Raw LLM Responses

Your idea is genius and the core concept can be used to turn the tables in a very beautiful and unexpected way. The ai depends on stealing pre existing data and information. We can poison the ai as you already told in the video, but in a more fundumental way. If the poisoned Trojans were unidentifiable even with advanced technology (so they cant simply filter them out and steal everything else but those with an additional ai) and if the error correction is chaotic and pattern-less. So the fix would need to be manual and not possible using automated algorithms to "clean" up the poison. Ai is logic gates, with some pseudo rng. It depends on patterns (like find the pattern in 1,2,1,2,1,2...) but not the pattern in chaotic stuff (like 2,0,6,99,3,3,1). And manual work means Paying employees. And these companies hate paying employees, they want as much automation as possible (capitalism to the max) . So the cost of the employees that would need to fix the algorithms manually could exceed would exceed the actual incomes. Im not an Economist just some thoughts. Maybe nonsense maybe genius. I guess if the internet were to be filled with those poisoned images ai companies would isolate their reference to private training data, but even thst would be a HUUGE step. Instead of stealing your and my art they would sorely rely on their private offline datasets, to avoid corruption due to internet being full of "corrupted" stuff. And last time I checked ai employees aren't the best artists, hehe.