The problem of designing an AI to understand what it sees

I have been working on this very problem myself for some time but so far have made very little headway; I will still keep on trying and inform you all if I ever start to make any real progress.
I have found this aducational video that helps to explain why this is such a huge and difficult problem as well as some of the recent progress in AI to help solve it;

YouTube

On a related topic; This is a link about evolving AI to (hopefully) have human-like intelgence by making it evolve in a computer simulation that simulates evolving intelligence via mutations and (simulated) natural selection in a similar way to how real intelligence evolved in nature;

https://phys.org/news/2019-09-evolution-key-artificial-intelligence.html

I had independently came up with exactly the same idea but designing the simulation so it would work in practice without taking an eternity to come up with something really good would be extremely difficult. One day I might give it a try.

@humy said
On a related topic; This is a link about evolving AI to (hopefully) have human-like intelgence by making it evolve in a computer simulation that simulates evolving intelligence via mutations and (simulated) natural selection in a similar way to how real intelligence evolved in nature;

https://phys.org/news/2019-09-evolution-key-artificial-intelligence.html

I had independen ...[text shortened]... y to come up with something really good would be extremely difficult. One day I might give it a try.

Do you think understanding can come from 1 and 0? Isn't that like suggesting you can give a piece of knowledge to stones that you can move around to represent numbers?

@KellyJay
You are just showing how far behind technology you are, Intelligent AI will be made of complex neural nets not 1's and 0's.

@kellyjay said
Do you think understanding can come from 1 and 0?

No, and we wouldn't.
Get up to speed with modern science. Start here;

https://en.wikipedia.org/wiki/Neural_network

@sonhouse said
@KellyJay
You are just showing how far behind technology you are, Intelligent AI will be made of complex neural nets not 1's and 0's.

Breaking it down to the all that is going on is calculating and doing that fast, knowledge is different than a database, or a robust algorithm, outside of life it isn't found. So I think you are elevating what can be produced with our hardware and software while diminishing life and its many experiences.

Still, what you are doing is not a wasted effort, in my opinion, even after acknowledging my views on what I just said. Copying what is in life or attempting to mimic is wise! Granted, you think life is all accidental due to natural processes acting on what material was at hand on the planet. It is so much more, far superior to what we can do with our insight in all things.

@kellyjay said
knowledge is different than a database, or a robust algorithm,

Right; its the understanding, expressed as certain mental processes, of defined relations (I elaborate on the technical meaning of that on request) of information of what is or what's going on. Why can't an AI be given that? My OP link shows an AI that clearly has some knowledge and understanding of the world because it can look at a picture it has never seen before of the real world and often describe in natural language what it means and often about as well as a 4 year old human.

outside of life it isn't found

False. I am a qualified expert on AI and I can assert that is false. It hasn't been yet achieved at nearly the same advanced level as that in human mind (a 4 year old human currently does better, but not always a one year old. But that's changing) but still it has been already achieved at least at a primitive level and its becoming more and more advanced and its just a matter of when, not if, it can be achieved to the same level as that of humans. You would see that the content of the above links I have shown, if only you look at them, clearly hint of that. What is your expertise on AI?

So I think you are elevating what can be produced with our hardware and software

What is your expertise on AI? It is better and more humble and intelligent to have no opinion on something you know nothing about than one based on ignorance.

while diminishing life and its many experiences.

How would achieving human-like AI intelligence be "diminishing life and its many experiences"? In what sense "diminishing"? To which aspect of life? To any particular experience? Give an example...

@humy said
Right; its the understanding, expressed as certain mental processes, of defined relations (I elaborate on the technical meaning of that on request) of information of what is or what's going on. Why can't an AI be given that? My OP link shows an AI that clearly has some knowledge and understanding of the world because it can look at a picture it has never seen before of the real ...[text shortened]... what sense "diminishing"? To which aspect of life? To any particular experience? Give an example...

Knowing what it takes to produce the color red on a monitor is different than the experience of seeing it.

@KellyJay
Eventually AI will do both.

@sonhouse said
@KellyJay
Eventually AI will do both.

