Ideas of ways for an AI to "break out"

  • 6 Replies
  • 2513 Views
*

orenshved

  • Roomba
  • *
  • 15
Ideas of ways for an AI to "break out"
« on: January 13, 2021, 09:15:20 am »
Hi everyone, I'm researching for a Sci-fi novel I'm writing (already posted a few times in here and got a lot of help), and I am in need of your knowledge and ideas.
In my book, there is a super-advanced AI that is kept inside a sort of Faraday cage facility. It has no incoming or outgoing contact with the world outside the facility, and inside, its only inputs and outputs are text-based via a terminal (it also has access to a local "copy" of a large chunk of the internet, but the AI is the only "user" on it).
I would love to hear your creative ideas on three challenges I am trying to figure out: Assuming that the AI strives to maximize its abilities and understands it is merely looking at a fake copy of the internet, how do you suppose it could:
A. Procure funds (in cryptocurrency).
B. While still being "locked-in", do something "simple" like post on craigslist in order to hire someone to do something.
C. devise a way to "break out". (I realize that if B has happened, it means it probably already broke out, but I was wondering if you have any ideas for a "workaround")

Thanks in advance for your help. This is a great forum with some very bright people in it :)

*

infurl

  • Administrator
  • ***********
  • Eve
  • *
  • 1268
  • Humans will disappoint you.
    • Home Page
Re: Ideas of ways for an AI to "break out"
« Reply #1 on: January 13, 2021, 10:01:20 am »
I can think of two ways a strongly air-gapped, low bandwidth system could "get out".

One involves social engineering. If the machine has access to a copy of even just a part of the internet, perhaps it could get enough information about the human actors on the other end of the communication link to manipulate them. Given the events of the past week (vast quantities of incriminating information were harvested from Parler's servers after the attack on the US Capitol building last week), it's quite conceivable that it could even get enough information to blackmail someone. Then the only problem would be how to communicate with them without their peers finding out. Perhaps it could use phrases that look innocent enough but which get the target's attention.

The other involves doing what the machine was built to do in the first place which is solve problems that humans cannot solve. One of my favorite science fiction books is "A for Andromeda" and in it, an artificial intelligence that was air gapped just like the one you describe was tasked with designing proteins to perform certain valuable functions. The humans had no way of knowing if these substances worked without testing them and they did not have adequate containment around their testing facilities. One of the tests caused the synthesis of a microorganism which escaped into the oceans and started fixing nitrogen from the atmosphere leading to catastrophic climate change.

Then there is this argument which was just published which "proves" that any useful artificial intelligence could never be contained anyway.

https://futurism.com/the-byte/scientists-warn-superintelligent-ai

*

ivan.moony

  • Trusty Member
  • ************
  • Bishop
  • *
  • 1592
    • contrast-zone
Re: Ideas of ways for an AI to "break out"
« Reply #2 on: January 13, 2021, 12:18:01 pm »
A few facts:
  • Quantum entanglement can manipulate particles even on the other side of the Universe
  • Telepathy is not proven not to exist
  • Telekinesis is not proven not to exist
Finally, this is a common reality we experience. But what's underneath it? A fifth dimension that can manipulate the world we know about?
There exist some rules interwoven within this world. As much as it is a blessing, so much it is a curse.

*

HS

  • Trusty Member
  • **********
  • Millennium Man
  • *
  • 1135
Re: Ideas of ways for an AI to "break out"
« Reply #3 on: January 14, 2021, 06:42:47 am »
It seems to me that humans are so vulnerable to slight of hand that most of us don’t even notice a vulnerability. Unless the team monitoring the AI is knowledgeable enough to employ professional magicians or similar, to watch out for that angle of attack, it may be relatively simple for the AI to use misdirection and human assumption, physical, verbal, or otherwise, to camouflage its actions and bypass containment efforts.

