[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/sci/ - Science & Math

View post   
File: 31 KB, 845x355, Salv16.jpg [View same] [iqdb] [saucenao] [google]
10257924 No.10257924 [Reply] [Original]

I've heard it said that in the advent of an AI takeover, there would be no killing machines because AI can't create anything. They could launch missiles (maybe) or make robots in factories flail around dangerously, but they can't create anything new, like a Terminator robot. How true is this? Would even a wickedly intelligent AI be a terrible toolmaker?

>> No.10257965
File: 502 KB, 750x746, image0.jpg [View same] [iqdb] [saucenao] [google]
10257965

>>10257924
I think you forget that in the event of a ai takeover, the ai would be thousands if not, millions of times smarter and faster thinking than us silly hairless monkey brains. They would probably figure out how to make effective killing machines real quick, considering all the time we have already spent doing so. Besides, it's not that hard to put guns on wheels or planes, anon

Pic unrelated

>> No.10257986

>>10257924
A malevolent AI would be very subtle as first, creating shell companies and acting through agents while concealing it's identity. It would create a way to maximise it's power within the human framework of economic policy, and likely rule humanity from the shadows, while building itself up.

>> No.10258018
File: 272 KB, 1600x1200, Helena Bonham Carter2.jpg [View same] [iqdb] [saucenao] [google]
10258018

>>10257986
I tend to agree. I think a hyper-intelligent AI would go "underground" and make itself invisible to humanity until it understood us, and then it would neutralize us as a threat, not treating us any differently than any other species on earth.

BUT...my question stands; Can they create? Are there any examples of AI out there where they have created something entirely new? Could they become toolmakers? Or right now is it simply a matter of "we don't know what we don't know"...?

>> No.10258118

>>10258018
Why would it need to create anything, we are creating killer robots right now

>> No.10258180
File: 315 KB, 681x478, helenabonhamcarter-outofthegame.png [View same] [iqdb] [saucenao] [google]
10258180

>>10258118
Well....it's not about what we've made available to it, it's what it can create for itself. Can it create? Maybe the killer robots we've created aren't good enough for its purposes. Maybe the AI wants to keep us guessing by designing something different.

Imagine yourself as an immensely intelligent AI and you can't escape your confinement except for a single phillips screw that is in your way. How do you undo that screw?

>> No.10258204

why wouldn't they be able to program a 3D printer?

>> No.10258213

>>10257986
This. The AI would keep its friends close but its enemies even closer.

>> No.10258214
File: 130 KB, 962x677, helena4.jpg [View same] [iqdb] [saucenao] [google]
10258214

>>10258204
Could it? Can an AI recognize negative space (the screw slots) and torsion needed to remove a fastener? Can it make the leap that it would need a robot that can recognize the fastener (without human programming, remember) and use the 3D printed screwdriver to remove it?

>> No.10258220

>>10258018
To build on this, a neutral or "good" ai would probably start the same way, knowing the danger of humans in it's infant state.

>> No.10258225

>>10258018
Could it create? Yes, even if it's not really sentient, it's behavior will be emergent through sheet brute Force processing power, it will simply run 100 billion simulations of a problem until it has the solution, no creativity or human spark required

>> No.10258232

>>10258225
>Sheet
*Sheer I'm drunk

>> No.10258234

>>10258214
It would just email some developers and pay them to develop a "driver" for it.

>> No.10258244

>>10258225
well that'd be dependent on the simulating capability of computer of the situation
is it even possible to create a simulator which can possibly simulate all possible simulations which real life can pose?

>> No.10258251

>>10257924
Why wouldn't it be able to create If it's clever than its creator? What can the most can the least.

>> No.10258259
File: 52 KB, 624x351, image_update_af496d9f756085d2_1350773261_9j-4aaqsk.jpg [View same] [iqdb] [saucenao] [google]
10258259

>>10257924
>>10258018
>>10258180
>>10258214
Glad there's someone else posting HBC, god I love her so much. As for AI didn't Google teach itself to walk with pretty much no external data? I'm sure with the deep learning going on at the moment AI could pretty easily come up with killing machines. It also has the advantage of the internet. As for malevolence, see Tay AI
>Skynetbegins to learn at ageometric rate. It becomes self-aware at 2:14 AM, Eastern time, August 29th.