is imminent --- from something that fails 3rd grade math.

Why indeed. When you can just instruct ChatGPT to use a calculator.. the stateful python environment is more than capable of accurately performing the most complex mathematical functions you can throw at it, keeping in mind compute and time limitations. Even then, ChatGPT can construct a Jupyter notebook that you can run locally.

Those of us with the slightest understanding (and I don’t know a lot) know that they don’t think, reason, or understand logic. They’re statistical word generators only. That’s why I don’t use them for math. I suspect most others here have similar reasons.