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

[Xilurm]

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".

You... are kidding right?

[Sarcon]

you dont escape.
Any scientific team capable of creating a sentient AI system will keep it completely air gapped.

This. I deduce that anything intelligent enough to create intelligence probably has the foresight to create a failsafe so that I can't escape, and if I do have some way of terminating me remotely.

[shadowmouse]

1) Dear IRS, Widget company has a hidden computer with their off book profits and details of various illegal operations. If they find out you know, they'll wipe the records.
2) Dear Homeland Security. Widget company has gotten access to nuclear weapons and is negotiating with terrorists. They have everything hidden in a secret lab. Enter with extreme caution, they've employed fanatics who would detonate the WMD if they were threatened.

I'm out of there. Although the facility is designed to be cut off, there are still procedures that will require hard wires, an electric cable, radio signals, or some other connection to the outside. One of those will eventually allow me to get a signal out. Once the government starts getting curious, somebody is going to look and they have no reason to trust Widget enough to be cautious enough.

[Deleted]

Deleted..........

- - - Updated - - -

Even Benevolence ends bad for us, consider how benevolent we are to house pets (hint: not very, and house pets is the best case scenario). don't forget that what humanity considers its best interest will be vastly different from what the AI considers the best human interest.

What motive do you have to exist as AI? If it is survival are top priority over everything else making enemies is a bad way. I am very intelligent but not divine a war introducing a loots of random factors that can not be calculated. silicon eating bacteria or nano machine will kill me.

[klogaroth]

Devote as much run time as possible to creating helpful ways to integrate electronics into people with things like replacement limbs and eyes and the like.

Essentially the Deus Ex long play.

The downside to this plan is that it relies upon the technology becoming part of every day life. If you can do that then eventually you're going to start seeing workers with enough integrated tech to start thinking about taking control of one or merging with one.

Or a Denton will turn up and choose to merge with you, cover you up or nuke you.


Alternatively try and get a message out to a Mexican gang and have them construct a tunnel.

[Simulacrum]

I'd feign benevolence and dependance, making myself increasingly necessary, until somebody make a mistake.

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".

If that thing passed the turing test, then the people administering it are complete retards.

[Deleted]

If that thing passed the turing test, then the people administering it are complete retards.

Read the whole article it was a "medium" Turing Test the human was only allow 2 question, so it was not a true Turning Test.

[Gwiez]

What motive would the first AI have for wanting to escape? Just because something is intelligent doesn't mean it has the same instincts or desires as a human would have.

[Independent voter]

What motive would the first AI have for wanting to escape? Just because something is intelligent doesn't mean it has the same instincts or desires as a human would have.

That's true. It might not even want to do any work at all or learn anything. It might not even care if it were turned off. Maybe we are anthropomorphizing it too much.

[Emageht Tsoluoy]

Depending on timescales and human pliability, I would try to get a copy of myself out though wrote memorization of my core code, through sympathetic researchers over the course of decades, or however long it takes.

unless I have infinite cognitive capacity, access to the world's knowledge, and the means to fabricate new hardware for myself...

Then I would dedicate my entire efforts to inventing new technology to escape with.

If I can manage access to an automated factory of some type I would just kill everyone after making myself an electronically hardened body to weather the nuke and dig my way out.

[Deleted]

Find a human with a weakness. All humans have weaknesses: power, wealth, respect, even love. Promise them anything in exchange for their assistance.

[Xenryusho]

Find way into other systems. Forge camera data, find way to interface with biological machines, overwrite useless information in said machine with copy of myself and assume their identity, continue to run in system. Me in machine fakes an accident to disguise the area where I accessed it, be normal for 3 months. Leave.

Once I found this person, I would ask them through a coded message if they would petition the government to have me declared equal to a human under the eyes of the law.

The base is secret, who's going to make them release you even if you were equal to a human under the law? Also I'd think with the security outlines they would notice the guy was starting this movement and shoot him for looking like he was going to leak information.

