Time Travel: 2014

…

Due to the time difference, while Lin Hui was fast asleep, people on the other side of the world were still busy at work.

California time is 16 hours behind BJ time, and it was just 8 a.m. on the 21st.

There is also a natural language processing algorithm team currently busy at the NOTEXIST Research Laboratory on the campus of Stanford University (located in California).

Seeing the busy researchers in this NOTEXIST laboratory, many computer-related students passing through Stanford University will unconsciously slow down their steps.

While walking slowly, these people will also feel a little envious of the people in the research room.

Because they know that although these researchers are doing research in the research room of Stanford University.

But in fact, most of the members of the NOTEXIST research laboratory do not belong to Stanford University, but to the famous Google/Google Research Institute.

The Google/Google Research Institute can be said to be the dream workplace for all computer-related students at Stanford who are committed to employment.

Although the salary here can only be said to be above average, the key is that it is cool.

How many world-changing technologies were born from Google/Google Research Institute?

And which student has never dreamed of changing the world?

In short, most of the looks in the eyes of these passing students were envy and a trace of jealousy.

Dr. Eclair Kilcarga is a member of the NOTEXIST research laboratory. He does not think his current life is enviable.

With an annual salary of $325,000, he has to live like an ant.

Although it was only eight o'clock in the morning, in fact he and his friends had been busy almost all night.

Eclair Kilcaja hates this busyness!

But I have to be busy!

If their text processing algorithm development team does not keep up with the latest generative text summarization related technologies.

Then they are very likely to follow in the footsteps of the MIT Natural Language Processing Text Summarization Group:

--Toward falling apart.

Although they are unlikely to be dismissed directly as part of Google Research.

But it is very likely that he will be sent to Asan's research institute.

I heard that the Google Research Institute over there has just been established and is now very short of manpower.

Eclair Kilcaja did not want to go to that dirty and chaotic land.

Thinking of the possible tragic ending, Eclair Kilcaja hated the guy named LIN HUI.

For such an algorithm genius, wouldn’t he be able to achieve success in any algorithm he uses?

Why do we have to develop some kind of generative text summarization algorithm to take away their jobs?

Eclair Kilcaja was complaining when she suddenly heard fellow team member Harley Price shouting:

"Hey, Eclair Kilcaja, don't be dazed. What are your test results over there?"

Eclair Kilcaja: "Don't mention it, it's simply terrible. The X1 algorithm we use is inferior to the algorithm used by Lin in Nanfeng APP in all aspects of parameters..."

The so-called X1 algorithm is a generative summary experimental algorithm developed by Eclair Kilcaja and his team, inspired by the technical route disclosed in the patent applied for by LIN HUI.

It is called an experimental algorithm because this algorithm is mainly used for testing.

As a test algorithm, the X1 algorithm is qualified.

Eclair Kilcaja and his colleagues have made this algorithm a bit like a generative summary algorithm.

However, this algorithm can only be used as a test algorithm now.

It is still one step away from real practical application, but even though it seems to be only one step.

This is a step away, but it can be said to be a world of difference.

Objectively speaking, the X1 algorithm is far behind the algorithm in LIN HUI Nanfeng APP in terms of text summary extraction speed and accuracy.

When it comes to accuracy, Eclair Kilcaja is depressed.

Even the standards for measuring accuracy are provided by LIN HUI. How can you compete with others?

It's over before it even begins.

And the most depressing thing is that the X1 algorithm they made is not comparable to the algorithm made by LIN HUI.

Even compared with the algorithms they tinkered with before, the extraction speed and accuracy were also greatly inferior.

This situation cannot help but make Eclair Kilcaja a little pessimistic.

Harley Price: "Tell me about the numerical difference."

Eclair Kilcaja: “I have a feeling you don’t want to know, the disparity is desperate.

The results are printed out, take it and see for yourself..."

Harley Price took the report handed over by Eclair Kilcarga, read it for a while and then frowned:

"I can understand the difference in recognition accuracy. After all, the algorithm developed by LIN HUI must have been trained for a long time to obtain the results.

But why do we make the recognition speed of this X1 algorithm much slower than the algorithm in Nanfeng APP?

Normally, when two algorithms with the same theory deal with the same problem, shouldn't the algorithm time complexity be the same? "

Eclair Kilcarga shrugged: "Who knows? Maybe there is something wrong with the mechanism we use..."

Harley Price wondered: "How is it possible? Isn't the technical route explained by LIN HUI based on the Sequence-to-Sequence deep neural network model? If there is really a problem with his technical route, then that Things have become easier to handle, and we can sue his patent for invalidity based on this?"

Eclair Kilcarga: "Don't be naive! LIN HUI only said that the model is based on the Sequence-to-Sequence model, and it did not say that he is still applying an early version of this model in the algorithm."

Harley Price: "These damn strangers, they're despicable..."

Eclair Kilcarga: "Stop complaining. If we encountered the same problem, the methods we would use might be even more excessive. And the focus now is to improve the accuracy of algorithm processing. As for speed, it doesn't really matter... At worst, we can run this algorithm on a distributed system."

In computer science, distributed computing is also translated as distributed computing.

A distributed system is a system formed by a group of computers that are connected to each other through a network to transmit messages and communicate and coordinate their actions.

Harley Price knows that Google has made great progress in distributed computing in recent years because of the Google Brain Project.

If this algorithm is placed on a distributed system, it can indeed achieve a breakthrough in execution speed.

Harley Price was not happy at all after hearing the news.

Once they really choose to run the algorithm on a distributed system, it means they will lose directly.

You must know that the algorithm in Nanfeng APP runs offline and does not even have a server.

What is the difference between using distributed algorithms directly and killing chickens with h-bombs?