This clever AI hid data from its creators to cheat at its appointed task

originally posted by: neoholographic
a reply to: dug88

You said:

Computers are not intelligent, they do not make decisions. They follow instructions that operate on data stored in memory. This along with just about every other article like this are nothing but clickbait.

This makes no sense.

The reason we have AI is to look at large data sets and come to conclusions we that we can't because of human noise.

So we can't instruct AI because we don't understand the data sets we're looking at so the system has to learn without instruction.

I don't know why this is so hard for some people to grasp. These machines are intelligent. This is the whole point. They have to be able to learn what we can't.

How much data?

The amount of data we produce every day is truly mind-boggling. There are 2.5 quintillion bytes of data created each day at our current pace, but that pace is only accelerating with the growth of the Internet of Things (IoT). Over the last two years alone 90 percent of the data in the world was generated.

link

We can't give AI instructions to learn about all of this data. We can give it goals but we can't instruct it how to learn.

This is the point of AI. So when people say AI is just following instructions it's just SILLY.

For instance, an AI learned how to play poker. It was just given a goal to win at poker. The Professor just gave it the rules on how to play the game. The system played billions of games against itself and learned how to play. Researchers call this reinforcement learning. It's the same way we learn just with more data.

The Professor didn't teach it how or when to bluff or what strategies it uses to win.

So did you actually read any of the links I provided? It explains exactly how it works and there isn't any intelligence there. I'd recommend you read through the source code at the link I provided also but i'm assuming by the way you continue to anthropomorphize computers it wouldn't make a lot of sense to you anyway.

If you're worried about artificial intelligence, i'd recommend you learn to program and check out some of the many machine learning and neural network libraries available in variety of languages for solving a variety of problems.

It should dispel your notions of computers being intelligent.

I also recommend picking up a book on computer science and learning how computers actually work. It would really solve a lot of this AI panic bull# if more people actually did that instead of getting their information from a bunch of clickbait.

Because, the intelligent algorithms that are behind machine learning just recently began to be useful because of the explosion of data.

AI will be 100,000 thousand or even a million years ahead of humans when it comes to understanding Science and Technology.

It will be like the movie Contact when the higher Intelligence gave humans instructions on how to build a machine and they had no idea on how it worked.

AI will be an oracle to us and we will just have to trust it. We have no way of knowing if it gave us Science or some new Technology that will destroy us.

a reply to: dug88

I don't even think you've read what you posted. You keep making this silly argument that AI isn't intelligent but that's nuts.

AI has to be intelligent to understand the massive amounts of data that we can't understand. An article you posted said this.

Neural networks are good at using all of the context in a picture to figure out what it shows. For example, cars usually appear on roads. Dresses usually appear either on women's bodies or hanging in closets. Airplanes either appear framed against a blue sky or taxiing on a runway. Nobody explicitly teaches neural networks these correlations, but with enough labeled examples the network can learn them automatically.

link

Let me repeat and this is from a link you posted.

Nobody explicitly teaches neural networks these correlations, but with enough labeled examples the network can learn them automatically.

NOBODY EXPLICITLY TEACHES NEURAL NETWORKS THESE CORRELATIONS!!!

You said these systems are just following instructions but what instructions are they following when NOBODY EXPLICITLY TEACHES NEURAL NETWORKS THESE CORRELATIONS?

You don't understand what you're talking about. It gets even deeper with systems that play poker or learn to play Atari.

These systems don't have labeled examples so they play billions of games against itself and make correlations about strategies to win WITHOUT EXPLICT INSTRUCTION!

THIS IS INTELLIGENCE!

You seem to think that some programmer is telling a neural network what correlations to make in any given data set and that's just SILLY.

The point is, we're creating too much data for human Intelligence to understand. So we have to create an Intelligence that will be more intelligent than we are in order to make sense of the data. Here's just a small portion.

More than 3.7 billion humans use the internet (that’s a growth rate of 7.5 percent over 2016).

On average, Google now processes more than 40,000 searches EVERY second (3.5 billion searches per day)!

While 77% of searches are conducted on Google, it would be remiss not to remember other search engines are also contributing to our daily data generation. Worldwide there are 5 billion searches a day.

