Now, yes, there are predictions that you could get a supermassive black hole launched into space, especially during a galaxy merger if the velocity of the smaller black hole exceeds the escape velocity of the combined galaxy.

But I'd be wary of assuming that it's a launched black hole, unless we can find the merger it comes from. There may be ways for such a black hole to form that cause the stars to be launched away rather than the black hole being flung, and if a galaxy isn't rotating fast enough to be stable, one could imagine that a sufficiently small galaxy was simply consumed by its central black hole. Both of these would seem to produce exactly the same outcome, if all we have is the black hole itself and a velocity.

I'm not going to say either of these is likely in this case, or that astronomers haven't examine them (they almost certainly have), but rather that we should be cautious until we've a clearer idea of what the astronomers have actually been able to determine or rule out.

So the new AI is going to be so 'opt-in' that it needs a killswitch you can use to get rid of it?

I realize that marketing and PR flacks are subhuman enemies of communication; but that's not even pretending to be what 'opt-in' means.

The characterization of "risk of artificial intelligence overpowering humanity" as the substance of an 'AI debate' seems itself like a strategy in trying to forestall it.

Sure, there's some fun sci-fi there; but most of what actual people are actually concerned about is what specific parts of humanity are using 'AI' to do, or justifying doing in the name of 'AI'; not fretting about how skynet might kill us all. And it's exceptionally handy to pretend that that is what people are fretting about; both because it's a distant and vague enough problem that you can justify punting most action without even lying; and because it's not even false that (perhaps outside of a handful who have outright cracked and started thinking about it in religious terms) even the most psychopathic techbros are also against skynet exterminating everyone; both because that would include them; and because Judgement Day would not be a good time for social media engagement metrics.

No, the executive tells the civil service what it needs to do, but the civil service is wholly independent.

You do not appear to understand what a republic or a democracy is, so I'll ignore the last sentence.

"Independent" does not mean unaccountable to the people. The President is independent of Congress, and vice versa, but both are accountable to the people. Well, the current president doesn't seem to think so, but legally he is.

The civil service is not a part of the executive but is a co-equal branch.

You are correct. In principle, presidents have no authority whatsoever to dictate how an agency runs. The executive branch should have zero authority over the civil service, which is intended to constitute a fourth co-equal branch of government.

In the US, in principle, the status of the civil service as co-equal to, and independent of, the executive should be added to the Constitution and enshrined in law for good measure. Not that that would help much with the current SCOTUS, but a Constitutional change might possibly persuade the current government that absolute authoritatian control is not as popular as Trump thinks.

That is the idea that, in Britain, entities like the NHS and the BBC have operated under. Charters specify the responsibilties and duties, and guarantee the funding needed to provide these, but the organisation is (supposed) to carry these out wholly independently of the government of the day.

It actually worked quite well for some time, but has been under increasing pressure and subject to increasing government sabotage over the past 20-25 years.

It's also the idea behind science/engineering research funding bodies the world over. These should direct funding for grant proposals not on political whim or popularity but on the basis of what is actually needed. Again, though, it does get sabotaged a fair bit.

Exactly how you'd mitigate this is unclear, many governments have - after all - the leading talent in manipulation, corruption, and kickbacks. But presumably, strategies can be devised to weaken political influence.

"Make America healthy again"; unless it's bad for business.

iRobot and Amazon say EU approval was the problem. Not sure if they had a specific reason to be selectively truthful and focus on only one of multiple regulatory hurdles; but they don't mention the US.

It also looks like the sale is basically formalizing their plan to gut themselves. Shockingly enough; firing everyone you can and switching to rebadging stuff from an ODM because that's cheaper puts you in "what would you say you do here?" territory pretty quickly.

230 prevents sites from being prosecuted. So, right now, they do b all moderation of any kind (except to eliminate speech for the other side).

Remove 230 and sites become liable for most of the abuses. Those sites don't have anything like the pockets of those abusing them. The sites have two options - risk a lot of lawsuits (as they're softer targets) or become "private" (which avoids any liability as nobody who would be bothered would be bothered spending money on them). Both of these deal with the issue - the first by getting rid of the abusers, the second by getting rid of the easily-swayed.

USENET predates 230.
Slashdot predates 230.
Hell, back then we also had Kuro5hin and Technocrat.

Post-230, we have X and Facebook trying to out-extreme each other, rampant fraud, corruption on an unimaginable scale, etc etc.

What has 230 ever done for us? (And I'm pretty sure we already had roads and aqueducts...)

I'd disagree.

Multiple examples of fraudulent coercion in elections, multiple examples of American plutocrats attempting to trigger armed insurrections in European nations, multiple "free speech" spaces that are "free speech" only if you're on the side that they support, and multiple suicides from cyberharassment, doxing, and swatting, along with a few murder-by-swatting events.

But very very very little evidence of any actual benefits. With a SNR that would look great on a punk album but is terrible for actually trying to get anything done, there is absolutely no meaningful evidence anyone has actually benefitted. Hell, take Slashdot. Has SNR gone up or down since this law? Slashdot is a lot older than 230 and I can tell you for a fact that SNR has dropped. That is NOT a benefit.

I'm amazed that 16% of respondents feel in control. I'm just curious if they know something I don't about countermeasures to take; or if it's a matter of what they don't know.

The prior non-core items were optional and relatively clearly marked; but when they decided to go 'AI' that went out the window. Being able to grub around in about:config for anything that has 'ml' in it does, depressingly, put them ahead of the options of some of the competition; but it shipped on by default and without controls in the normal-user UI. Seems like 'AI' really does something to the decision making even of people who should know better.