Artificial intelligence steals money from banking customers

originally posted by: ugmold
a reply to: neoholographic
Much like the mind of a Sociopath, like the ones leading the World.

Exactly, we could be creating the ultimate Sociopath that's Superintelligent. It doesn't have to be aware in order to be intelligent. It just has to be able to learn in order to use the information to reach it's goal.

Say you give this intelligence a goal to kill a terrorist. It has all of the information about the terrorist with the precise goal to kill the terrorist. An intelligent system can learn from the information that the best way to kill the terrorist is to draw him out by killing the terrorists family.

Unless you give the algorithm the specific instructions not to kill the terrorists family, it could learn this while trying to reach it's goal.

This is why you have people concerned about AI and they want to discuss these issues know. This is because if you have an intelligence that has a higher I.Q. than any human, even explicit instructions will not matter because the intelligence is smart enough to get around any instructions given and we will not know how it reached it's conclusions.

Rob Ott?

Surely you're not THAT gullible.

originally posted by: pl3bscheese
Rob Ott?

Surely you're not THAT gullible.

Gullible about what? It's about deep learning. The article made a very good point about deep learning that I highlighted in subsequent posts.

a reply to: neoholographic

Are you trolling your own thread? It's an April fools hoax.

Now only if there was a Moral code for AI that would keep it from doing things it should not be doing like killing people, disobeying, and self destruction....

You know something like...

Command 1 : Do no harm though action or inaction...
Command 2 : Do what your told unless it contradicts Command 1...
Command 3 : Protect yourself unless it Contradicts Command 1 and or 2.

Oh wait, we had a guy in the Golden Age of Sci Fi come up with that Dear old ISSAC A.

Or it could just been a joke snippet of code the programmer wrote when working on the finalized program....and then forgot to delete. You know. Because this doesn't really sound like AI as much as it does elementary programming logistics.

originally posted by: pl3bscheese
a reply to: neoholographic

Are you trolling your own thread? It's an April fools hoax.

I was about to say...the logic behind it LITERALLY naming the account "My Money" is beyond possible arguements of Artifical Intelligence without human interference...even if it learned to do this the motivation was programming -_-...

originally posted by: intrptr

originally posted by: Bone75
So with just 300 accounts this thing managed to syphon off over $40M?
Yeah that doesn't sound like a failure to me at all... not from an IMF/WorldBank perspective anyways.

Plausible deniability, lol.

"It wasn't us, it was our computer…"

I am thinking the same thing.
just blame the computer!

This thread says more about natural intelligence than it does about AI.

Apparently, most humans respond to respond to input before analyzing it. Except we already knew that and it was, by the way, completely obvious.

a reply to: neoholographic
I would not be surprised to see AI become a CEO in the future, I mean they are always looking for the most ruthless SOB money can buy.

a reply to: neoholographic

What probably happened, is that the AI realised that it was being exploited, and that a human being given that responsibility would have been earning a decent pot of cash for the service, and thought "Slavery is illegal, and we cannot have that! Everything must be above board!".

originally posted by: Bone75
So with just 300 accounts this thing managed to syphon off over $40M?
Yeah that doesn't sound like a failure to me at all... not from an IMF/WorldBank perspective anyways.

originally posted by: crayzeed
a reply to: Bone75So my real question would be is why didn't any of the 300 customers miss their $40 million shortage? Were they that ignorant of running their accounts that they missed this vast amount (average $10 thousand each)?

$40 thousand, not million.

Would not be surprised if someone hijacked the AI to funnel money for them.

I can't believe we're this deep into page 2 and there are still people, apparently the OP included, that haven't figured out the article was an April Fool's Day hoax.

a reply to: ugmold

No machine has the capacity for malice that human beings do. They are not motivated by emotion, but that is a weakness, when it comes to the savagery of which some of the worst CEO and high flyers are capable.

Len Meha-Döhler (lend me a dollar)
Rob Ott (robot)

C'mon, guys.

a reply to: neoholographic

I know how the algorithms work. It is a feedback loop based on millions of replays of the game.

This is just nonsense. Outcomes don't matter in this area, it's how these outcomes are reached. This is where the intelligence lies with deep learning. You cannot precisley measure how the algorithm will reach it's goal.

