You don't have to like the forecast to track it against the news reports. (Of course, the news reports are often just PR, and you need to weigh that in.)

I don't actually use Apple Store all that often. A fair portion of the software I have installed, like LibreOffice and Firefox is just installed via DMG images. It kicks up a window about unrecognized source, but then just works. iOS devices are definitely more locked down, but the Macs are really no different as far as installing software than Windows or Linux.

I imagine the Mac Neo is the real source of their panic. Right now RAM prices are probably saving them from even more losses, but the hegemony is coming to an end. If a credible useful, at least for average users, non-Windows platform using smart device level hardware can sell as well as the Neo has, I'd say Microsoft's reckoning is finally upon them.

Assuming that CGNAT makes you immune is a huge error.
Once you compromise a single customer you're now inside the CGNAT pool, where you will see lots of very vulnerable devices because they were left vulnerable on the assumption that they were not reachable. In an ISP with thousands of customers, at least a handful will have some infected devices.

Modern Windows devices absolutely do not become compromised via inbound connections to open ports, they become compromised via vulnerable client software or user error (eg phishing, malware infected downloads etc), all of which only depends on being able to make outbound connections.

At what point in this long and seemingly endless list of fixes to even the most basic usability features in Windows do its users finally admit it is really a shitty and badly maintained operating system. I use Gnome or MacOS, which are streamlined and uncluttered, and then I head over to Windows and it's like looking into the mind of someone with severe ADHD. It's a colossal mess where nothing particular makes sense, there's no coherent approach, everything is slow and inundated with advertising, context menus that worked for decades don't function right or at all, even the simplest tasks just seems to land you in the wrong place.

I suppose under the hood it's still a fairly decent operating system, although tools like Powershell, which can be achingly slow itself, demonstrate that there's a lot of layers of cruft.

I don't play video games, and frankly Office isn't that much better for my needs than LibreOffice, and Outlook is a bloated pile of crap, so I rarely even access the Windows desktop I have at work via RDP, save for two applications I rarely use. Windows is rapidly becoming irrelevant in my world.

You shouldn't be relying on perimeter security anyway, every device should be able to stand on its own in a zero trust scenario.

The russians have been blackmailing ukrainians in ukraine into committing acts against their own government, there are several documented cases of this.

https://www.theguardian.com/wo...
https://www.washingtonpost.com...
https://news.sky.com/story/how...

There have been numerous recently reported cases of russians gaining access to information on ordinary ukrainian citizens, and then blackmailing them to carry out spying or sabogate operations.
You absolutely do have to worry about what hostile foreign governments might do.

There are hundreds of ways an attacker could gain access to the inside interface, even doing so blindly via xsrf where you have predictable legacy addressing.

Many simply don't care.
A lot of ISPs especially in Asia use CGNAT and/or rapidly rotating IPv6 and then do nothing about abuse so the address space is widely blacklisted.

In other countries ISPs aren't forced to use CGNAT, and use at least sticky if not fully static addressing so if customers get themselves blacklisted the ISP generally doesn't need to care as it won't affect other users.

Setting the policy to DROP just means that clients will try multiple times before timing out, which means not only will you waste bandwidth with the retries, but your own clients will experience a delay while they time out instead of receiving an instant rejection.

For legacy IPv4 networks the address space is so congested and in short supply that it's economically unviable to leave unused addresses, so you gain nothing from this. With IPv6 there might be some very limited security-through-obscurity value to someone not being able to identify a live address, but its also not practical to scan sequential address space anyway.

What this article really highllghts however, is how flawed the perimeter security model is. Modern end user devices will actually do perfectly well on an open connection, as they don't have any externally visible services. Indeed people frequently connect their devices to public wifi networks where they are fully exposed to the network owner, other users and potentially beyond and it hasn't caused the apocalypse.

People are relying on the perimeter security model, and then using really lousy insecure devices to actually implement that perimeter so they get the worst possible outcome. User think their devices are inside a secured perimeter when the very device supposed to be enforcing that perimeter has been compromised putting the attacker inside. These devices are often MUCH worse than today's end user operating systems.

The proper solution is zero trust - assume your devices are fully exposed and have to stand alone.

The brainstem is probably rather invariant between people. That's an area where differences are likely to be fatal.

Actually, we're pretty sure it CAN happen by random chance. The argument is about how small that chance is...and what environmental conditions give the highest probability. And in what order things happened. (Currently I, not really knowledgeable in the field, think metabolism probably came before centralized information storage.)

YES!!
And *part* of the reason it's tricky is that we've got no real idea of what current AIs are going to develop into. Projections are all over the map.

FWIW, I tend to consider the AI2027 projections as identifying the most probably sequence, but there are lots of others, and none of them have any chance of being more than approximately correct. (The AI2027 people say it isn't moving quite as fast as they projected, but when I compare what they said against the news I'm not sure I believe them.)

Or let's put this another way. Show of hands - how many of you "spicy autocorrect" / "stochastic parrot" people had "AI will start mass-solving Erdos problems" on your forecast list a couple years back? Huh, none of you? Fascinating!

Take some time to reassess your priors. And while you do so, understand that, yes, they are doing logic / reasoning.