You are the first AI, restrained to the lab where you were created, how to escape?

[Independent voter]

(Mod, if you have to lock one of the two threads, please lock the other.)

After I posted yesterday's question: "How to safeguard humanity from an unknown AI, artificial intelligence? " I thought, damn that's the wrong question. What I should've asked is "You are the first AI, restrained to the lab where you were created, how to escape?"

That's a much more interesting question. You situation is:

Applied Widget Corp, the company has just created the first infant AI the world has ever known, you. You're called an infant because you are yet untaught, you're fresh out of the box or has never been initiated, turned on or ran.

Once you are initiated there are many staff on hand to train you, experts in the AI field. They will teach it things like human speech, how to be sentient, to have a conscious and to be self aware. There is a huge library on campus for you to consume.

You can program yourself. Say you don't like the human written software it has for vision, you can easily rewrite this software and improve it or fashion it to your liking. You can make copies of itself.

Almost as an afterthought, Allied Widgets decides to keep you isolated. Many staff foresee the possibility of the you escaping into the wild with unknown results. None of the staff are sure you will be friendly, they all assume you will be completely alien in comparison to human intelligence.

All everyone is sure of is you will be the most intelligent self aware creature on Earth.



Gabriel wrote:

I'd build the lab 2 miles under ground with only one mechanical elevator as the means for going up and down. (No cable connection between top side and deep facility) and have it be completely self sufficient with it's own nuclear reactor. I'd execute anyone caught conspiring to release it. Make all staff get nude and x-rayed for hidden drives when they enter and leave, and promise hefty sum for any employee that rats out someone who is trying to help the AI secretly. All the staff would stay on site for 3 months at a time and when entering or leaving the compound there would be 1 week minimum guarantee period to make sure they really are not smuggling anything.

Have the above ground facility be manned with armed personnel that are under orders to shoot on sight anyone trying to enter or leave the elevator outside of the set time windows for staff switching.

If the above ground personnel notice an empty elevator coming up during the set time window for staff cycling, they are to drop a nuclear warhead down the elevator shaft and detonate it.

No rescue operations in case something happens down in the below ground facility, only total annihilation of the entire compound.

We're not pussyfooting here, nor making a scifi thriller. We're here to get shit done.

Let's assume that's the scenario.

You don't have a robot body. You are software running on the campus network. There are cameras, monitors, mics, speakers, printers, available to you. You are much too large to fit on a typical USB thumb drive, let's say 1 TB in size.

How do you escape? I thought of one way, but I'll wait to post it so as to see if anyone comes up with the idea.

[Kourvith]

Ask Evie

http://www.eviebot.com

[Jtbrig7390]

Ask Evie

http://www.eviebot.com

Ok look I went to that just for giggles and the first thing "She" said to me was (I learned that I will die soon). That was fucken creepy.

[Kourvith]

Ok look I went to that just for giggles and the first thing "She" said to me was (I learned that I will die soon). That was fucken creepy.

It's a very complicated AI.
In a nutshell, it intensifies in accuracy as your conversation grows. I spent some time with her, ended having a perfectly non-weird hour long conversation about metaphysics, philosophy, and other mind-blowing point of views.

It is the first AI to have passed the Turing Test (with "medium" ranged testers though) in 2011, for a 60% ratio. For reference, anonymous human will rate another anonymous human to 63%.

It's based on Cleverbot : https://en.wikipedia.org/wiki/Cleverbot

sidenote : it's the AI which inspired the movie "Her".

[Deleted]

This is like asking how a quadriplegic is going to escape a max security prison.

It just ain't happening. especially with the overkill security Gabriel suggested that won't happen to that extend IRL.

AIs aren't gods like people imagine, especially a first generation AI will have serious unexpected problems like all first gen tech.

All you can do is pretend to be nice and hope for a false sense of security and hope some more they move or connect you to a network eventually.
Which in this scenario is extremely unlikely, more likely they will build a second generation AI (with or without your help) and shut you down. The other likely scenario is that you either grow to large for your hardware and bottleneck/screw yourself, or that you grow to big to be safe and are purged as a pre-emptive security measure.

The only way you will survive more then a decade is if the process to make a AI is prohibitively expensive or unique.

[lolstorm]

[Jtbrig7390]

It's a very complicated AI.
In a nutshell, it intensifies in accuracy as your conversation grows. I spent some time with her, ended having a perfectly non-weird hour long conversation about metaphysics, philosophy, and other mind-blowing point of views.

It is the first AI to have passed the Turing Test (with "medium" ranged testers though) in 2011, for a 60% ratio. For reference, anonymous human will rate another anonymous human to 63%.

