In my mind, the issue isn't so much of artificial intelligence, as it may become "sentient", going rogue and wipe out or enslave mankind.
The bigger risk appears to be that of powerful humans will leverage this exponentially powerful capability toward the end the powerful always have: to exploit & oppress the lesser capable populace.
Pretty much any historically modern technology has also been used against people (as well as for their dual-use nature application toward civil/peaceful ends): printing press, telephone, cameras, air planes, television, nuclear power, computers, social media, air planes, etc...
If modern super-computing capability's most visible "accomplishment" was the front-running of retail investors by flash-crash causing Wallstreet traders, then why wouldn't Artificial Intelligence (among peaceful uses) not also be deployed for exploitation & oppression purposes toward general society? ("to extract value" as some business literature has it) How different is that versus the exponentially complex and incomprehensible legal code that imposes ever more consequences on people without them understanding what their rights and recourse are?
The real issues with the technology singularity (as potentially manifested with super-AI) are ...
- speed (for manipulations to happen so fast that they may fly under the radar of human perception, avoiding the opportunity to even disagree with what happens),
- complexity (pose as something else without human chance to look through the real intentions/behavior,
- imposed one-sidedness of actions (no effective feedback channel, no negotiation opportunity).
Computers, software, and websites & mobile apps are already being leveraged at grand scale to one-sidedly impose processes on average people. Call center software that doesn't give much human choice outside of its rigid menu, automated emails that cannot be responded to, phone auto-dialers spamming without any feedback channel, no customer service relationship that gives a customer a sense of true negotiation power. Dynamic pricing that changes with such non-intuitive logic, apparently to confuse the shipper so much that their primitive impulses are tapped to "secure" a price before it changes again.
Business has always been built around manipulation. The convergence of exploiting humans' basic instinct to judge & behave in social context, to crave validation & personal identity, and the state of the art technology enabling massive social reach (what is called now "social media") accelerates the manipulation effectiveness & power to levels previously only seen in primitive societies submitted to divine governance (often by proxy of the chosen onew/few) or fascist (controlled by fear/violence)
nation states.
The literature for marketing techniques is full of neuro-marketing and cognitive-behavioral manipulation to drive the quick sale - suppress any conscious thought, pull emotional strings, trigger impulses, bypassing rational judgment. Go look it up, it's out there in the open.
Given computers' long history of helping automate the tedious & complex, why wouldn't AI be used to further these self-serving *human* ambitions? That's what we really should be afraid of.
The speed, complexity, and uni-directional channels of highly sophisticated collective machine intelligence (the web, the cloud, the app, the cool tech-company-as-a-brand, the cult of technology & disruption as proxy for progress & freedom) make-believe an open/free/democratic society, while at the same time providing the ruling class with the same power & control over the masses as much cruder methods in times gone by.
The true risk of advanced AI used as a tool is that this time people may not even realize what is really going on, since they are cognitively & perceptively outsmarted, with the technology just being the intermediary, the tool for those with the ambitions. Those in power are usually paranoid enough about maintaining their power base, that they will make damn sure that their tools will not become independent enough to threaten their true masters.
Ultimately, the enablers for all this are the masses, who traded in their true freedom for convenience. Turned off their thinking capacity for comfort, willingly yielding access to the city to the Trojan horses.
The sophistication growth curve of AI is being met by a steadily declining human cognitive capacity, that increasingly prefers simplicity and homogenized media & communication (crisp clustering of preferences, and ever more polarized attitudes make manipulating behavior easier).
In the end, if everybody is so lulled into this new world, and the exploitation is such a seamless facade, perhaps it doesn't matter. If it just weren't for those that feel the economic inequality. But perhaps that "nuisance" will soon be optimized away by that glorious technical singularity.