The Wow I Can't Believe That News Story Thread

[Frownland]

So we're working with different definitions. I thought you were just referring to AI which doesn't apply to your arguments given the basics of how it functions.

Having met humans, human intelligence in a box is definitely not a threat. It's how intelligent the box thinks it is that's the threat. That said, you don't think we can develop degrees of separation where one AGI system will determine that x and y are the right choices, while humans are the ones who decide to actually implement x and y? What would AGI self interests look like that would make that control necessary?

[jwb]

Originally Posted by Frownland

So we're working with different definitions. I thought you were just referring to AI which doesn't apply to your arguments given the basics of how it functions.

Having met humans, human intelligence in a box is definitely not a threat. It's how intelligent the box thinks it is that's the threat. That said, you don't think we can develop degrees of separation where one AGI system will determine that x and y are the right choices, while humans are the ones who decide to actually implement x and y?

I did make sure to specify AGI from the beginning, which had the specific definition I'm using, as even your article notes

And no, I don't have any faith they will contain it. Given that virtually everything is networked these days and they can't even really properly secure the systems we use right now, I believe our advances in technology develope at a faster rate than our ability to control said technology.

[Frownland]

Originally Posted by jwb

I did make sure to specify AGI from the beginning, which had the specific definition I'm using, as even your article notes

And no, I don't have any faith they will contain it. Given that virtually everything is networked these days and they can't even really properly secure the systems we use right now, I believe our advances in technology develope at a faster rate than our ability to control said technology.

What self interests would AGI use its autonomy to preserve?

[OccultHawk]

Originally Posted by Frownland

What self interests would AGI use its autonomy to preserve?

We don’t know but we do know that everything we understand to be sentient is also inclined to survive and propagate.

[jwb]

@ OH

Exactly. It's a complete gamble. And technology proliferates in a decentralized ad hoc sort of way so even if there is a specific way to harness AI without it ever threatening us, it's hard to ensure that nobody is going to go outside those parameters and create something that could pose a threat.

There's an inherent incentive to create AGI because of the inherent utility of intelligence and seeing that AGI would match and eventually surpass human intelligence in every domain, that would include creating new, even more effective forms of AGI.

So with each successive generation of AGI, the engineers creating the next generation grow smarter and smarter because at the point the engineers themselves are robots.

This creates a feedback loop that allows for an exponential growth in AI that humans will not be able to keep up with because we are working with inherent biological constraints.

@ elph

There's no reason they shouldn't be able to, if we are ourselves biochemical machines. Without invoking magic to explain human intelligence it's hard to imagine that it's impossible.

But yes, this entire argument is predicated on the assumption that AGI gets developed. If it doesn't then obviously it doesn't pose any threat.

[OccultHawk]

Originally Posted by elphenor

We don't even know yet that AI can ever mimic human intelligence

To merely mimic would just mean to pass the Turing test

No two intelligences on earth are really comparable be it plant, insect, microbial, or animal (including human). Whether a machine’s intelligence is even comprehensible within the parameters of our consciousness is likely irrelevant to the amount of harm it may inflict.

We have a very ˈspeciesist definition of what intelligence is. We sure think we’re smarter than mosquitos with our books and buildings and all but if Kafka turned you into one you’d be the dumbest mosquito in town. And boy they sure can kill some folks.

[jwb]

@ OH

That's a good point as well. Certain insects like ants and bee colonies have a collective form of Intelligence that makes them har capable of making complex decisions as a group without any individual ant or bee being conscious.

They're all basically robots following very simple rules which collectively results in complex decision making as a group. We have a very hard time controlling or containing the spread of these insects as it stands, even with the supposedly limited Intelligence they have.

That's a system that was developed ad hoc by trial and error via evolution. It's easy to imagine with direct engineering we can develop systems far more advanced and potentially menacing.

@ elph

We won't "know" until it happens. But the general trend that AI us advancing and is already surpassing humans in specific domains of Intelligence such as computers playing chess, go, etc.

[OccultHawk]

we don't totally understand how human intelligence works,

We don’t understand how any intelligence works.

capable of making complex decisions as a group without any individual ant or bee being conscious.

We don’t know if they’re conscious. I personally think they are.

[OccultHawk]

Why are you so obsessed with the human brain as some kind of benchmark?

[OccultHawk]

Originally Posted by elphenor

I think it's obviously the most impressive thing on the planet

bees can get ****ed

scientists are going to make robot bees

Obvious to a human brain that thinks like a human brain.