AI can’t do sums, Part 1: Olmo-7b-instruct.

I have been testing assorted so-called Artificial Intelligence Large Language models on their ability to do a relatively simple calculation. I ask, “Calculate πφ/e where φ is the golden ratio.”

You can paste “π*φ/e” into Google’s search box  (it needs the * to indicate multiplication and not Greek text), and it will tell you the answer, 1.87000613369.

Wolfram Alpha does it more precisely, as shown:

Continue reading