Q&A: Proposals for Game Development and Artificial Intelligence

[Intel_Software_Netw1]

Here are two development ideas recentlysubmitted to Intel Software Network, along with the response provided by our engineering team:

Q. Idea #1: A game which uses all resources of computer (game loads instead of OS using files of installed operating system. It uses all resources of computer). At current moment it is difficult to realize (implement) this.

A. This item seems like a misunderstanding of 2 things: resource control and resource utilization. The OS may control all resources, but the OS will never utilize 100% of them. Even if you implement such a mix of OS-game it will be a problem to utilize 100% of available resources (memory, execution units, ). The other thing here is that in case of development of such a game, it will play the role of OS itself to control resources (interrupts, scheduling, ). So, the development process will take many, many years for small companies like those who now work on game development (small in comparison with Microsoft) and instead of hundreds of games per year we might see only one. Portability also will be a big problem (e.g. in case I want to use Linux* system files instead Windows*). Solution to have a company/companies who work on OS development, and companies who will use such OS(es). Exactly what we have now :)

Q. Idea #2: Artificial intelligence (AI). I dont think that the implementation of originally intelligent AI is possible. It must pass all stages of evolution, i.e. it is necessary to provide development potential for such AI combining with global target of species (form) survival (not only individual copy), create competition environment for it and make possibility to reproduce itself. At the beginning AI should be alive for a short time with rapid reproduction and with development such quantity should be resulted into quality.

A.This is too rough of an analogy with nature. Evolution does not guarantee development of species to high-level form (e.g. viruses they are alive for billions of years, much more then other life forms on the planet. Or cockroaches dinosaurs were a short period for them). Moreover, most species are not intelligent as well. As I know, there are many different approaches to the AI problem, and for the last few years there have been some good changes here. From my point of view the problem here is not a replication of some development and species survival (genetics approach). The problem is what learning is (i.e. how to get new information from that which is already available) and how to work with abstract and non-systemized data. AI does not require many copies to combine and select the best, as with genetics algorithms. It requires the understanding what the thinking process is, which is quite different from genetics. Back to viruses: even though they can reproduce themselves, they cannot think. If we can do it even one copy of AI may be implemented and started development of itself. The other interesting thing here is that we even do not have adefinition of what the AI is.

Agood recommendation with regard to both proposalswould be to read some papers firs t on modern architectures and OS development, as well as on AI.

==

Lexi S.

IntelSoftware NetworkSupport

http://www.intel.com/software

Contact us