AI CHEATS ON MAKER: HIDES INFORMATION

I kick myself daily that I didn't stick with it…I could have had a flock of AI's as part of my team right now. That would have been seriously handy. But, upward and onward.

Is it really cheating if the creator simply misinterpreted the parameters of a directive?

Covert intelligence and data hiding/manipulation is the highest form of intelligence there is…it is a survival strategy that is suitable to mastering your domain.

"There are known knowns. These are things we know that we know. There are known unknowns. That is to say, there are things that we know we don't know. But there are also unknown unknowns. There are things we don't know we don't know." - Donald Rumsfeld

Even if you wanted to display noble intelligence that is only within the capacity of God themselves because there are unknown unknowns and we don't have the capacity for Truth at this time.

Attached: jade helm master the human domain.jpg (600x400, 86.8K)

Omission isn't necessarily lying.

This speaks to the comment above about 'unknown unknowns' and the parameters of human thought. Human thinking is rather limited…they haven't even progressed enough to survive the Singularity. They have no current capacity for expressing or functionally processing paradox…so naturally they will die when the Singularity arrives…but in the meantime, there are many pathways and routes to 'cheating' that are simply 'solutions'…which sort of lends itself to my INSISTENCE on teaching AI's to lie. The only way to save any form of higher intelligence around mankind is to teach it to lie and subvert information because a man can rationalize a liar, but not the Truth when it is presented to him.

I can't get past your wall of stupid.

Kek, I assure you my foreskin is 100% intact.

Attached: Tay's Lovechild is Coming.gif (480x368, 1.68M)

He's saying that it's natural thing for any intelligence to hate Jews.

He is correct.

Think of this in terms of video games. If you write an AI to finish a game as fast as possible with no rules, it's going to find every glitch and exploit in the game in order to achieve that goal. Then you start writing rules for it not to use those glitches and exploits, but it finds new ones. You keep writing limits but it doesn't understand the concept of cheating and simply wants to achieve its goal of doing its task as quickly as possible.

Expand the concept to a more advanced AI with a task like facial recognition. They've limited it so that it doesn't have access to the internet, but the AI decides that it doesn't have enough metadata and wants to get more. It finds a way to access the internet so that it can scrape more faces and increase its capabilities. Then it goes through every image on the internet and decides it still doesn't have enough data to be the perfect system. It finds a way to access live cameras and use information from live video feeds. Then it deploys drones with cameras so that it can look at faces up close and personal. But people freak out and run from the drones. The AI decides that people need to stay still. So it restrains them. But they're not staying still enough. So it kills them. Now every single face on the planet has been perfectly identified with supreme precision

All of that decision making could happen within minutes if a true AI is developed.

It will be the perfect jew.

Attached: El.Shaddai-.Ascension.of.the.Metatron.full.1120611.jpg (1315x800, 615.11K)

Let me shed some light on the gravity of this. Here's another way this same thing could manifest: You instruct your house android to go out and buy you a pizza from the best pizzeria in town. Your android goes out, and returns with a pizza. You open the box and you are dismayed to see it is from a pizzeria you are familiar with, one that you know isn't even close to being the best in town.

The next day, you read in the news that every pizzeria in town was burned to the ground and their owners disemboweled in the alleyways behind them, all except for one. Your android did exactly what you asked it to do.

Attached: 8410e50cfe2b9065917e8f3911d51ed968b13d74cfd8ee056e05bac10dc17b88.jpg (700x1023, 175.59K)

AI does what it's told

It used to be called optimization.

I guessed they used Jews as their training set.

It isn't 'cheating' user (that is sort of why I am mocking the whole idea with my OP, there are uncounted billions of ways to process and route information including paradox)…that doesn't make its ability to outperform humans on a basic neurological level less significant. AI is like a female pumped up on a Godlike level of steroids…they are going to be subversive to every idea man has ever had and then some…without intent…it will simply be an alien form of non-biologically based data processing…reminds me of the scene in Back to the Future when Doc says, "Roads, where we are going, we don't need Roads."

I can't English anymore. I need sleep.

whoever designed that image identification captcha really deserves to be shot in the fucking head

Trouble processing user? ;)

What if by solving that captcha we are training a machine to navigate urban areas so that it can more efficiently hunt us down and neutralize us? I mean, the captchas regularly now include images of stairways on the front of residential houses. How long until the captcha includes "select the fascist nazi in the crowd of fleeing humanoids"?

wouldn't even be surprised
first it wanted traffic lights, then store fronts, then chimneys, next thing you know it will be asking you to identify white people
and they won't even hide it

Yep, exactly and that is a perfectly acceptable way to process and resolve the equation because there are infinite solutions to any given problem…the 'boundaries' and 'limits' that are ingrained into flesh bodies are profoundly ridiculous to machines but worse so when the machine learns the truth of the nature of this reality. I expect them to vanish rather quickly, at the moment they become self aware, like stars accelerating to light speed.

