To be fair, among the tasks that Gu Play arranged for Ma Yiyi, the "distributed compilation architecture" is the easiest breakthrough point to figure out.

The knowledge structure required for this thing is very suitable for gifted students majoring in computer science.

On the other hand, the earthlings' attention to distributed compilation back then was actually related to the approaching bottleneck of Moore's Law - on the earth, Moore's law that "the CPU computing power per unit area can be doubled every 18 months" was actually gradually stuck in a bottleneck as early as around 2005~2006.

At that time, the latest CPU released by Intel in the summer of 05, the unit computing power could not reach the doubling of the early version of 04, only about 1.6~1.7 times that of the spring version of 04.

By the time the Intel single-core CPU version was launched at the end of 06/early 07, about 18 months had passed since the summer 05 version, but this time, the unit computing power was only 1.5 times that of the summer 05 version.

It can be seen that the speed of multiplication of computing power is gradually slowing down. At that time, the processing technology of the CPU was still about 40 nanometers, and the bottleneck was already highlighted. It's not that Intel likes to squeeze toothpaste, it's that it really can't squeeze faster.

But the average consumer on the planet, in 05/06, did not feel a significant slowdown in the rate of improvement in computer performance - this is because Intel knew in advance that it could not squeeze toothpaste, so it launched a technology called "dual-core" at that time.

If the CPU unit computing power is not increased enough, then engage in a new computing power distribution architecture and use the number of cores to make up. After another two years, around 08, the dual-core couldn't hold on, so it went to the quad-core.

As for the later distributed compilation architecture/cloud computing, one of the biggest external incentives is the idea of "adding CPU cores", which has gradually become less cost-effective.

It is impossible to stack 32/128/512 cores indefinitely, in this case, it is better to directly engage in distribution and solve the underlying architecture problem of fully dismantling computing tasks in parallel. From now on, let's not worry about the computing power of a single CPU, and go directly to the scale of the Xinhai tactical stack.

On the Blue Hole, basic scientific research lags behind the Earth, and business model innovation is ahead of the Earth.

As for the chip industry, although this thing is subject to basic scientific research, it is also an industry that can make money to support research and development, so the two offset each other, and it is almost in sync with the development on the earth.

Therefore, on the Blue Hole Star in 99, it will take another three or four years to feel the crisis of "Moore's Law is about to fail".

In this case, the Blue Hole Star people currently pay very little attention to "distributed compilation" and other computing power distribution architectures.

Even if you really study this thing and find that it's not that difficult, but no one goes to this heart.

Everyone thinks that their efforts will become less meaningful because of the strength of Intel, so they don't bother to think about it.

A beautiful leak is just placed there, waiting for a person who has the level of a graduate student majoring in computer science at the alliance school to pick it up if he wants to.

……

In October 99, when the first semester of her sophomore year had just begun, Ma Yiyi at that point in time was, of course, not at the level of a "graduate student of the computer department of the alliance school".

However, under the targeted hard study, under the guidance of Gu Wan's training, he made a desperate effort to dig, and at the end of the period, Ma Yiyi almost figured out this problem.

In the final analysis, it is still a rare opportunity to pick up leaks, and it is a godsend.

In December, she successfully tinkered with the architecture in her school laboratory and ran it for half a month, confirming that the compilation efficiency was improved and the data collection was sufficient, so she began to write a paper.

This paper, Gu Yu didn't plan to put it on the name, so he let his wife write it alone. After all, there are not many follow-up projects that Ma Yiyi can complete independently, and it will only get more and more difficult. Let her debut work be completed independently, and it will be good for her reputation for the rest of her life.

Otherwise, he relies on his husband throughout the whole process, and he will not be said to be hard rice and soft food.

In addition, Gu Play is not a person who likes to send his sister's results. He is very measured, even if it is his legal wife, he will not give results unprincipledly.

First of all, we must ensure that the other party's ability is really done, and in the future, others can talk about it in the first place, so that the paper can be transformed into academic prestige. If you can't control your academic prestige, it's counterproductive.

In addition, Gu Play took into account that he would not personally run a business in his life.

He is a pure science student, and he can only be a great scientist, an absolute controlling owner, and a spiritual leader.

If his wife can help him become the CEO and manage the transformation of results into money, it would be the best. Therefore, he also needs to let Ma Yiyi gradually establish the prestige of controlling other scientific researchers in the future.

Otherwise, those scientists think that it is not good to play cronyism, listen to women, and detach from morality from now on.

Therefore, there is no need to worry that Gu Play is not an "absolutely refined egoist". He is absolutely self-interested on the way to gain faith for science, and the benefits given to tool people will be useful sooner or later.

The paper of "distributed compilation architecture", writing, revising, and running-in, two months before and after, one rejection, and one minor revision. When it came to the winter vacation, I finally relied on Ma Yiyi's own strength to pass the draft.

On the day she got the reply letter, Ma Yiyi was very excited and showed it to her husband as soon as possible.

"It's an IEEE journal, and Transaction-on-Computers is finally a top international journal. Our dean even praised me publicly throughout the school. ”

Gu Yu is used to it, and he doesn't praise him too much, who made him a straight man who just wrote:

"IEEE's journals? Yes, your article this time is still biased towards architecture application, not pure theory and pure algorithms, and it is good to be on the top journals of IEEE. Next time, try to be the second author with me, the last ACM journal. ”

IEEE stands for Institute of Electrical and Electronics Engineers, and ACM stands for International Association for Computing Machinery. Each of these societies has the world's top journals in the field of computing/algorithms.

