The current approach in which ChatGPT is trained likely allows for that. ChatGPT keeps claiming it is a "language model". In fact it is a reinforcement learning agent trained with proximal policy optimization. We've certainly seen reinforcement learning agents (trained with PPO) interacting with what happens on screen (such as playing StarCraft, etc) and outplaying best human players. So yes, I expect we'll see a lot of interesting stuff in the next few years.
And orbital infrastructure... Sounds impractical. Even if somehow possible, it'd be risky to have a dependency on such infrastructure. Fragile and susceptible to so many risks.
The difference between chemical/fission and fusion are orders of magnitude. It is <10km/s for chemical/fission. And 1000km/s for fusion:
"Today’s best chemical rockets produce propellant exhaust velocities up to 4.5 km/s. Fission (nuclear thermal) rocket engines could roughly double that, to about 8.5 km/s. A working nuclear fusion core producing thrust through direct exhaust of hot plasma, could readily generate up to ~1000 km/s exhaust velocities" - https://link.springer.com/article/10.1007/s10894-015-0034-1
Likely necessary for advanced space flight. Likely necessary to allow for "clean energy" on earth. Unlike advances to flamethrowers.
Using these resources to improve chemical flamethrowers is sub-optimal. A dead end. Nuclear fusion is a more likely answer to spaceflight and Earth's clean energy problem. It likely should be the priority.
Could tweaking the profit function result in a better alignment? For example, issue a fine for every confirmed sale that reached someone in the wrong age group. Upon the return of the unused product direct the proceeds of the fine back to the person that made the purchase.