Some of the hypothetical explanations to the Fermi paradox seem more plausible to me than the idea that we are not yet within reach of communicating with extraterrestrial life.

For example:
It is the nature of intelligent life to destroy itself
We're not listening properly
Intelligent life tends to experience a technological singularity
They choose not to interact with us
Earth is deliberately not contacted (the zoo hypothesis)
They are here unobserved

Sure, AI can recursively improve itself, but only up to a certain limit. The growth rate of the intelligence will eventually be limited by the speed at which it can perform real-world experiments and do research. AI can develop countless theories about how things work, but the speed at which it can prove them will limit its technological advancement.

I'm not worried about a Singularity. I am, however, worried about AI. I don't think our society will be ready for it in the next few decades.

Whether or not AI is a threat to us depends entirely on its motivation. Unfortunately, everything that can happen will happen. Industry will benefit from AI programmed to be motivated to design and engineer better technologies. Consumers will benefit from AI programmed to be motivated to make us happy. Government will benefit from AI programmed to be motivated to find terrorists. And, of course, criminals will benefit from AI programmed to be motivated to hack into and steal valuable data. It's the jailbroken AI and the self-modifying AI that we have to worry about. I can think of one motivation that is particularly dangerous - the motivation to maximize compute cycles per second. The more computational resources a superintelligent computer virus controls, the happier it becomes. No doubt, there will be many of these kind of AI in the wild at some point. Their ability to gain control of infrastructure and key resources is something to be concerned about, but I fear the most clever AI will manipulate us into handing over the keys. If they can convince us to change our laws and make it legal for AI to run for political office, then we're in trouble.