Author: the workplace "reporter Dan stone
The general trend in scientific frontiers of computer technology is pregnant with the major changes, although today the computer's artificial intelligence or special wisdom, but who knows in Microsoft Research Asia this is a global leader in computer research institutions will not be on a certain day of the "health" of a computer's universal wisdom, especially the male of this encounter xufeng technology madman.
If Xufeng male with Chessmaster ksparov too, he wins the possibility of almost zero; if the latter is his design will play chess computer too, so the chance of winning xufeng male could be 100 per cent.
Although there is already a decade away, people still clearly remember May 1997 of the chess machine century wars.
Computer "dark blue" overcome Chessmaster ksparov, become worthy of "world champion". Since then, the "dark blue" behind the xufeng male was the father of the "dark blue" title. Now, the "dark blue" and "Dad" sat in my eyes: plain appearance, grizzled speed very quickly. Often to one subject, he is going to stalled for more than ten seconds, for fear is in the data processing in the brain, then began to talk.
Some years later, the male will not xufeng will become the father of the other, unknown, we just knew that he was the great crazy studied harder, he became the father of the "what" are reasonable.
Man-machine-game or a game of everyone?
Do scientific research of people seem to have such a proposition "stubborn", this stubborn in obvious choice male to Carnegie Mellon, meet the "man-machine-game" topic, the more obvious.
He always said he went "man-machine-game" road "is a chance". "Professor came up to me and asked me to help me to complete a single can of computer chess. I like playing chess, like computer, also know let computer society chess efforts has lasted for many years, but until then, the computer still has the power of amateur level discs. I like to do it, but I can't agree with Professor plan. ”
They are going to follow the chess set of 64 squares make a chip, so it was a great thing.
"I think this is stupid, Professor said that today's technology can take these things into a wafer, why do 64 block Professor insisted that pop?: If you speed up computer speed, the inevitable increase in the volume of the computer, I said ' not sure ', I came into conflict with the Professor, Professor that I don't want to do. Of course, then I decided to do, and not to gamble with the professor. ”
"If you are angry, that price is too big.
Instead, I found it was a can do. Computer game research is on your computer can handle the constant challenge to the limit. "He said. Now looking back the value of the "dark blue" is in the historical process of value, the more is to our research provides a lot of reflection on the path. "For example, now at Microsoft we do is hardware acceleration, consider these hardware devices to get WebSearch speed boost. ' Dark blue ' is a software and hardware combination, although with some differences of research now, but the idea is that you can use that time to the deep blue of the development process.
Back then, get yourself another tutor support, xufeng male intends to make with 6 months they require two years to complete the chip.
For him, this is an unprecedented journey. He every day from morning to night sitting in front of the computer. The first month, he put the 36000 Crystal a long 6.8 cm width 6.7 cm chip, each line has to be redesigned and spent 4 months, he put those line is a line art, there are tens of thousands and tens of thousands of lines, with the majority rely on hand-finished. Weekly working hours exceed 70 hours.
The first six months of the beginning, he put the 36000 Crystal with all circuits are all installed on the chip, then a line is not added.
Then start testing, the place where the error occurred. If the working time before is still relatively long, finally this month, he needs to get up every day at midnight. "After the June deadline", his chips really born, not only to work properly, but also prove that 64 chips system also 10 times faster.>
He gave his first success is called "Crystal test".
Since then, xufeng male madly in his chips, enthusiasm, seems to be endless. "I want to be a faster machine. "Four years after the birth of the" food for thought ", it has the 200 block of chip and two processors per second can of 70 000 points. Half a year later, the "food for thought" beat the Grandmaster Bengt Larsson ·, fame. 1989 xufeng male with two friends into IBM, began the manufacture of "deep blue". Still later, as we all know.
Someone once asked xufeng male, if you talk to a game of "deep blue", who would win? "he said," I can open the rear door, he went to see it on how operations.
"This can be regarded as a humorous answer, also answered a lot of people for artificial intelligence to replace human intelligence. "This is actually very simple, when we remove the deep blue socket that there is no way. Now, similar to the ' dark blue ' this is artificial intelligence, more, and special wisdom subhuman universal wisdom. One day the machine have the universal wisdom, may only occur for domination of the wisdom of the day. But it seems now that computers have not the capacity. ”
It is said that he now has to fuck "computer game" and "old industry" — this is the game of go.
"Mobile computing" big future
Xufeng male is now responsible for Microsoft's platform component group, devoted to the hardware-related research.
"Some of our hardware for Microsoft Project, but most of the projects for mobile applications and services. The core software algorithms, especially the service-side hardware acceleration software algorithm is the main goal of our current. Our areas of interest include: FPGA-based hardware platform, and accelerate some software algorithm or the system. With the help of this platform, we keep the original system performance (such as speed, capacity and power consumption, etc.) to get the order of magnitude improvement. At present we have successfully accelerate some machine learning algorithms, and is the platform to new areas, such as data mining, Web services, mobile, etc. Currently our primary mission is to accelerate at Microsoft use of various types of machine learning and data mining algorithms. "He said.
"According to Moore's law, we consider mobile computing in the future will become the primary form of calculation, so previous studies focused on this area.
In terms of software, we focus on system software, as well as software for special equipment. In terms of hardware, is exploring how to implement wireless power supply system. Game peripherals and related software, the use of camera and inertial sensors, we are looking more natural flow of game control. They also will be extended to situations outside the control of the game. ”
Xufeng says that his work is always a difficult to break.
In fact, everyone who has unlimited potential. Most people never perfect ability to express themselves, because they never wanted to know exactly what you want, never had a trying to grasp something. There are very few people will realize that they really want, feel it is in the front of the summon, desperate to grab it. But precisely because xufeng-the strong calls for his hard work, playing a talent that inspires him, beyond all obstacles.
No comments:
Post a Comment