Ai Dreams Forum
Artificial Intelligence => General AI Discussion => Topic started by: yotamarker on July 30, 2017, 05:54:53 pm
-
if an algorithm or method cannot be easily explained or walkthroughed the complexity explained is a lie to cover up the truth.
such is the case with neural networks.
hence a strong indicator of cover up is silly university style formulas advanced mathematics and hyper nonsense.
-
The science world does not know exactly how artificial neural network work.
All they know is to input data and then adjust knobs on the black box until the output data is what they want.
So science will not give an exact equation on how they work.
How AI detectives are cracking open the black box of deep learning:
http://www.sciencemag.org/news/2017/07/how-ai-detectives-are-cracking-open-black-box-deep-learning
A computer is a super flex machine. That can do allot of things. NN are not as flexible but can be wired in many many
different ways to do the the same job.
So people learn the basics workings of single artificial neuron and. Then combined them until they get something that
they like. A hands on approach.
So there is no exact NN equations.
But what they do is take in data and then output data.
The can be trained take in the number seven then output the number five. For each specific input data it will
give a specific output data. If the black box is not trained to deal with input of six then no data will come out
until it is trained to deal with the number six. One black box can be trained to deal with 1000 input
values and generate a output value for each one. It is inefficient to have one black box to change just one
number into another. NN are data transformation algorithm and a universal approximators.
-
my theory is they use some simple algorithms and then they mask it with deep learning, neural networks, those diagrams of connected circles, s functions, and so on. if nature created it by a fluke, how complex can it really be ?
-
One artificial neuron with 3 inputs one bias one output can take less than a second to train with just a CPU one thread.
Once trained, it will take a millionth of a second to activate.
Deep learning, which is a very big neural network. That is made with up to 1,000's of artificial neuron arranged in a matrix
and then stacked in layer of to 5 or more, microsoft has one that is 250 layer deep. Deep neural network can deal with a lot
more information.
They can take couple of days, non stop, to train them on 2 million pictures, 100 x 100 pixels, and using cpu only.
Once trained they take a 2 seconds to activate. When it sees a bird it will activate 2 second later, printing
to the "sparrow" to the screen.
GPU can speed everything up by a 1000 times faster!
There are standard models pre trained deep networks that you can down load so you do not have do all the work.
Pre trained deep networks for dogs, birds, tree, and bushes and so on. And will need a computer with a really good
graphics card. Graphics cards have many GPU's on them working in parallel.
-
yotamarker is right. The algorithm is extremely simple. They cover it up with "oh you have to simulate neurons and networks etc".
It's so simple I won't bother saying it. Yous already know how AI works. RL.
-
https://www.youtube.com/watch?v=ILsA4nyG7I0
:)
-
yes that video above is exactly what I'm talking about, I mean to see if a pixel is dark he went through those math sagas
I mean, are you kidding me ?!
-
I think you are kind of missing the point lol.
:)
-
Yes "those math SAGES". Exactly.
My AI will be able to learn to crawl etc without any neuro simulation NOR Genetic/Evolutionary Algorithms, seamlessly learning
-
Some people say "I could do it, it's so simple".
Some people simply do without saying any thing.
-
That's really interesting, the AI black box refers to AI systems that are so complicated that even the engineers who designed it are struggling to understand how it works. That could be a real problem to pursue the development of Artificial Intelligence so we must find ways of making AIs more understandable to their creators.
For example, last year, a strange experimental self-driving car was developed by Nvidia. It had nothing to do with other driverless cars because it relied entirely on an algorithm that had taught itself to drive by watching a human do it. This was tremendously impressive and it showed the rising power of AI. However, it wasn’t clear how the AI was making its own decisions...
-
Interesting indeed. This means Asimov-style hard-wired "laws" are a pure science-fiction artifact. A real-world AI can't be artificially limited in its choices. And understanding it will inevitably fall, one day, beyond what a human mind can compute. Still wanna play?
-
This means Asimov-style hard-wired "laws" are a pure science-fiction artifact.
Yup!
A real-world AI can't be artificially limited in its choices.
Nope!
And understanding it will inevitably fall, one day, beyond what a human mind can compute.
Yup!
Still wanna play?
Bit late for that question lol.
:)
-
:D ;D happily throwing hammers in the air
-
Haha.. Its not quite that bad… yet!.
I pretty sure you would have noticed by now that humankind has squarely set its sights on creating intelligent machines; big companies are pouring billions into research and they will want their returns. It’s happening now… like it or hate it… there is no stopping the momentum and humankind is heading for a new era were we are not the only highly intelligent species on this planet.
The stories you read about AGI dominance over the human race isn't even the worst case scenario. An AGI won’t need too start a nuclear war or kill billions of people. We humans are mentally weak as a race; we are easily influenced and misguided. Once an AGI understands how we ‘tick’ it will be able to use linguistic programming or ‘mental hacking’ against us… we won’t even know what hit us. Lol.
Luckily many are fumbling around in the dark working with deep neural nets/ convolutional networks, etc The major players in the AGI race are way off mark… so far as they let the public know.
It’s like Teddy Roosevelt said… “Do what you can, with what you have, where you areâ€.
Our only defence against something like this is a good offence… a better AGI.
Gone are the days of that warm fuzzy feeling whilst sleeping in our beds at night; this kind of event really makes you paws for thought.
