Within another 5~10 years I doubt most people will be able to distinguish it from something a real person created at all.
Maybe, or maybe not. The issue with machine learning software is that despite what they are marketed as, there is zero intelligence in them. They basically work via trial and error, and to know they have errored they must have references. And with an increasing number of generated content flooding the internet, differentiating the real reference material from generated material will be difficult, as in not cost effective. So we will end up with generated content teaching the model to generate content, and the progress we have seen will be effectively halted.
Maybe, or maybe not. The issue with machine learning software is that despite what they are marketed as, there is zero intelligence in them. They basically work via trial and error, and to know they have errored they must have references. And with an increasing number of generated content flooding the internet, differentiating the real reference material from generated material will be difficult, as in not cost effective. So we will end up with generated content teaching the model to generate content, and the progress we have seen will be effectively halted.
The big philosophical question (with practical implications) is, wether our intelligence evolved differently.