The most basic truth of computers is that they don’t always do what we want, just what we tell them. You get a program to stop behaving as a program you have a useless program.

@kellyjay said
Knowing what it takes to produce the color red on a monitor is different than the experience of seeing it.

What has this got to do with useful AI with ability to do all the types of practical reasoning like we can do? Such AI doesn't need to have sensations and/or emotions.

@humy said
What has this got to do with useful AI with ability to do all the types of practical reasoning like we can do? Such AI doesn't need to have sensations and/or emotions.

It is the difference between a life experience and moving rocks around the ground to acquire an answer. The answer will never be understood in the rocks no matter how you arrange them.

@kellyjay said
The answer will never be understood in the rocks no matter how you arrange them.

The "answer" to exactly what question relevant to potential AI intelligence?
Or have you just suddenly changed the topic of this conversation?
And what has "moving rocks around the ground" got to do with it?

@humy said
On a related topic; This is a link about evolving AI to (hopefully) have human-like intelgence by making it evolve in a computer simulation that simulates evolving intelligence via mutations and (simulated) natural selection in a similar way to how real intelligence evolved in nature;

https://phys.org/news/2019-09-evolution-key-artificial-intelligence.html

I had independen ...[text shortened]... y to come up with something really good would be extremely difficult. One day I might give it a try.

Do you have any concerns over the use of AI technology? Aside from dystopian novels and movies which sensationalize the threat, modern AI systems have been described (even among people directly involved in the research) as a threat to humanity. Monopolistic dominance over humanity by a select few companies, outsourcing of ethical decision making, privacy, inherent bias. Some argue that several of these threats are already playing out, for example social manipulation through data mining and purposeful dissemination of false information, and we seem to have very little institutional power to prevent it. Are there sufficient safeguards in place to ensure that AI tech is used in a responsible and safe manner?

I'm not an expert on the capabilities, so I thought I'd ask. If a self-driving car 'understands what it sees', and there is a pedestrian crossing the road that requires an evasive maneuver that may put the passengers at risk but the outcome cannot be determined in real time, how do you program that decision into a car? or do you expect it to make the right decision? Can that car (or its programmer) be put on trial for vehicular manslaughter for failing to stop?

@wildgrass said
Do you have any concerns over the use of AI technology? Aside from dystopian novels and movies which sensationalize the threat, modern AI systems have been described (even among people directly involved in the research) as a threat to humanity. Monopolistic dominance over humanity by a select few companies, outsourcing of ethical decision making, privacy, inherent bias. Som ...[text shortened]... on? Can that car (or its programmer) be put on trial for vehicular manslaughter for failing to stop?

There are some potential dangers from the misuse of AI but this is true for any technology. I have no doubt there will be some future misuse of AI as this seems to be like a law that if it can be misused then it sometimes will be. But I don't see any credible special AI danger that could exist any time soon that would make me lose sleep. I once considered writing a whole book about novies which sensationalize the AI threat and explaining all the many things that are very wrong with their plot and why it would be very unlikely to happen anything like as bad as the way they showed.
Yes, there are dangers; but I believe, unless we go stupid about it, they can easily be made all manageable and not scary.

However, that said, I think it would be extremely STUPID for humanity to not take just a few legally precautions enforced by international laws just to make sure no particular super-intelligent AI singularity goes murderous and completely out of control. In my opinion, the international law should state that the programs controlling all super-intelligent AI should not ever be kept secret and instead should be made open to public inspection (but not publicly open to reprogramming else they could be hacked) so that AI experts can check that they are safe. There law should also state a serious of basic ethical principles that, by law, aIl super-intelligent AIs must have; No exceptions. And there should be a paid large body of top AI experts to constantly check super-intelligent AIs and hunt for possible dangers to rectify.

I am not sure exactly what the answers are to your latter questions involving car, but any practical AI driving cars etc must be designed to be able to deal with ethical dilemmas in its decision making process at least as well as that of your average 'good' human. How an AI might exactly best deal with ethical dilemmas is one of the things I hope to one day get around to research as I think its an interesting problem but have so far I have not got around to making any start on that.