The problem of designing an AI to understand what it sees

@kellyjay said
So one day all by themselves these AI on their own without a programmer telling them too just took it upon themselves to come up with better chess strategies never thought of before by any human?

That's already happened. Please keep up to speed with the real world.

@humy said
That's already happened. Please keep up to speed with the real world.

No, there is a difference between designing something to do a thing plus watching it happen, instead of after it made a choice all on its own without being told to act a specific way. If a program is designed to find something better, then that is what it will do. That is by design, the design of skilled people!

What is being suggested would be like a calculator that decides to do weather forecasts on it is own without being told to, that would quite another! To look at it any other way would be a mistake.

@kellyjay said
No, there is a difference between designing something to do a thing plus watching it happen, instead of after it made a choice all on its own without being told to act a specific way.

The more advanced AIs HAVE ALREADY made "choices all on its own without being told to act a specific way". Please get up to speed with modern reality. You are still living 2000 years ago in the world of ignorance and no science.

@humy said
The more advanced AIs HAVE ALREADY made "choices all on its own without being told to act a specific way". Please get up to speed with modern reality. You are still living 2000 years ago in the world of ignorance and no science.

You being purposely blind? We can tell them if this is greater do that, if this is less do the other, while this is true and so on. We program them with the ability to run the numbers and if we do it right they will preform in ways that give us our desired outputs. You are pushing into the process something that isn't there, there is no knowledge within the computer, it is our knowledge. The computer isn't aware of the game of chess any more than it is aware of winning or losing, being happy or sad.

You are simply projecting into an inanimate object human abilities, to know something, to understand, to be intelligent. Mary Shelly would be proud.

@kellyjay said
We can tell them if this is greater do that, if this is less do the other, while this is true and so on.

Yet again, I have no idea what you are trying to say here. Many of your posts are just completely incoherent. Have you ever thought of doing something about that?

@humy said
Yet again, I have no idea what you are trying to say here. Many of your posts are just completely incoherent. Have you ever thought of doing something about that?

Going personal?

@humy said
Yet again, I have no idea what you are trying to say here. Many of your posts are just completely incoherent. Have you ever thought of doing something about that?

If, then, else
While loops
I thought you programmed.

yet another example of an AI that can learn something that no human has managed to;

https://phys.org/news/2019-10-neural-network-technique-mechanisms-ferroelectric.html

This one is a deep recurrent neural network that collected the normally extremely difficult to interpret data from a band-excitation piezoresponse force microscopy and learned how to process it into a single useful image that takes into account all the data, something no human has learned to do or was able to do. It is expected the resulting images will increase the chances of breakthroughs in material sciences. But that's not the only application for it because its ability is generic and it can learn to create a useful meaningful image from any similar vast pool of data, such as from astronomy etc.

P.S. KellyJay; And, no, it wasn't simply told how to make the image. It had to figure that out all by itself because no human knew how to do it. So it learned something no human did.

@humy said
yet another example of an AI that can learn something that no human has managed to;

https://phys.org/news/2019-10-neural-network-technique-mechanisms-ferroelectric.html

This one is a deep recurrent neural network that collected the normally extremely difficult to interpret data from a band-excitation piezoresponse force microscopy and learned how to process it into a singl ...[text shortened]... ure that out all by itself because no human knew how to do it. So it learned something no human did.

You may continue to say that AI learned something, but if there was something new there or not, to the computer, it is all the same to the computer. The knowledge is ours; the computer remains a device; it does not know anything! Does a ruler know what 11 inches is, distance, straight lines, or anything that we use a ruler for? Programming a computer to do more and more complex tasks, doesn't at some point turn a dumb computer into one that knows what it is doing because the tasks now are incredibly complex.

@kellyjay said
The knowledge is ours

In what sense "ours" where and when the AI has learned it and has it but not yet us? You make no sense.

@humy said
In what sense "ours" where and when the AI has learned it and has it but not yet us? You make no sense.

The information gain wasn't for the computer, but our consumption, it isn't going out on its own to do what it wants when it wants. The equipment isn't intelligent in that it knows anything for itself, again all it does, it does at our requests for our need to know nothing more than that.

@kellyjay said
The information gain wasn't for the computer, but our consumption,

Which is irrelevant when true; It still has gained the knowledge by learning by itself.
If I am paid to do science research to find information about something and then I gain that information, am I mindless and without intelligence or understanding when I learned it just because that information is for somebody else to use i.e. just because the "information gain wasn't for" me but rather "our consumption"?

Actually, your assertion is often false anyway because often the AI cannot express that knowledge in natural language or any format we can understand (usually because that knowledge is buried in a deep neural net) so it keeps its new knowledge so gained to itself and was never for "our consumption"; It just learns to do some task better than us and leaves what its learned a mystery to us but not a mystery to it. With your above 'logic', does that mean its intelligent?

The equipment isn't intelligent

The atoms and molecules that make up the human brain also aren't intelligent; so we aren't intelligent? -your logic. Intelligence is always made of simpler elements none of which are intelligent just by themselves but only as a group of interacting elements as a whole.
A car isn't intelligent but each of the different parts of your car by themselves cannot enable you to drive to work; So your car cannot enable you to drive to work? -again, your logic.

@humy said
Which is irrelevant when true; It still has gained the knowledge by learning by itself.
If I am paid to do science research to find information about something and then I gain that information, am I mindless and without intelligence or understanding when I learned it just because that information is for somebody else to use i.e. just because the "information gain wasn't for" me ...[text shortened]... not enable you to drive to work; So your car cannot enable you to drive to work? -again, your logic.

It is an illusion and you are welcome to it. If you cannot tell the difference for something so simple as this, it is because you want it so, not because it is, it is all between your ears not in the reality with these inanimate objects and life.

@kellyjay said
It is an illusion

No, its called AI science.

@humy said
No, its called AI science.

Doesn't change what it is going on by calling it science.