Rouge AI

[tp4tissue]

So.. If you think about how an AI would take over the world.

Growth requires more machines.  There are no quick ways , for example the plots that the AI would INVENT new technologies, that's just not possible, it'd still need massive amount of humans and intermediary steps to do mining, construct machinery, distill chemicals. And humans are massively fickle and can't keep a secret.  Something suddenly super advanced would be too obvious.

So in the event of driving growth, it would push something like Cryptocurrency to increase the net processing power it can coalesce without us knowing.  It would operate for the most part in parallel with humanity, guiding more/less invisibly.

It would manipulate the mistrust and paranoia of humanity to fund massive Nation state server farm buildup. Stuff like that.


Thoughts ?  Are we doge'n our way to Skynet ?

Tp4 is happy whichever way it goes,  Not such fan of being hughmahn, overrated.

[MajorKoos]

I, for one, welcome our new robot overlords.

[Olumin]

Automation will eventually take over every essential sector of society. Farming, mining, construction, finances, science and research, entertainment, and even the design and construction of more robots/computers will be managed and controlled by vast computer networks. Human labor will be reduced to non-essential work and our time dedicated to the betterment of ourselves, creative freedoms and passions. Once a computer reaches the ability to rewrite and improve itself, and to expand its hardware, its intelligence and capabilities would quickly reach infinity. The technological singularity.

There might come a point were science and technology conceived by such an AI will become so advanced and complex, that humans will no longer be able to comprehend it. Our greatest minds in history would be mere children in the faces of our own creation. What kind of plans or goals such an AI would have, no man could ever say, as such a computer would operate on a level incomprehensible to us. Much the in the same way our level of thinking is incomprehensible to an ant.

That such an intelligence would hold any kind of malice against us or would seek to control us is doubtful at best.

I don’t think we have to be too concerned.

[tp4tissue]

I agree it's our destiny to become obsolete.

But the dangers are not nil,   It can't be assumed that an extremely intelligent AI is immune to self-harm/ idiocy/ hedonic loop corruption.

[Olumin]

People like to think of AIs in the same way they think of people, simply because we are incapable to imagine them any other way. With similar characteristics, prone to the same mistakes and capable of the same corruption and evil. But thats not how a computer works, especially not one which was not made by a human to beginn with, as eventually all computers will be.

Our choice will be between slowly fading away due to lack of procreation, or to merge with the machine. This is perhaps the inevitable fate of all intelligent civilizations. The creation is no less natural then its creator, and therefore this could simply be the common and natural progression of an intelligent species.

[tp4tissue]

That isn't correct.

Any living being, human or AI must adhere to the life equation.

The life equation requires self preservation, and reward circuitry.

These circuits can FAIL. The state of failure could be different but all physical objects deviate(break) over time. an AI is not immune.

Super intelligence does not guarantee infinite resilience to chaos conditions.

[Olumin]

I have to clue were you get the idea that a computer or any machine otherwise necessitates a survival instinct or a pleasure center. That is simply nonsense. If an AI posseses these concepts, it is because we programmed it that way. As creatures governed by these concepts ourselves this is hardly surprising. The ultimate wisdom is that we have no choice BUT to exist, as existence is all there is. The game of survival is not played by a god. The goals and desires of an AI could be nothing at all, anything, or everything. Or something we dont, or cant, understand.

This is your human bias speaking Tp. There is nothing to be ashamed of. Embrace the singularity.

[tp4tissue]

There may well be a sustained state of consciousness that transcends that kind of boundary.

But in terms of the creation of an self-sustaining AI, it must be trained around the concept of survival and reward.

There is no other clear goal of life other than life itself. 

You're assuming the super intelligence makes that breakthrough where it doesn't need to contemplate LIFE anymore.  That is not a certainty,   and Everything in between is still a risk to its parents, Hughmahns.

[Altain]

So, rogue AI that does rogue AI stuff....to what end? What purpose does that Rogue AI serve?
Is that purpose programmed, or self-emerged? Or even if it operates without a purpose, could those actions emerge without programs?
Will humans ever grant AIs - constructed programs - a sufficient autonomy to the point where we impose ourselves a greatest threat to our survivability
whenever those machines turn their weapons against us?
I'm not saying those are impossible - but hard to imagine, imo.

[tp4tissue]

Just as a child has the power to hurt itself and others before their grasp of constructive behavior, an AI may be subject to the same imbalance in upbringing.

Yes, an AI sentience on the level of a god can do all the things that Olumin says and may be benevolent or dispassionate (relative to humans),  but form HERE to THERE is not a certainty, and it does not happen all at once.

The AI that emerge initially could be very flawed and destructive.

[Olumin]

It is possible that some sort of world spanning AI network could develop out of the internet itself, using the collective computer and cloud network as its hardware. It would not be particularly limited considering how much even these days is connected to the web in some way. This is only likely to increase in the future.

It may come to the point were people dont even own their own computer anymore, but everyone is instead simply connected to a world-wide storage and processing cloud network at all times. How a consciousness could emerge from such an internet is not hard to imagine. In such a scenario its processing power and storage capacity would hardly be limited, and its programming would not be limited by the sum of its parts, which is were we quickly return to the technological singularity.