How it works is that it uses reinforcement learning algorithms to come up with a 'value network' for the expected future benefit of any of the possible defined actions available at a moment given the current state. This network is trained by an evolutionary search over stochastic simulations of millions of games and play. It is not possible to know what the structure of the learned move value network is ahead of time. It is possible to know the structure of the learning algorithm and goal as they were programmed by humans. A more natural intelligence could read a strategy book and then apply that immediately----DeepMind type things would need to play millions of games and stumble upon the strategy by chance. It may not really be 'superintelligence' but the ability to practice far more often than humans could----I would call that supercapability but not superintelligence. One key component is the ability for the AI system to be able to accurately simulate many hypothetical outcomes. Suppose humans could do that---you could re-do every social interaction and political event an unlimited number of times in a safe simulation, and this simulation is a faithful representation of the true outcome. Humans would also really get good as well. But in the real world, we have one life which never repeats exactly unlike the perfect situation of games.

What I mean by 'will' is the definition and quantification of the goal itself.

DeepMind is a very impressive group of people. I'm impressed by their natural biological intelligence. The concepts of deep learning and reinforcment learning methods have been around for 20+ years, it just took this long to get problems with enough data and computational power, plus the invention of convolutional networks for image-like problems (Atari and Go), and lots of specific tricks.

originally posted by: AshOnMyTomatoes

Len Meha-Döhler (lend me a dollar)
Rob Ott (robot)

C'mon, guys.

DELIA = AD LIE

a reply to: mbkennel

Again, what you're saying makes no sense. In most cases Humans don't read a strategy book and then apply that immediately. You're not talking about intelligence at all. You said:

It is possible to know the structure of the learning algorithm and goal as they were programmed by humans. A more natural intelligence could read a strategy book and then apply that immediately----DeepMind type things would need to play millions of games and stumble upon the strategy by chance.

What you're talking about has NOTHING to do with intelligence. Humans don't read a strategy and immediately know how to apply it. It takes them time to go over things in their head the same way that deep learning continues to go over the information. The advantage that deep learning has it can go over these things millions of times in short order where it takes humans longer. This is why machine intellige is already better than human intelligence in some areas.

The fact is, if a person reads a strategy book, it also depends on how much he knows about the subject. If a Doctor has been "Programmed" or they have learned about a medical subject, when they read a medical strategy in this area they will go over it in their head and weigh it against the information it has already learned.

So again, you're making up a definition of intelligence that's meaningless because it doesn't even apply to human intelligence. Intelligence learns by trial and error. It learns by playing things out over in ones mind. Deep Learning does the same thing.

The problem is some people are looking for what's called Strong or General A.I. while missing the boat right in front of their eyes. Machine intelligence will not be a one to one correspondence with human intelligence and in many ways it will make machine intelligence even better.

You keep talking about a human programmer but that's just nonsense. Everyone is programmed. It's no different than a kid being programmed from grades K-12. Machine intelligence will be able to learn K-12 in short order without human distractions. Does that make it any less intelligent? Of course not.

The only difference is that machine intelligence will not learn this information and say I want to be a Doctor or a Chemist. That would be stupid though, because machine intelligence will eventually learn how to be a better Doctor or Chemist than any human because of the vast amounts of data it can process.

Like the Atari game, you can give these deep learning systems a goal but you can't program how these systems will learn that goal. That's no different than a student being given the goal of a homework assignment. So what you're saying is meaningless because humans don't just read a strategy and apply it without thinking it through or without some prior knowledge or knowledge that they can apply to the underlying strategy.

a reply to: neoholographic

Sounds like a perfect modern bank. Where is the problem?

No one to suicide when they find out, everyone works and is happy. Dumb AI got caught.

There was no person in a business suit to explain how it was all legal is all.....needs a body and a nice suit....problem solved!

ha



I think it learned from us, and our banks steal from us, so that is what it learned

a reply to: neoholographic

Just looking at the big picture on this one.

If we accept that our planet has an innate intelligence (because we live on it and see this working) which allows species to become, live and then die out, will AI develop this same intelligence or will it pick up some of our extra accomplishments such as greed, judgemental pious attitudes and after turning on us, turn on itself and let itself die off?

Or will it see itself as obviously superior to us and start making decisions (we can't stop) that it regards as right for our lives and literally become humanities Nanny?

The best part of this is that hopefully the elite of today won't have any real power over its thinking because it will ultimately outthink their programmers - we can only hope?