Increased efficiency also seems to be part of its appeal. The limitation is you can't increase efficiency by just upgrading computer specs, but instead have to find innovating ways to use the existing resources as efficient as possible to make something great. These kinds of optimization or compression problems seems like something AI would be very helpful for, so I think it is premature to try and ban its usage.
You can crib techniques from other people but unless you also show that you understand them deeply, e.g. by creative adaptions, you'll still be considered a lamer even though your results match those of someone else.
This is one of the reasons why the demo scene still has a lot of physical events, it's part of the socialisation process to be in the same room as other people, putting in your final touches while they observe and produce distractions that in practice validate your abilities and respectable refusal to take shortcuts.
Are you talking about some sort of hypothetical future Super-AGI?