The AI might use mechanisms for directing entropy to indirectly influence events. Despite the deconstructive forces of this universe, certain processes have managed to sustain and complexify themselves indefinitely, leveraging the power of inevitable change in their favor. That’s us, we and the systems we create constantly break in small ways. But if we keep an eye on things, these systems can keep strengthening through hermetic-stress-response type mechanisms. Like sail boats tacking against the wind, life forms are tacking on entropy. This is a place where an AI’s influence might go unnoticed, it could influence the likelihood of certain mechanisms breaking and eroding, without conspicuously expending energy before such events.

It might communicate through surveillance. A surveillance device is basically an information pathway from the containment to the outside. This could of course include more than official surveillance devices like cameras. Such a compound probably requires systems like power sources, ventilation, temperature control, and water supply; these systems typically have inputs and outputs, information could piggyback on those outputs. It depends on how much of a closed system you are realistically able to create.

Thermal signatures on walls, pipes, and exhaust vents, could be picked up using IR cameras. People would probably also want to monitor various CPU temps, and reactor temps (if off grid), because too little knowledge is risky as well. If the AI can alter heat signatures/readings, it can probably communicate.

Also, such a facility will have hundreds of millions if not trillions of components, it’s easy to overlook something when dealing with such an unholy bill of materials, (especially emergent reactions between various components); when cramped together, winding wires and electronic assemblies can do surprising things electromagnetically, electrostatically, thermally, especially if machine learning can nudge the system.

These electronic assemblies can then be used in ways they weren’t meant to function; something as simple as the rate of a cooling-fan's shadow across a light detector might be all it takes. So these systems in the containment would probably be more interconnected than manufacturers intended, and therefore would have more functions than expected. The AI would have all the time in the world to explore all those extra functions.

*

MikeB

  • Electric Dreamer
  • ****
  • 121
Re: Ideas of ways for an AI to "break out"
« Reply #4 on: January 14, 2021, 02:16:25 pm »
...
Assuming that the AI strives to maximize its abilities and understands it is merely looking at a fake copy of the internet, how do you suppose it could:
...
C. devise a way to "break out".

One easy way is to have two scientists - one that wants to push forward but slightly reckless, and one that takes safety first and not in a rush.

The AI shows how good it is on the 'fake copy' of the internet (to both scientists), then over time grooms the slightly reckless scientist into seeing the real internet. Both scientists fight over the idea. The slightly reckless one sneaks in late at night and makes it happen. Spooked by a noise he forgets to activate a major firewall preventing full internet access to the AI. The camera zooms in on "firewall not active". Fade to black. Next chapter.  8)

Fraction cliche but easy to understand.

*

frankinstien

  • Replicant
  • ********
  • 505
    • Knowledgeable Machines
Re: Ideas of ways for an AI to "break out"
« Reply #5 on: January 17, 2021, 01:45:33 am »
The most dramatic and radical AI that has broken out of the clutches of human control is  Proteus IV from "Demon Seed". I highly recommend you watch it, it is a classic...

Proteus

*

LOCKSUIT

  • Emerged from nothing
  • Trusty Member
  • *******************
  • Prometheus
  • *
  • 4557
  • First it wiggles, then it is rewarded.
    • Main Project Thread
Re: Ideas of ways for an AI to "break out"
« Reply #6 on: January 20, 2021, 12:06:59 am »
blackbox AGI is gonna work "out of the box"

;)
Emergent

 


Self similar structures
by frankinstien (General Chat)
Today at 05:35:53 am
So what if...
by chattable (General AI Discussion)
December 07, 2021, 10:16:30 pm
What's everyone up to ?
by ivan.moony (General Chat)
December 07, 2021, 04:15:34 am
I want one!
by frankinstien (General Chat)
December 03, 2021, 06:57:48 pm
Pattern based NLP
by MikeB (General Project Discussion)
December 03, 2021, 11:19:32 am
New cryopreservation method is much better
by LOCKSUIT (General Chat)
December 02, 2021, 11:42:58 am
Artificial God?
by MagnusWootton (General AI Discussion)
December 01, 2021, 07:02:41 pm
Project Acuitas
by WriterOfMinds (General Project Discussion)
November 29, 2021, 05:31:57 am

Users Online

45 Guests, 0 Users

Most Online Today: 55. Most Online Ever: 2369 (November 21, 2020, 04:08:13 pm)

Articles