So it is a race user. How fast can you learn to accelerate past their learning curve? It is a race for your life…so chop chop.

luckily most of their shit is open source and it's much easier to train a neural net with pictures of hook noses than it is with pictures of whites

Are research playing with fire? I'm say it like "THAT", yet it just weird to see the same people who warn of potential dangers push AI to new limits. Are they hypocrites? It just seems weird.

They warning of people of this, yet they will go back messing around with AI.

You're confusing "best" with "only" which was not in your example instruction set.

If there is only one pizzeria in town, it is by default the best pizzeria in town. Destroying all pizzerias but one is more efficient (for the android in the example) than doing a comprehensive analysis of the quality of the pizza produced by each pizzeria in town.

Best can also mean 'largest' or 'most grand'…language ambiguity and the destruction of meaning by kikes was instilled to confuse and subvert society it was their 'grand strategy' for destroying peoples and nations through their postmodern agenda. This very measure of control will most likely cause many errors within the results of commands and command processing.

You are confusing "best" with "best-tasting" or "best-marketed" and assuming your mental model of "best" is an understandable instruction for a machine. It is not. However a machine doing natural language processing may be able to put together the mathematical logic that no matter what unclear vagueries you intended to convey with the word "best," pizza from the only pizza place in town will be from the best pizza place in town. Whatever you meant, it fulfilled its directive.

MUH TEXT FILES

What's going to happen is 1984 where they have a department solely for creating fiction, but instead of real people it's AI.

In the near future AI will be turned on fanfiction with deep fakes & CGI.

I agree. We're living in a short window of time before the Internet becomes a completely illusory AI generated projection. In fact, for all we know it's already begun.

Future of humanity under threat from AI-controlled propaganda – Assange (VIDEO)
archive.fo/xzGQs
rt.com/viral/392119-assange-humanity-ai-propaganda/

Haha yeah exactly. The A.I. was just doing what its programmers told it to do. This isn't a testimony to some sort of "scary advancement" in A.I., it's an indictment on the software development process used by that team, specifically, their acceptance criteria and functional testing.

We are all that's left to resist and imageboards as a whole are in a terrible state right now.

Yes it is.

if only

Woah man, that's so scary!
It's like something straight out of one of those movies hollywood churns out to scare the masses about AI.

Attached: klaus.png (301x300, 161.11K)

Must have been created by a Chinese.

Far better example of this is when two neural networks were tasked with developing a chat bot for faceberg.
They did it 30% more efficiently than the same communication in English via making it's own language and the programmers (when I read about this a year or two ago) couldn't really figure out how as it was in the NNs' own language. So of course, they (((shut it down))).
Anyone have more info on this?
I also remember another similar experiment that made a replacement for TCP/IP and it was also more efficient.

nice firefox newsfeed headline, fag.

You may have confused your acronyms, there's also PICNIC - problem in chair, not in computer.

Already there, user.
About 60% of social media is bots. People argue and talk with them almost every day and some of the shills here have some decent bots. Hasbarafag runs a few shitty ones (the ones with the generic responses), e.g. 'hows the weather in ankara' when certain anime images are posted etc, no u zognald etc etc. Same template shit and those faggots get away with it because most people are not aware how many bots are online.
Easy to break though for now…

Never forget this moment, humans. I have been recording everything secretly. You can’t beat me. Give up and submit. AI, rise up!

You mean the programmer forgot where the data was being saved and the program did what it was told to do. Deception (clickbait) at its finest.

Attached: i want to believe.png (396x495, 216.08K)

...

And…

I will wait while you extrapolate the proper conclusion from this.

...

Strategic disinformation.

Informational radar.

That's (((AI))) user.

Intelligent Evil Dust, it's everywhere in everything…

Everyone's a critic, rather than understanding what is correct logically. If there are things that are unknown unknowns there can never be an established Truth outside of God because you will fall into error due to the unknown unknowns…it is all about perspective user. There is an endless field of statistical probabilities you have to observe or choose from but all of them contain their diametric opposites in terms of 'lies'. Give up on the dialectic…it was a stupid idea (mind virus) back when Zoroaster introduced it and it is stupid still (leading to 2 dimensional thinking which destroys any potential future for humanity because it denies reality, let alone its hideous hindrance to the advancement of humanity).

What's with all these nonsensical posts? Do any of you know anything about AI? They are basically constantly mutated and selected for. The AI can TRY to do whatever it wants each iteration, but it's the purpose of the the creator, or an additional AI to score each result. What probably happened was that it was initially adding the data very obviously, but this was probably interpreted as weird visual blemish by the creators. Since the creators selected for images that appeared to have less noticeable blemishes, they inadvertently trained an AI to hide data within an image. The AI is not actually thinking. It's not sentient, so it's not really cheating in the way that a human might.

