What is considered artificial intelligence
What is considered artificial intelligence?
 |
| What is considered artificial intelligence |
Holy Grail of AI (Artificial Intelligence)
A long time ago, we were talking about intelligence in general And the way you can model intelligence as a process of improvement This is a hill climb algorithm. - Yes, this was an example we gave. We used Evolution as an example of an optimization algorithm, or an optimization system anyway. And then we were using that as a way to talk about other types of intelligence. We talked about the game of chess for AI very briefly. That kind of thing. So the question is: What is the difference between the type of artificial intelligence that we have now? The type of artificial intelligence that might play chess, drive a car, win a danger, or whatever - In exchange for the ideas that we have from AI in the future? This kind of science fiction Synthetic is what we call a real AI. What really has an effect? Is it simply a matter of solidarity or is there something different? The real distinguishing factor is the year. What this means is the breadth of a range of regions It can get better. So if you take AI chess,
it is very smart in chess, and it definitely is Useless in almost any other area. If you place AI chess in a google self-driving car, you can't drive a car that has no concept of what a car is. It does not have any of the necessary structure of Cognitive architecture to drive a car. vice versa? Google car cannot play chess You cannot win at risk. Where we have a practical example of general intelligence. It is human intelligence. right? Human brains can do many different things. In many different areas. Swallowing. Including brand new domains. Areas that we did not develop in particular. Actually a chess game, right? We invented chess, we invented driving. Then we learned to be good in them. So, general intelligence is, in other words,
a different class of things. Because it is an individual improvement system capable of improvement in a very wide range of different areas. And if we can build artificial general intelligence. This is a kind of holy grail for AI research You have one program or system that can solve any problem we address Or at least address any problem we address. - Finally, Brazilian Bradford ... Turing idea the test What amazes me with what you say is that it is a very specific field Pretending to speak man. Yes, in the sense that it is a very specific domain. The Turing test is necessary but not a sufficient test for general intelligence. hum. You can, it depends on how the test is formatted, correct Because you can say well, if artificial intelligence had to pretend to be a human, convincingly The original Turing test was in short chat using only the text But you can say, to convince me that you are a human: tell me what step I should take in this chess game To convince you that you are a human, tell me how I will respond in this driving mode Or what is the answer to this serious question? So in your Turing test, you can test a wide range of other areas on purpose. But in general, the conversation is one area Hum,
yes, you can craft the real Turing test, this way But it will get longer and be more reactionary. Another way of thinking about general intelligence is field-specific intelligence But where the field is the realm of the physical reality. And if you can reliably improve the world itself. This is what General Intelligence does. Have humans changed the world to meet their needs? Sure, so when you say change the world We clearly change the world on a large scale But everything that humans do in the real world is in the sense of changing the world to be better for them, right. As if I was thirsty and there is a drink there, then take it and put it on my lips and drink it. Change the world to improve moisture levels So I am, sort of, optimized I am using my intelligence to improve the world around me In a very abstract sense. But also very practical. - But on a larger scale, as you say on a larger scale, Build it and water the field put a tube on your house and then I will need a tab. -Yep. - He does the same thing but on a larger scale.
True, there is no solid border between these two matters. It is a similar platform for activity. The idea that you want things to be in a different place from where they are So you use your intelligence to come up with a series of actions or a plan, That you can implement, which will best meet your values. And that This is what real AI, the AI in general, does. If you can see The metaphor of optimization is still there, right. You have, this vast country area, which is all possible countries in the world Remember before, we were talking about dimensions And what kind of problem would be if you had many dimensions. (So when we have 2D space ...) This is what kills Amnesty International's general implementation Because the world is so complicated.
It's an exceptionally high dimensional space. With the example "I drink wine", I got the same thing again. You have a state of the world and it is a place in this space And you have another condition in the world which is the one you just ate a drink. One of them is the highest in two utility jobs. It is higher in my order, preferring the request of the countries of the world. So I will try to move, I will try to change the world From fewer places for preference to higher places. This gives you a way to articulate Develop plans and implement smart actions and behaviors in the real world Mathematically speaking. No, you cannot implement them just because they hum Because of this enormous dimensional problem. - All these dimensions, if you try to break the dimensions of the infinite force, it will fall very quickly. Yes, yes, immediately. Change the world. True, if this seems a little threat Hey ho. (Laugh)
Post a Comment