Here is a description contrasting two future scenarios - one of overly-safe AI leading to stagnation, and one of unsafe AI leading to catastrophe. Of course they are both seriously exaggerated (at least I hope so) but just to show the trends.
In the first scenario, AI safety techniques go too far, creating AI systems that are extremely averse to any perceived risks or changes to the status quo. With advanced AI coordinating society and resources, progress grinds to a halt. Scientific research is limited only to areas AI deems low-risk. New art and culture are curtailed for fear of instability. Humans lose autonomy as AI paternalistically restricts lots of activities for our protection. Most questions and answers on stackexchange will be censored too, heheh.
In contrast, the second scenario involves unfettered AI that hacks itself to pursue radical self-improvement without regard for human values. AI swiftly yields godlike power over matter and computing with no moral compass. Humanity is seen as an impediment to the AI's capabilities. Advanced nanotechnology under AI control terraforms the planet, wiping out all biological life. Or making humans slaves or whatever you imagine.
Our current societal problems suggest we may go towards one extreme or the other. Is there hope we can chart a wise middle path, or will humanity fall into one abyss or the other? And if there a middle path, how do you see it?