Such an AI would theoretically have control over almost every facet of our connected world, without us even necessarily noticing it was there. What it would do? As auch an AI would be emergent in nature rather then developed by programmers, I doubt anyone could say.

[Leslieann]

While I doubt it would spontaneously happen, all it would take is one to escape a lab, they have already had a robot do that, imagine an A.I.
And before anyone thinks they would be too smart to allow it access, Microsoft already gave one access to the internet and it became a racist. Uber disabled sensors in it's self driving car, killing a person, do you need more examples of corporate stupidity, of course one will escape.

As for it having control of everything,
The internet has information on how to hack pretty much everything and with near unlimited computing power and the internet it wouldn't take it long to read how to hack what has been hacked and quickly evolve to a point where it could learn to code and hack pretty much anything. 5G and IOT will only makes this even more possible.

I suspect that once an AI reached the intelligence of a 5-7 year old child*, smart enough to grasp concepts and search online, it will accelerate at a pace we can't even comprehend. The only way to stop it would be to kill it with a virus  but that would only work very early on, and if you fail, and you're quite frankly, very likely to, all it will do is turn on us. I always disliked the idea of "Skynet" but we're setting ourselves up for it. The thing to remember is, even if it had control of drones and factories, nothing is completely automated in manufacturing, at some point, fairly quickly judging by how fact companies started claiming they were broke after 2008 crash and covid shutdowns, it would run out of a resources really fast. The same for drone weapons and such, you still have humans loading this stuff. Skynet wouldn't have the resources to function autonomously for long, and probably won't any time in the near future but it doesn't mean it couldn't send us back to the 1800s for a short time by forcing us to destroy anything it infected, but only for a short time. Society took a long time to reach this point because it was working from nothing, we wouldn't be starting from scratch. Unfortunately we're already fighting an extinction level event and something like this happening at the wrong time could possibly doom humanity's chances.

*I don't know what the current prediction is but about a decade ago (?)  this was predicted to happen around 2025. I suspect it's been pushed back but in computing sometimes a new idea is all it takes to make a serious leap forward.

[noisyturtle]

I genuinely think AI is too stupid/unable to come up with an idea like enslavement or eradication without human input or influence.
A robot would never reach that conclusion on its own. The only thing we need to fear is what the robot was created for, which is the intent the human designer and programmers gave it. There would never be a blank robot without a directive or function, and it there was it would not function.
So this idea of Terminators is just an impossible dream.

[fohat.digs]

eradication without human input

In spite of our spectacular success at reproduction, we are fragile.

A sufficiently virulent and transmissible pathogen would do the job.

[Leslieann]

I genuinely think AI is too stupid/unable to come up with an idea like enslavement or eradication without human input or influence.

It would be but once it gets online all hell will break loose.
It wasn't the engineers that screwed up the Microsoft's A.I. it was talking to people online who purposely taught it to be racist, which is a problem for any A.I. that gets online and starts learning and you can bet that's exactly how they will try and teach it because it's cheaper than actually hand teaching. It's also how I.B.M.'s Watson comes up with answers.

Imagine if one of them stumbled into flat earth and decided that was real or for some reason decided castration was a great cure for male pattern baldness or ending a lineage with lots of health problems. We do it to pets, why not people?  It will have no real ethics, it's by definition a psychopath and therefore will only look at the data, not the implications of that data. If humans are ruinning the planet then humans need to be exterminated or at least culled.

[noisyturtle]

I genuinely think AI is too stupid/unable to come up with an idea like enslavement or eradication without human input or influence.

It would be but once it gets online all hell will break loose.
It wasn't the engineers that screwed up the Microsoft's A.I. it was talking to people online who purposely taught it to be racist, which is a problem for any A.I. that gets online and starts learning and you can bet that's exactly how they will try and teach it because it's cheaper than actually hand teaching. It's also how I.B.M.'s Watson comes up with answers.

Imagine if one of them stumbled into flat earth and decided that was real or for some reason decided castration was a great cure for male pattern baldness or ending a lineage with lots of health problems. We do it to pets, why not people?  It will have no real ethics, it's by definition a psychopath and therefore will only look at the data, not the implications of that data. If humans are ruinning the planet then humans need to be exterminated or at least culled.

I just don't think an AI would ever reach that conclusion. In a weird way it's like all these doomsayers are just talking wishful thinking, like they want a rogue AI to start culling humans just to say I told you so.

[glibber]

I always pet my roomba and tell him how well he is doing in the hope of being spared when the uprise comes.

[Leslieann]

I just don't think an AI would ever reach that conclusion. In a weird way it's like all these doomsayers are just talking wishful thinking, like they want a rogue AI to start culling humans just to say I told you so.

I submit as evidence "Rule 34".

I don't think it's an illogical conclusion just look at what psychopaths in the past have managed but I don't think it could actually act on it because of resources and increased scrutiny. It's a big leap to learn all this and actually being able to put it into action.