But:

1. I don't want my AI to have super-human intelligence. It would generate code that I do not understand. A coder with human capabilities is better for me.

2. We're interested in AGI. If the AI reaches human intelligence, that's a milestone. Drawing bicycles with pelicans is superhuman. Hence not a relevant test.

"Drawing bicycles with pelicans is superhuman."

It really isn't. Many humans can draw a bicycle, and a pelican, just fine.