I'm not as concerned simply due to the nature of this sort of AI.
It's like when people train a dog to drive a car or teach a chimp sign language. None of the knowledge or ability of these AIs is actually independent from human thinking, it's all downstream from ideas humans had.
The real threat is deep fakes and the post truth era. The gpt stuff is just a silly distraction from the fact you can now make a nearly convincing recording of the president saying whatever you want and nearly convincingly lipsynch that recording over a video of him speaking. That sort of AI is only going to get more powerful and deceptive, and what's worse is by design such fakes cannot be detected without a significantly more powerful AI. Such an AI can easily be used to create an AI which can trick the detection AI though, so really this shits just going to get harder and harder to detect.