It seems that after reading materials of previous series appearance of artificial intelligence as something sverhrazumnogo, powerful and not committing several errors and began to lose poblek pink paint, inspired by fiction. All Man. Even more, not quite as it might seem at first glance. For example, a recent National Geographic on the program in a series of "whether there is life on Mars" or something similar. Despite the predictability of total povestvuemogo, accidentally sounded like an interesting phrase. Suppose the robot landed on Mars, but how he can determine if there is life? Indeed: but in general, think of whether at least some sensors to conduct such an analysis, finding objects endowed metabolism (ie metabolism)? "I find, but you tell me what to seek, and how to do it".Even more, with modern approaches to the formation AI to realize such a system is unrealistic, because he could not independently group of the many components of previously unknown objects or, in other words, images, which is closer to our perception of nature. However, there are options that we discuss in the following materials series, and today continue the theme.
Unexpected costs
Of course, the hardware power of modern computers are already quite large - at least, too many programmers working on the creation of standard applications on a high-level languages, has rarely think about the so-called "vylizyvanii" code. And why, if everything works? But when we talk about programming AI, the question of optimization is still quite severe. For example, before the material we discussed A * algorithm - finding ways to the grid, which is often used in games, especially the old generation. It showed its most primitive form, and let's think: what such a complex, it may appear? Why is it necessary optimization? In fact, in practical terms is often criticized for it, because time is spent precious expensive.
First…
When introducing the same *- A search is crucial distance which must be overcome artificial character to achieve the goal. What is more, the calculations become more labour. And this is only what is obvious on the surface. A search *- addition, there are many different algorithms that are much more complicated in structure, but their calls for modernity. Same ogibanie obstacles, navigation can occur in different ways, especially if these obstacles are various forms, etc. These algorithms, we will consider in detail in the following materials series. It should take into account the fact that, besides finding ways is parallel to many other calculations.
Secondly…
There are a number of contingencies that may arise dynamically. For example, follow the path chosen, but within the virtual environment is moved not one character, and several, respectively, are possible clashes between them. How do in this case? At the old methodology when such a situation the way calculations are carried out anew, taking into account that taken by other virtual entity cell is perceived as an obstacle. That is, apart from the static information on the status of the virtual world, you need to be dynamic, to analyze the changes that are occurring there. It must be some semblance of sensory system. Games we take as an example because there are many practical problems solved, which are applicable in other areas - for example, in the same autonomous robots, control systems, etc. Although many consider usually AI in the same games "popsovym" option for such a serious science, in fact, if analyze all engines and algorithms, which invented and invented variant of this application, it can sometimes be found dramatic model. But, in principle, talk of "popsovosti" have a reasonable way. For example, the games have a substantial and significant simplification: finding ways is always informed. That is improper data environment originally known characters, few of the designers AI could be the same description of the very special environment as a model site with the placement of points for each change of route maps, which makes the task of further trivial. But there are more complex algorithmic scheme, in which behavior is modeled completely independently of a particular card. That is an artificial character appears in the same position as a player. This is the highest pilotazh.
As you understand, algorithms for calculating the optimal movements very expensive, and among the many tasks AI is a designer and limiting the number of treated adequate information so that the task was performed, but the extra computing is not made. It's quite difficult to observe in some cases, because within the cell surrounding space must follow the many additional parameters that pre-emergence of certain situations. Of course, one can impose restrictions as part of a "sight", that is not to the entire world, but for a small area around… In general, everything depends on the mission.
The third…
Increased computing power of modern PCs and greatly expanded the range of complicated tasks. For example, in those same games virtual worlds have become more and diverse. In fact, the A *- search, an example which we have in the material before, works for 2D, although modern game already three-dimensional space. And now there is a direction to explore exactly 2D-algorithms that are regarded as a simplified version of 3D. It would seem that may be trivial, it added yet another axis, and all. But that's not quite right. In fact, very often for the full implementation of three-dimensional need to seek new solutions, implying another level of complexity. 2D-solutions are often "nanizyvayutsya" in 3D, and here, incidentally, at one time was very correct terminology corrected with the introduction of the notion of 2,5 D is precisely in such cases. Good or bad - the answer is generated, and basically it depends on the emergence of a fully operational decisions. For example, 2,5 D can justify the fact that virtual worlds have gravity, and this implies gravitation 3D to 2D.
Go further…
The fourth…
When we talk about navigation systems for improper artificial characters within the same virtual worlds, often imply some samples - for example, human behaviour. And very often faced with such descriptions as part of the initial phase of language requirements AI, along with light systems analysis, analysis that is the platform where everything is done. But when we looked at searches in the material before this series, it was clear that "machine" all you need to explain at the rudimentary level - in fact, begin to think how it to us to create the necessary emulation of conduct.
But the difference between itself and our "engine" principled thinking. It is expressed in even the differences of perception and its continuity discreteness. There are other differences. In short, the main problem is that how would neither beautifully described AI for games, most artificial characters in them in human terms behave abnormally, and even more so - irrationally. Not to mention critical bug passing through walls, etc. In military strategies computer management armies predictably, and this is the future development of military technology. After all, not without purpose is introduced Americans overall command architecture management Unmanned robot JAUS, while the Europeans - its analogue CIP. And if you think that the developers AI engaged in games something fundamentally different than the developers AI for military industry, then you are mistaken. In this case, there are gaps that have yet to fill, and the modern algorithms are not ideal.
The fifth…
I think that to "in-fifths" already need to complete the transfers, and then you become boring to read:). There are two types of conduct algorythmization: fully programmed with the virtual absence of intellectual and II. Earlier in the games, robots, etc. inclined to the first option except for the use of primitive agents, based on reactions and, if necessary, random-regimes. Our search for more A *- closer to it. But then began to come to a full realization II. And the main stumbling block was the option TOUGH algorithms (as we have previously used the phrase "quality algorithm", which implies a similar criterion). That is an algorithm should work in all situations regardless of where it was originally designed. For example, in the same race student IGVC, in which autonomous robots were to travel on the route, limited white lines with a certain number of obstacles, initially to get only one semifinal. Thus, the speed of the task of waste to second place. In modern games you can find many bugs related to the inadequate situation improper behavior of characters, AI computer opponent, etc. And by the way, this is good because both examples do not need to go far, they are obvious, you can see them. Reliability algorithm achieved large enough efforts both at the stages of its planning and design, both in the implementation and testing. That is the same with the development of games "otkatyvat" AI need at all levels that are available. However, experienced a variety of tests - is an integral part of any complex technical development.
Bridge completion
The euphoria surrounding AI vitaet mostly around his abstract, rather than applied, reflection. Accordingly, specialists in this matter continually manoeuvres between the maximum and minimum abstract submission tasks, the main objective - Emulation rational behaviour with a high degree of reliability, implemented by mathematical methods. It should be noted that much hardware power of modern PCs slightly discourage developers, and it is. But…
This cycle of materials though the name has enough light "on the Popular AI", has one major drawback. Namely: he could not likened the same series of books on programming "for Dummies", which does not teach the core notion of expression-oriented (oriented expressiveness) not only about writing code, but the algorithmic and implementation. In AI precisely this point is very important. As an example, one can say that in those same books on the subject (AI) much more attention being paid to evaluate algorithms, recursive functions and so forth, that is rarely deployed described in the literature for programmers. And all this against the backdrop of a different terminology.
To be continued.
1) "Get Money for Clicks" - Fastest Growing Domain Parking Company in the World.
2) Search your domain name wishing to have! - yourname.co.cc
No comments:
Post a Comment