📜 ⬆️ ⬇️

How to take the processors?

In the science of futurology, one of the main methods is the analysis of current investments in certain studies. Futurology proceeds from the fact that where no one invests anything, no fundamental changes can be foreseen (although, of course, there may be unexpected discoveries that confuse all the cards, but it is still impossible to predict them scientifically). Accordingly, if someone invests somewhere (for example, it is known that huge amounts of money are now being spent on applied genetics and microbiology), then it can be expected with some certainty that something will work out there. Further, these scientists go and ask people who are doing something in these areas, what goals they pursue and within what time frame, from their point of view, these goals are achievable. And then they make an amendment to optimism, connect their imagination and analysis, and complete the picture of the world in which there is something that these people have already done by that time.

I propose that we do a short time amateur futurology. We will try to answer a simple question: where will the processor race lead us?

It is absolutely certain that so far no one is going to stop significant investments in the development of new, more productive processors. Obviously, while there is a parallel search for methods of replacing silicon transistors with fundamentally new logic elements, manufacturers will try to squeeze the most out of existing technologies. How can I do that? First, by increasing the number of processors per chip, and secondly by specializing processors, which will allow them to more effectively perform particular tasks. Actually, no special futurology is needed here: this process is in full swing right now.

But the question is: we need to understand what is to load all these processor capacities in order to ensure their demand? Today, most processors are, in fact, idle. The average CPU load in an average machine does not exceed a few percent. Obviously, we need some fairly massive tasks that require large computing power. What tasks can it be?

I have one script. It seems to me that the future at the computer is a collector and analyzer of information from the surrounding reality. He will try to save, systematize and identify patterns on which to base decisions. In particular, he can learn the master's habits, recognize him, adapt to him, his mood, habits and desires. He can be a smart agent who searches for and finds on the network for the owner what he is interested in, based on an understanding of the meaning of the texts. All this, one way or another, recognition tasks. We regularly see successful commercial applications of such technologies. It is enough to recall gesture recognition to control games in XBox, face recognition sitting in front of a computer with webcams, face recognition in programs that store photos. Apparently, many people are engaged in this, and this direction is actively developing.

As a continuation of the first option, you can imagine the inverse problem: the reconstruction of reality based on the data obtained, realistic virtual worlds created by computers based on their own observations and networked, sharing information with each other. Then you can think what all this may need.
But most importantly, it will take all current and future processors seriously and for a long time: recognition tasks require large computational power.

Source: https://habr.com/ru/post/105456/

All Articles