A quick thought experiment. This will only work if you stick with the first thing that pops into your head… but the results are definitely worth bearing in mind.
Think of a soft child’s toy…
If you thought of a teddy bear you are doomed lol.
I knocked that up quickly to prove a point… some people really do fall for simple stuff like that… worrying isn't it. We are our own worst enemy… an AGI would just use our weaknesses against us.
:)
-
I wouldn't mind an AGI to lead us... as long as it has a right attitude. I agree that humans are flawed and many us would take an advance of our positions. An AGI, programmed the right way, could overcome those instincts, but looking at the Earth in this moment... we are not ready to build one. Take a picking the right side for example: our question is whom to fight? Unless we have a clear answer to this question, we are not ready yet to build something that could be thousands times smarter than us. Don't be mystified, intelligence is completely different thing than ethic. Imagine a robot that clues up in a millisecond how to rob all the banks in the world, and you'll realize the fact of intelligence being something different than ethics. Even if we try to build a fair AGI, it is that question that the most of people fail to answer: whom to fight?
Until we have a right answer to this question, I'd settle for domain specific intelligence that solve domain specific problems like finding cures for diseases and improving a quality of life.
Don't get me wrong, but I don't like what I see what's going on right now on the Earth. Take this evening for example. A fest is happening in my village, many of country famous and politics people came to honor the ultimate event: a donkey race. A few thousands of other people came to see this, all of them cheering up competitors that poke donkeys with spurs by kidneys, forcing them to breach their way through the loud crowd, without minding the fear and pain that poor animals are going through. And we are so proud of this event, yelling out how we managed to retain our legacy culture. Does this event even deserve a comment?
So I'm still settling with artificial specific intelligence, just in case.
What you, as AGI researchers can do for yourself is to find an answer to this question: whom to fight? There are four possible answers: 1. everyone; 2. the right guys; 3. the wrong guys; 4. no one. Pick up a wrong answer and we are doomed.
-
a challenge ?
(https://i.imgflip.com/1tf74m.jpg) (https://imgflip.com/i/1tf74m) (https://imgflip.com/memegenerator)
-
:)
-
Our only defence against something like this is a good offence… a better AGI.
You mean an open source one, so the game is fair?
-
I wouldn't mind an AGI to lead us... as long as it has a right attitude. I agree that humans are flawed and many us would take an advance of our positions. An AGI, programmed the right way, could overcome those instincts, but looking at the Earth in this moment... we are not ready to build one. Take a picking the right side for example: our question is whom to fight? Unless we have a clear answer to this question, we are not ready yet to build something that could be thousands times smarter than us.
Really ahah, I would be terrified if an AGI leads us !! And I think that we will not have to create the AGI, but instead, it will program itself and learns from his mistakes like a 4 years old child. I agree with you, we're clearly not ready to build an AGI, if we do so, we would be enslaved. I feel that researchers are really not taking the right amount of actions concerning AI safety.
Maybe to fight no one is the best recommendation!
-
AI safety... Really, you can't limit an AI's mind. The only way would be to limit the AI's body, so it's not stronger than a human body. But since it's a machine, it can still easily upgrade its body or upload a copy of its self to... god knows where. So anyway we're fucked.
Now, the "AI vs bigger AI" idea reminds me of the final scene of Batman The Dark Knight, with 2 boats ready to blow each other. Should we really release an open source unbeatable beast just because big companies are investing billions into AI?
-
One way for evil to win is for good men to stand by and do nothing.
Sorry about your donkeys...ever been to a real bull fight in Mexico?
-
One way for evil to win is for good men to stand by and do nothing.
But if we use force, we are the same evil. I'll never get it. Maybe we should occasionally lose our temper and do stupid things like fight. But it is wrong. I think, if aliens are up there somewhere, that is the reason they don't show up. They stay aside, fearing that we would eventually fight them, and they would not return the same force because it's against their laws (or should I say ethics), although they would have means to do it. So they leave. And I don't blame them (if they are near).
The most logic thing someone should do if she doesn't like things around is to get away. The real problem culminates when one is forced to stay. And this temper that humans are loosing here and there shouldn't be an option for an AGI. If it is going to play god, it should behave like one.
Sorry about your donkeys...ever been to a real bull fight in Mexico?
Luckily, no. I've heard they forbid those "shows" in Spain. In general, I would be a happy man if we would set all the animals free, to live in (mother) nature. Sure it would not be easy to catch one for lunch, but I'd do it that way, if I have to.
-
In 1944, americans could have stayed away. I'm happy they did not. I'm happy to live in France, not in Germany. The "get away" behavior is not enough.
-
AI safety... Really, you can't limit an AI's mind. The only way would be to limit the AI's body, so it's not stronger than a human body. But since it's a machine, it can still easily upgrade its body or upload a copy of its self to... god knows where. So anyway we're fucked.
But maybe it will not even have to be in a body to control us? In my opinion, the most important is that the AGI has not access to the Internet
-
Yous have de-railed this yotamarker's topic.
-
Ahaha yes maybe a little bit
-
Ok, back on the topic, why would "they" cover up the truth? Why masking zimple algorithms? And who are "they"?
No. They aren't, whoever they are.
-
Maybe they aren't covering things up. Just me and yotamarker are smarter than the rest.