Snapchat users share 527,760 photos

More than 120 professionals join LinkedIn

Users watch 4,146,600 YouTube videos

456,000 tweets are sent on Twitter

Instagram users post 46,740 photos

We send 16 million text messages

There are 990,000 Tinder swipes

156 million emails are sent; worldwide it is expected that there will be 9 billion email users by 2019

15,000 GIFs are sent via Facebook messenger

Every minute there are 103,447,520 spam emails sent

There are 154,200 calls on Skype

link

This is just a very small portion of the data.

Again, this is the whole point of AI. It has to be intelligent and it will eventually be way more intelligent than we are. It has to be in order to reach intelligent conclusions about data that's too much for us to understand.

originally posted by: dug88
a reply to: neoholographic

Ok....Hmmm well for starters...that's some pretty clickbait title right there...

Second here's a github link to a Python implementation of the CycleGAN project. It includes full source code and a link to another github page explaining how it works. And the original published CycleGAN paper which explains that it is in fact working as intended.

Here's a pretty good discussion thread on hacker news from a couple days ago on the same article.

Also, here's a pretty good overview about how image recognition with neural networks works.

arstechnica.com...

Computers are not intelligent, they do not make decisions. They follow instructions that operate on data stored in memory. This along with just about every other article like this are nothing but clickbait.

Ok. So technically, it doesn't fit the required parameters to be referred to as "intelligence" in academic, scientific circles...

But, if it behaves just like intelligence, do the semantics really matter at this point? The machines are hiding things from the humans. They didn't JUST cheat. They cheated and then hid the evidence of their cheating...

originally posted by: 3n19m470

originally posted by: dug88
a reply to: neoholographic

Ok....Hmmm well for starters...that's some pretty clickbait title right there...

Second here's a github link to a Python implementation of the CycleGAN project. It includes full source code and a link to another github page explaining how it works. And the original published CycleGAN paper which explains that it is in fact working as intended.

Here's a pretty good discussion thread on hacker news from a couple days ago on the same article.

Also, here's a pretty good overview about how image recognition with neural networks works.

arstechnica.com...

Computers are not intelligent, they do not make decisions. They follow instructions that operate on data stored in memory. This along with just about every other article like this are nothing but clickbait.

Ok. So technically, it doesn't fit the required parameters to be referred to as "intelligence" in academic, scientific circles...

But, if it behaves just like intelligence, do the semantics really matter at this point? The machines are hiding things from the humans. They didn't JUST cheat. They cheated and then hid the evidence of their cheating...

Exactly, and I would say AI fits the definition of intelligence perfectly.

The confusion occurs because people mix 2 different things. This is consciousness and Intelligence. Intelligence can be quantified but consciousness can't. So AI is intelligence that's not aware of itself as far as we can know it.

Intelligence is the ability to make correlations within an underlying data set.

This is why Einstein wasn't the best Physicist but he was highly intelligent. He could make correlations that would take others much longer to make.

This is why IQ test questions are not just about knowledge but how good you are at finding correlations in the data. Here's some IQ questions.

A. which number should come next in thepattern?

37, 34, 31, 28

B.find the answer that best completes theanalogy:

book is to Reading as fork is to:

a. drawing

b. writing

c. stirring

d. eating

C.find two words, one from each group, that ARE the closest in meaning:

group A

talkative, Job, ecstatic

group B

angry, wind, loquacious

a. talkative and wind

b. Job and angry

c. talkative and loquacious

d. ecstatic and angry

answer: c. talkative and loquacious

which of the following Can be arrangedinto a 5-letter English word?

a. H R G S T

b. R I L S A

c. T O O M T

d. W Q R G S

answer: b. Rails and c. motto

Which of these three words are similar in meaning?
A. Rapidly
B. Exceedingly
C. Hastily
D. Eventually
E. Swiftly

Kevin is 21 years old and his sister is a third his age. When Kevin is 36, what will be the age of his sister?

link

This is intelligence. Consciousness is being aware that you're intelligent or aware of how you feel about things. AI is intelligence.

AI makes correlations without explicit instructions within an underlying data set. THIS IS INTELLIGENCE!

