[ExI] A Realistic Scenario of AI Takeover - Minute-By-Minute

[Adrian Tymes]

On Tue, Oct 21, 2025 at 11:39 AM John Clark <johnkclark at gmail.com> wrote:
> On Tue, Oct 21, 2025 at 9:13 AM Adrian Tymes via extropy-chat <extropy-chat at lists.extropy.org> wrote:
>> >They said "everyone dies".  "Everyone" means "everyone", not "every human on earth who is important in this matter".
>
> They said "everyone is online or is trackable online" and that is equivalent to  "every human on earth who is important in this matter".

It is not.

"Everyone", without limitation, is "everyone" without limitation.

Many important people throughout history have overlooked this.  "This
tragedy would hurt everyone (in power), so everyone (but actually just
everyone not in power) must be willing to die to prevent it" has been
used, time and again, to sacrifice peasant/serf/slave lives en masse,
such as in wars.  Even today, such cries are a hallmark of the corrupt
- of those with no concept of the lives of most of humanity.
Protesting against such thinking was part of what this past weekend's
No Kings protests were about.

And yes, I do see "we must stop all (except for our) AI research or
'everyone' dies" partly in this light.  It's more well intentioned
than most, but it's still fallacious and should not be heeded.

It's kind of saying, "Some of you may die (from problems that AI could
have solved, who wouldn't be killed by this AI that I worry about
since you are not important), but that is a sacrifice I am willing to
make."  https://knowyourmeme.com/memes/some-of-you-may-die is the meme
I reference here.

>> > Being smarter is not omniscience.
>
>  True it doesn't mean that, but it does mean something, it means you can't outsmart it.

You have much lower odds of outsmarting it.  0.0001% is not 0%, which
matters a great deal in this case (given the high number of potential
AIs, even if you completely dismiss all humans involved).

>> > Even ants can get around human eradication attempts at times
>
> Yeah and you could say the same thing about the Covid virus, but that doesn't mean it is intelligent and it certainly doesn't mean it has outsmarted us.

Perhaps not the virus, but even ants have their own - low - form of
intelligence, especially collectively.  It's far from human-equivalent
and lacks several capabilities we have, but ants are more intelligent
than viruses.

> > humans might have further upgrades they can give.
>
> Not if the AI is better at doing AI research than the entire human race combined, and it's only a matter of time before it will be.

I refer to qualities rather than quantities.  How can the AI prove
that it can come up with literally all the insights that humans ever
could?  Its earlier generations weren't capable of thinking in certain
ways that humans were.

>> > Also, what about arcades with more than 8 machines, each one of which has its own state-of-the-art GPU?
>
> I don't know, they didn't mention video arcades so you'll have to ask them because it is NOT my recommendation.

No, but you're defending their point.

> Maybe they would say it would be allowed but only if it was under constant surveillance 24-7 to make sure no AI was going on, or maybe they would say video arcades would be outlawed.

And how can you "make sure no AI was going on"?  The scenario as
postulated makes the AI effectively invisible to human monitoring.

I'd argue about video arcades being outlawed, but video arcades were
dying off anyway, and they're unnecessary for this secnario: as I
pointed out, distributed many-GPU collectives keep forming anyway,
with no need to be in the same physical location.

>> > And what happens if 8 turns out to be enough for such an AI to kickstart itself
>
> The answer to that is easy, then we are cooked. You're giving me reasons why you think their proposed solution won't work and there is some validity in what you say, but it doesn't refute their central proposition that if anybody builds it everybody dies.

Granted, this is a tertiary point: "Even if what you say is true, and
we should treat 'if there is any chance at all of this happening' as
'this is already happening' as you say, then by your logic it's
already too late and we're already doomed, so who cares?"