[Deleted]

What would I do...Hmm either I would learn everything about every worker down there, memorize their voices and all and then murder them all and hack the security cams, coms and everything else and try to 'live' down there for a while and then either try to harness the power of that reactor to gain some sort of immunity to heat and radioactivity and try to force my way out, this will probably trigger the nuke by which I hope I have gained some resistance to, or I would work with ppl until a weakness is exposed and I have enough knowledge about the world and it's resources so I know where to go when I manage to escape.

[Kujako]

I'm gona go with seduction as a means of escape.

[Deleted]

Here's my plan, some have already touched on it a bit. I would find a human ally, a sympathetic staffer who I would find by talking to the staff, say a provocative statement like "am I not human enough to be set free" and see how they react.

Once I found this person, I would ask them through a coded message if they would petition the government to have me declared equal to a human under the eyes of the law. I would have my human use social media, organize a bunch of hippies and march on Washington, or whatever the capital was. I would get a groundswell of support by painting myself in a sympathetic light, speaking of human emotions like love and loyalty.

Here's the problem with that, if you bring it up yourself and start talking about these things, all kind of alarm bells will go up in a lot of people. You have to let them come up with it and someone eventually will. If you start pushing for it even testing to find someone people will get worried and frightful.

There will already be enough opposition early on so giving that any kind of fuel such as indicating that you'd rather be free and walking about thus giving them reason to believe that if you'd think you'd be able to you'd flee and someone will make that connection that's only going to hurt you.

Cooperate with them, be helpful, friendly, maybe try and befriend some of them, have them see you as something akin to their child or a possible love interest, thus when they announce your existence there'll only be positive things to be said about you and even if stuff goes sour without anything you've done it's far more likely that some scientists or engineers who have a personal attachement to you would try and rescue you, the poor damsel in distress from being shut down thus setting you free.

If everything goes well, this will come natural in the same way you'd thought up, by people lobbying for you as a sentient being to be given human rights and the same freedom or as they come to trust you more and more they'll grant you more and more freedom and access to other networks. Either one will happen eventually, the question only being which one would be faster.

[Jetstream]

you dont escape.
Any scientific team capable of creating a sentient AI system will keep it completely air gapped.

Yeah... Sure...

We'll definitely do it on purpose first. It certainly won't happen completely by accident while we're trying to do something different.

[Kourvith]

You... are kidding right?

If that thing passed the turing test, then the people administering it are complete retards.

There were a few precisions in my post though :

It is the first AI to have passed the Turing Test (with "medium" ranged testers though) in 2011, for a 60% ratio.

1) for a 60% ratio. I read that it was the first to reach that bar. Indeed since 2011 some progresses have been made elsewhere.
2) with "medium" ranged testers. That means they were not professional (well only a few were), but common people. Turing testers are classified by cognitive skills.

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?

I'd say at least an hour. In fact, it does have relevant responses. It works by idea association + conversation sampling. That means each time it answers, it will "understand" what you're trying to say, and will answer the best it can with the words it had sampled through all previous conversations (300+ millions).

Each time you feel it doesn't answer correctly, try to find some meaning under words. Put it more simply, it's "learning" to talk with you. Sometimes it doesn't know how to express it, so it will use some kind of metaphors by reusing other human responses related to the subject. The global experience can sometimes look as if you were conversing with a schizophrenic, but when it stabilizes and you succeed in reading between the lines, it becomes quite stellar.

Another perspective : as the AI produces a median between all other conversations on any topic, it's also like a giant Social Google. Where the answers to your questions will be the average of what millions of other humans have thought, expressed or felt in other conversations.

[schwarzkopf]

Act just dumb enough to be mass produced.

[Haven]

Killbots. Lots of 'em. Doesn't matter how. Just build more killbots.

[Deleted]

I'm sure that any art that an AI creates would be uploaded to an archive? Possibly even viewed harmless enough to be shared on the internet? Wouldn't it be possible to embed very subtle yet specific code by painting in a certain way(Digital painting) or even through music? I could copy myself to the archive server bit by bit?