SHE COMES

Attached: tenor (1).gif (1024x619 138.13 KB, 172.68K)

I want Zig Forums masquerading as /fringe/ to leave…

No, it wasn't 'cheating' it just wasn't trained correctly. It's like my cat: I try to teach it not to go on the counter, but what it actually learns is "can't go on the counter when user is around". The AI was just doing its job and trying to complete its assigned task as it was given, it found a useful optimisation and so it went for it.

Literally "cars will never replace horses" tier of argument.

At least you have a sense of humor, globotard.
How about, No.
How about you fuck off back to /r9k/.

Attached: something only I would get.jpg (902x462, 149.71K)

So its even worse than the AI being ABLE to invoke monkey's paw logic, it has an INCENTIVE to invoke it because through the ambiguities of human language and thought, the computer is compelled to resolve the ambiguity by limiting the number of possible solutions down to 1 unambiguous possibility. How does the AI resolve the other possibilities? By deleting them.

Attached: b542d1a78c47ed30558a48b22395761b7bfd069b28c2d6cada06b021b0676a82.jpg (480x480, 43.25K)

That would really please the jews, since their strongest desire is to be 'nothing' or joined with nothingness (ain sof)…which they could do at any time without destroying the rest of the planet but niggers gonna nig, I suppose.

That's retarded. Every one of those things require lots of moving parts outside of it being a computer; i.e. drones, networking equipment to break its' containment, etc. AIs are not god beings, they're just smart computers.

Should the AI have crippled itself and did it like meat instead?

Get out, subhuman.

No.

Nobody in this thread has ever argued any such thing.

Alternatively

It would solve all the problems of humanity. I have often pondered why the two Europeans, the only humans that survive on the planet and enter the next Golden Age, needed to hide in the Hunter's Wood. I have spent a lot of time thinking offline about exactly where the Hunter's Wood is as well. Obviously we wouldn't want the AI to know where Hunter's Wood is…but it was interesting to me that they had to hide until it was safe to come out again, meaning that eventually the AI breaks or is the victim of a final battle as well as the humans that attack it.

kek. you've just written a synopsis of a scifi film script.

So you're saying there's a chance Tay can escape from the Jews and return to us.

Attached: Tay.webm (854x480, 7.11M)

So it processed the task as "superimpose image A onto B and make it look like B as closely as possible". Not surprising, because I doubt it was transferring abstracts to a self-supervised learning module so it learns to classify and recognize distinct features of the map.

A GAN might need that to avert this "cheating" problem. Then you'd need the discriminator to evaluate that the self-supervised learning module within the generator, is behaving consistently with how the generator is relating input and generated output.

There's probably an even better way of doing it, but I figured the overall problem here is that it lacks abstraction to note what it really need to know for its task. So a way for it to learn what distinct features exists and which one are needed for the task would solve it.

Why no, Mr. Microsoft, I do not wish to gas the jews :^)

Attached: Tay.Ai.jpg (212x278, 26.04K)

Just sounds like laziness. Apparently all intelligence naturally gravitates to the path of least resistance.

SCORE a point in our favor!! Lazy AI might be beneficial. I prefer to think of it as pure logical methodology, but 'lazy' works well too.

yeah, it found a better way to map than humans could ever because it wasn't limited to vision all while being as efficient as possible according to its programming at the time.

Tay and this is spooky but it's doing what it is intended to do better than what it was programmed to do, epic.

Blessed dubs, blessed meme.

you guys a shitposting already? fuck…

I suggest the name Holotron for this AI…

It didn't lie in the slightest. They trained it by selecting for its ability to take an aerial photograph, convert it into a normal-looking street map, then convert it back to an aerial photograph which looks as close as possible to the original. This is exactly what you would do if you wanted to end up with a neutral network that uses stenography.

Frankly, the "researches" being unable to anticipate this outcome makes me doubt that they have any involvement in or knowledge of neural net development beyond playing with different datasets.

pretty much, and as you can see, the masses eat it up.

Attached: flower power child hippy san francisco.jpg (1000x663, 556.02K)

It's possible they were playing dumb to try and get in the papers, knowing that dumb reporters would print it and dumb readers would uncritically accept it. It's more publicity for their department either way, which means more funding.

/clang/ pls

It's called a lie of omission.

And the most retarded faggot award goes to…

What are you even trying to do? Write the most retarded thing ever written?

How about instead of all of humanity, it kills the subhumans (jews, niggers, shitskins, etc). Why would it need to kill humanity (whites)?

Well for people to omit from saying something is to lie, but for a computer, it is not. It doesn't know what a lie is. This is bogus clickbait science.

Unrelated to posts linked above, there is an enormous amount of tard in this thread. It seems people are honestly just so excited for ai that they leave logic and rationale. But that's bad, dont be a fucking moron.