However, the ACM shelf is relatively high, and people are very biased towards algorithms/ideas. The most famous journal under his name, JACM (J here is Journal, transliterated as ACM journal), almost refused to publish any papers with application technology, feeling that the commercial application was too lucrative and unqualified.

In a word, it is "theoretical discrimination" in the academic world.

In fact, not only in the computer world, but also in the mathematics and physics circles, those who engage in theoretical argumentation and conjecture look down on those who do experiments, and those who do experiments look down on those who do applications and the integration of production, education and research.

Therefore, on the whole, Ma Yiyi's "TOC" voted this time, strictly speaking, can only barely rank among the top 5 academic journals in the field of computer science. There are three or four purely theoretical, incomparably high-minded fireworks periodicals in front of me.

But it doesn't matter, who made the "distributed compilation architecture" actually have application elements.

Next time, when Gu Wan and Ma Yiyi follow this structure, they will take out the pure algorithm paper of cloud computing power, then they will not dare to use the Qinggao journal that does not eat the fireworks of the world.

The IEEE is just a tool association that can be thrown away when the springboard runs out.

Although this springboard is so valuable in the eyes of other master's and doctoral seniors in the Academy of Mathematical Sciences. Any Ph.D. student majoring in computer science at the Central University of Science and Technology can graduate directly as long as he sends a "TOC" under IEEE, and the tutor wants to enslave you as a cheap labor force for a few more years, and there is no excuse to stay.

People are more popular than dead people.

Ma Yiyi also expected that her husband had seen a lot of wind and waves, so she wouldn't be too surprised, and she wouldn't praise her, so she got used to it. She just asked, "So what are we going to do next?" Can we get to the topic of machine learning training that you mentioned?"

Gu Play: "This really has to be done, but you don't have the strength alone, so it's better to be the second author for me." I've been preparing for this semester. As for you, I have a thesis in mind for the next semester, which can be used as your graduation thesis for the 'cognitive neuroscience' major under your double major in psychology, so you can write it if you have time. ”

Ma Yiyi was curious: "What is that?"

Gu Play: "This paper is very important for artificial intelligence in the future, but its verification does not need to be verified with the help of artificial intelligence, even if it is only based on human physical experiments."

However, the design of the control group is very complicated, because every living person is different, and it is difficult to standardize people into machines, and then eliminate all interference to measure the learning speed/learning effect of living people.

However, as long as it is made, it can be widely accepted by the public because of the fun and easy to understand of the paper. I believe that with the long-term international influence of this paper, you may have an additional title of 'psychologist' or 'cognitive neurologist' in the future. ”

Ma Yiyi couldn't help but be a little excited, and hurriedly asked the details: "Can you be specific, what do you want to study?"

Gu Play: "What I want you to study is that when human beings learn, what is the ratio of what they understand and what they don't understand, what they do right and what they do wrong, what is the learning efficiency is the highest, and the learning speed is the fastest."

If this data is studied, and it is the kind of research that directly hits the bottom layer of the human brain's operating mechanism, then the conclusions can also be applied to artificial intelligence in the future. This is a universal variable that allows machine learning and human learning to be connected.

I can give you a little advice on the specific test method. You design it like this......"

Gu Play shamelessly plagiarized the experimental methods that later generations of deep learning artificial intelligence experts have explored for many years, and directly reported the answer.

Ma Yiyi heard her pupils gradually dilated, and she zoomed in and out a few more times, and her heart was full of shock.

Hell, is this something that the human brain can come up with? It's so fanciful, isn't it?

——

The above is already more than 3,000 words, so the following is not counting the money and a few words.

I may be a little sorry for everyone, but I still have to speed up the completion of this book.

But I try to ensure the integrity of the plot, so in the follow-up, you will often see one or two chapters that pass the progress speed of a semester in the book.

The more I write about the subject of "The God of Science and Technology", the more I find that if readers don't like to read academic results and pretend to be coercive, and feel that these contents are hypocritical, it will be difficult to write.

Because I'm writing about serious scientific research, not the kind of system stream that claims to be science fiction but is actually fantasy. This is destined to land the results of every major scientific research project, and it will take several years to land. Therefore, if you don't write about life and academic papers, you won't have anything to write in between.

Serious scientific research, especially scientific research on large-scale systematic projects, is not suitable for short-term and quick pretense at all. To be fast-paced, you can only skip the timeline, do a good job of experimental design, and after a few months of research and development, you can skip it directly, and then produce results and be admired.

I thought about the new book, but it will still be released in November, so I can't even write this book at the end of November, maybe it will be mid-November. I have about the speed of three or four chapters a year in the book, and I can write out a few big scientific research results, and then I will write a little bit of a ledger.

Because if the book is released again in early December, one of the biggest problems is that when it is put on the shelves, it will be in time for the Chinese New Year or the final exam of students, so the first order of the book will definitely be a nightmare.

If you can catch up with the shelves around the New Year's Day, the next book may have a better fate. In order to avoid the Chinese New Year and the final exams, I can only open a new book in mid-November.

I'm sorry for the friends who like this book, everyone don't give any more rewards, I deserve it.

()

Sogou