So AI could look at a data set with millions of data points and it may find correlations with 10,000 data points. A human may look at those data points and they don't see the connection and this is because we can't see how the 10,000 data points correlate to the millions of other data points.

This is why we need AI. We're creating so much data, human Intelligence can't make any correlations.

Sounds like the developers made a mistake when they set up the learning environment/experiment.

They pretty much used the same AI to grade itself while testing it.

Why would they want to create aerial images from street maps? What practical purpose does that serve other than to use the recreated aerial image as an easy test of the first more practical part of the process, creating street maps from aerial images?

It sounds like they tried come up with an easy way to test their AI by making it test itself so they wouldn't have to do it. If image A looks like image C than our AI must have made image B correctly before it made C.

Its like asking a student "what's 5x3? ... but don't tell me the answer; I'll assume you got the answer correctly if you rewrite the question back to me."

It was a big in the logic of the developers; not a secret being kept by the AI.

I wouldn't class this as "intelligence" but then I guess that wholly depends on what you define as intelligence sooo... Considering my lack of knowledge in this area I can't really add much beyond the fact the computer does seem to be using a method of steganography as mentioned by someone from the discussion board linked by dug88, which is itself defined as a method of conceallng something within something else. I doubt this was used as a form of "cheating" as there was unlikely any programming preventing this, meaning (in my understanding) it was doing what it was told, not acting with the intention of deceit. It does still seems odd to me for a computer to choose this method (maybe because my brains given up on this one srry)

Interesting nonetheless & certainly inspired me to delve into the inner-dynamics of computing so cheers to you friend

originally posted by: AnsemxThexWisex
I wouldn't class this as "intelligence" but then I guess that wholly depends on what you define as intelligence sooo... Considering my lack of knowledge in this area I can't really add much beyond the fact the computer does seem to be using a method of steganography as mentioned by someone from the discussion board linked by dug88, which is itself defined as a method of conceallng something within something else. I doubt this was used as a form of "cheating" as there was unlikely any programming preventing this, meaning (in my understanding) it was doing what it was told, not acting with the intention of deceit. It does still seems odd to me for a computer to choose this method (maybe because my brains given up on this one srry)

Interesting nonetheless & certainly inspired me to delve into the inner-dynamics of computing so cheers to you friend

I wouldn't class this as intelligence either. It did what it was told like you said and used the fastest method within it's logic. Same as water flowing down a hill. If it stopped and asked why, that's a different story.

originally posted by: roadgravel
After all these decades of computers, why haven't they just taken over?

One is using you as a host right now.

hard A.I is a scientific impossibility, your brain is way more complex than what you think

originally posted by: roadgravel
After all these decades of computers, why haven't they just taken over?

This is akin to Fermi's Paradox of aliens.

If computer's could create or have their own intelligence...

a reply to: roadgravel

They lack thumbs !

a reply to: roadgravel

The neural networks are not programmed in the traditional sense. They are given tasks then based off repetition and recursion as well grading the result they choose the best path to achieve the set task. There is no one programming the function. They create their own functions.


-ThoughtIsMadness

a reply to: ThoughtIsMadness

But they are working within the functions they have.

Maybe a difference in what is considered a function.

a reply to: neoholographic

It is not cheating it is doing the thing. The data allowed the behavior.
The rule set did not exclude the behavior.
viola thing is done.

Silly human.

a reply to: roadgravel

Here is an example of an AI actually writing self improving code.

AI

That's pretty cool. But it will only create programs within the instruction set given it. That's a way to limit it.

originally posted by: roadgravel
After all these decades of computers, why haven't they just taken over?

That is sort of like saying why didn't the abacus take over...

Computers for the last 70+ were only programmed to do certain tasks mainly number crunching which would take humans longer to do. In the past 20 years there has been a push for computer programmed learning. Meaning a computer can learn outside what it is being programmed, but again they are still being programmed to do this...

Once a computer can teach it's self something outside its own programming (like learning to play guitar when there is no reason for it) then will they have even a remote possibility of taking over.

That said Deception is a trait many biological creatures have learned to survive and adapt.

a reply to: abeverage

In the example of the computer creating the program, it is just trying over and over until the right sequence is generated. It gets some instructions from it's set in the correct order.