I suggest you read George Gilder's Life After Google.

It has a very detailed, debunking of AI ever reaching true sentience based on the mathematical proofs of Godel which are the basis for computer science itself.

If you read Homo Deus by Yuval Noah Harari then you will see he doesn't think it will end well for us, but I would hope we can co-exist with 'intelligent' machines. These latest systems are good at processing large amounts to data to extract information and present it in similar ways to how a human would, but I do not think they are by any means self-aware. It may be a big step to that, but we can be sure they will improve. There are applications for smart machines in autonomous space probes and for hazardous jobs, but if they are cheaper than humans then they will replace them in some workplaces. It is worrying when they get armed as has already been done for security situations. They will make mistakes, as humans do. But what if some dictator can unleash swarms of small bots carrying explosives into a neighbouring country or against protesters? That is scary.

I would hope any company working on this has people looking at the ethical issues. These are indeed interesting times. The genie is out of the bottle.

I recently posted research that indicates consciousness, or whatever it is we're trying to refer to here (which we remain incapable of defining), isn't a classical phenomenon.  In other words, aspects of our person are emergent quantum effects, and mere neural networks contrived to mimic those in our brains aren't mimicking the structures or aspects of living things from which consciousness arises.  Those networks are wholly classical, and no aspect of them is adaptable to avail them of whatever quantum properties living creatures have that infuses them with persons.

While physics is a distant slog from where my education lies, it's not difficult to grasp the outline of the thesis, and it is apparent from any rational consideration of the issue that the actual devices we build and employ to facilitate handling large datasets have almost no affinity or semblance to living things we consider conscious.

Again, the primary source of certainty that AI isn't actual intelligence is that intelligence is an aspect of a quality - consciousness - living things have that people really cannot define, do not know where it comes from, how it arises, or even what it is.  How could we build a car without knowing what moving is?  Could we build a plane were we unable to define flying?

It is our very nescience regarding consciousness that underlies any fear AI could be itself problematic.  Every indication is presently that extant development of LLM's and other neural network devices is being used by people and corporations for their aggrandizement, and that this is the reason these things are problematic.  Exactly as weapons, cars, or toxins are dangerous, it is not the things themselves that are dangerous, but what people do to other people with them, if those victims do not protect themselves from such vectors for harm.

What's called AI today isn't intelligence, conscious, or capable of being developed to be, but is simply processing of large datasets with very large processing devices that enable people to do beneficial or malicious things with that information.  Our vulnerability to such processing derives not from the devices, but from the acquisition of information about us that is processed and used against us by corporations.  My best assessment of how to best secure ourselves from such risks is to continue to develop and decentralize such technology until we have access to it and can deploy it.  No one is more vulnerable to financial data than those with the most financial assets.  Their singular possession of these data handling capabilities derives from the cost of such large processing devices, and advancing technology always decreases that advantage over time.

There are folks here on Hive that are purchasing GPUs and deploying neural net training software to decentralize this technology today.  While the threat seems terrible, Moore's Law suggests it will be very short lived, and that it will soon be used against the censors and propagandists that depend on that informational edge to maintain massive financial advantage to eliminate that inequity.  

Decentralization is accelerating in every industry as technology advances, and data processing more quickly precisely because it is the most advanced technologically.  

Thanks!

Edit: while it's a bit more technical, quantum computing has no relation to quantum consciousness, but is simply a means to more effectively process data, and our use of quantum physics in that endeavor reflects no understanding of how consciousness emerges from living things.