I am not afraid of robots. I am afraid of people

TBH, I'm more concerned about their imperfect control. At least if some group of executives has perfect control, then it's possible for someone to force them to turn it off. As with Satya having his team "fix" an issue in 24 hours, though, there's a dog-balancing-plates phenomenon with the current state of development.

the wealthy profit-maximizing executives paid for the engineers who trained the AI, and ran it through the legal department designed to protect the wealth and status of the corporation they are executives of. Under present circumstances, I don't know what an "autonomous" AI might even look like; someone must have trained it, and if another AI trained it, that was also at some point in the chain trained by these same institutional actors...

An AR-15 is a single weapon that can be used in a specific context by one individual to a limited degree. AI owned by major organizations is a highly complex thing that can be used by all sorts of multiple actors simultaneously, at scale, in all sorts of ways ranging from good, to iffy to ambiguous to dangerous to extremely dangerous. AR-15's in the hands of mentally unstable people or those with malevolent intentions of any age can be very dangerous things, but the nature and level of the danger is of a totally different, much more limited type to powerful AI tools in the hands of mentally unstable powerful people, or self.interested leaders and organizations..

Note; Edits for a bit more context.

AI doesn't have be out to get us to cause lots more social and economic instability (though ironically ChatGPT seems to be poised to replace a lot of white-collar workers - unlike the robot apocalypse predictions about replacing blue-collar ones... not that both can't happen simultaneously, of course).

>I saw a bunch of stories recently about how social media has increased mental illness problems amongst our children, yet I don't see the same level of alarm about that

wait til you see what AI will do to children's brains, once the last remnants of human interaction are excised from their lives.

(The chat boxes themselves could easily exaserbate the mental illness issues for children.)

One critical difference is the growth factor. AR-15 shootings don't double every year. Teen suicide don't double every year.

https://epochai.org/blog/compute-trends

Another is prevalence. This tech can be the ghost in the machine animating all sorts of processes, interfaces, devices, content. It will affect you at work, it will affect you at home, it will affect you in society. This is the entire point of developing machine intelligence. How safe, reliable, and controlable is this tech as of now? On that list of yours, there is a piece of technology (primitive technology btw) that is illegal to have: a fully automatic rifle. Calls for regulation and restriction are not unreasonable. How we define and enforce them is the challenge, but the need should be self-evident.

> the motives of these elites

And yet another reason. The powerful variants of this technology has very high financial barrier to entry. The sort of MI that could truly fuck with us is only in reach of nation states and companies like Google, Microsoft, etc. The latter set (we are repeatedly told here and elsewhere) are only obligated to further the profits of shareholders.

A pause also lets the emerging wave of open source and operationally accessible projects (such as *.cpp, etc.) to catch up a bit. This disempowers the "elite" but it is an undeniable plus for the plebes.

The cumulative impact on society of all of the violence that happens in the United States is likely lower than, for example, the invention and wide distribution of the smartphone. AI is likely to have a larger impact than that.

You're talking about only one small part of that violence.

Nobody is banning assault rifles? Aren't school shootings extremely rare in countries that do ban AR-15s?

Yep. And the issue is that because AI will depend on some form of expensive ASIC, AI compute and knowledge storage will be heavily centralized, guarded and sooner than later we'll start seeing rentseeking behaviour once again.