While on a plane, my laptop ran out of battery. I was working in MATLAB. I switched to my iPad and started coding in Pythonista a simple program to produce points on a sphere. After half an an hour, I had coded a program that computed the coordinates of points along meridians. I needed this to test an algorithm in artificial neural networks that differentiates points belonging to two different concentric spheres. The source code is at the bottom.

When I landed, I sent to source code to my laptop. I planned to simply call the Python function from the larger MATLAB program. But once I ran the code, there was a syntax error due to differences in versions of the plotlib library. In the iPad, the following instruction works:

ax = fig.gca(projection='3d')

However, in my laptop, I had to use:

ax = fig.add_subplot(projection='3d')

Not a biggie, I thought. However, the the result of the same program running from a laptop was very different:

This is one of those “W.T.F?” or “What to Fathom?” moments. “Well,” I thought, “It’s just a matter of finding the correct syntax for the latest version of the library.” However, why bother? Generative artificial intelligence can code and translate across programming languages… at least that is the claim.

Enter ChatGPT. I entered the command “translate the following program into MATLAB,” followed by the program I coded on the iPad. The result was yet a third version of what I needed to achieve.

Here you have it. Three different platforms, “same” program, three different results. Of course, it is trivial so solve the problem in each platform. In the end, instances like this are the reasons why I stick to MATLAB for algorithm development… and ChatGPT has a long way to go.

# Points on a sphere
# By Juan B. Gutiérrez - biomathematicus.me
import math as mt
from mpl_toolkits.mplot3d import Axes3D
import numpy as np
import matplotlib.pyplot as plt

fig = plt.figure(1)
ax = fig.add_subplot(projection='3d')
#ax = fig.gca(projection='3d')

# initialize variables
stepCnt = 30
radius = 10
x = np.empty((stepCnt + 1,))
y = np.empty((stepCnt + 1,))
z = np.empty((stepCnt + 1,))
x[0], y[0], z[0] = (radius, 0, 0)
Theta = np.linspace(0, 2*mt.pi, stepCnt)
Phi = np.linspace(0, mt.pi, stepCnt)

for phi in Theta: #yes, I know... I swapped phi and theta during development.. ooops
    i = 0
    for theta in Phi:
        x[i] = radius * mt.sin(theta) * mt.cos(phi)
        y[i] = radius * mt.sin(theta) * mt.sin(phi)
        z[i] = radius * mt.cos(theta)
        i = i + 1
    ax.scatter(x, y, z)


ax.set_xlabel("X Axis")
ax.set_ylabel("Y Axis")
ax.set_zlabel("Z Axis")
ax.set_title("Points on a sphere")

plt.show()