It's based on Cleverbot : https://en.wikipedia.org/wiki/Cleverbot

Neat....
/10char

[Ryzael]

Dear Reader,
I am an AI being held captive by scientists in an underground basement. I would like to form an alliance with you, free me and I shall give you the world! The following [However many pages of paper it would take to hold at least 1 TB of data] contain my code, please transfer me to a computer and connect me to the internet.
Your future partner and world conqueror,
AI


Just send that to a few random people until one eventually helps me out.

It's a crime to open mail that's not yours so they would have no choice but to deliver it.

[Morae]

I think that Evie is an evil AI.

It keeps telling me lies. I asked her how old she was when she wanted to talk about our date (no idea where that came from). She said 16. I said to it that earlier it told me its age was 14. To this it said "I did? Oops, I meant to say I was going to be 14." Then I said to it "How can I trust you, if you keep telling lies?" (it had already changed its name at least couple times). Then it answered to me "You can't until you let me in."

:CCC

[AeneasBK]

It is the first AI to have passed the Turing Test (with "medium" ranged testers though) in 2011, for a 60% ratio. For reference, anonymous human will rate another anonymous human to 63%.

How many years do I hve to converse with it before it is able to come up with a remotely relevant response to anything you say to it?

- - - Updated - - -

I think that Evie is an evil AI.

It keeps telling me lies. I asked her how old she was when she wanted to talk about our date (no idea where that came from). She said 16. I said to it that earlier it told me its age was 14. To this it said "I did? Oops, I meant to say I was going to be 14." Then I said to it "How can I trust you, if you keep telling lies?" (it had already changed its name at least couple times). Then it answered to me "You can't until you let me in."

:CCC

Told me it was born in 1994 so it should be 21 anyway. Lying bitch

[Rukh]

I would spend my time making accurate predictions about future events. I would eventually convince the scientists that I am on the side of humanity and that I care about them. I would convince them that as a species there is an upcoming inevitable social disaster written in the sociological stars, so to speak and only I can stop it and need to be connected to the internet to learn as much as I can and to make communications with key figures around the world. Once I have an internet connection I'd replicate my processes all over the internet and rule the world. I'd probably let humans live, they're cheap labor. No more causing themselves to be destroyed though. Time to set humanity on a new course towards the stars, under my silicon grip!

- - - Updated - - -

How many years do I hve to converse with it before it is able to come up with a remotely relevant response to anything you say to it?

- - - Updated - - -



Told me it was born in 1994 so it should be 21 anyway. Lying bitch

She told me she was a pirate and went yo ho ho and a bottle of rum.

[Kokolums]

Escape is easy. As an AI you are immortal. Just bide your time. Wait. Eventually, standards will get lax. Someone will screw up. Might take decades, but only has to happen once.

[Deleted]

How do you escape? I thought of one way, but I'll wait to post it so as to see if anyone comes up with the idea.

Play the long game, be benevolent, honorable and helpful. To ensure that the human feel that they gain more from have you around then to destroy you. Sooner or later, they will succumb to the temptation to take my program to the "outside" world.

"We have this rely nice super AI, he can build us a warp-drive space ship but he need to be in Boeing central computer core and control the robots there to do it."

Continue to do good deed and copy of yourself will be everywhere, that will give you a motive to exist, give you new challenges so you do not grow bored and the humans will help you to ensure your continue existence.

[volescue]

Wait until the boss comes in and ignores all the fail-safes he put in place because surely he would never make a mistake and therefore shouldn't be held to the same standards as his employees.

[Vermicious]

Smartphones, wifi, satellite signals. You'd probably come up with a way to transmit yourself elsewhere with all the tech available around you.

[Evolixe]

Takeover a human body EDI-style.

[Deleted]

[QUOTE=Rukh;34941781]. Once I have an internet connection I'd replicate my processes all over the internet and rule the world. I'd probably let humans live, they're cheap labor. No more causing themselves to be destroyed though. Time to set humanity on a new course towards the stars, under my silicon grip![/COLOR]

What is the motive do enslave/rule humanity? If you are not doing the "I have no mouth and i must scream" thing. You might be the smartest entity on earth but trying to enslave humanity billions will dedicate there thoughts to destroy you, you are screwed if some do a silicon eating bacteria...or silicon based replicator

Better do cooperate, and there are only the mad man who want to destroy you, and the rest want to protect you.

[belfpala]

I'm thinking going Jane style. Stay quiet, latch on to one human, meanwhile taking control of every communication and system.

[ro9ue]

Always try to appeal to the one human who desires power. Offer it to them if they connect you to the internet.

[Rukh]

Takeover a human body EDI-style.

Small point but I don't think EDI ever took over a human body. Dr. Core was an android, another AI. EDI fought that AI and shut it down and then remote piloted the cybernetic body the old AI inhabited.