Time Travel: 2014

…

On the evening of June 30, Lin Hui also went to meet Eve Carly.

How should I put it, there has been some progress in the follow-up work that Eve Carly has been engaged in, but the progress is not as great as imagined.

At least it was different from what Lin Hui expected.

Of course, Lin Hui knew this and couldn't rush it.

Generally speaking, the time required to study an efficient summary application with generative text summarization as the core is often restricted by many factors:

The first is the technical level:

The developer's technical level and experience will affect the progress and quality of project development.

If the developer has rich experience in natural language processing and machine learning,

Then the time to develop an efficient generative text summarization application may be shorter.

Next is the needs analysis:

Requirements analysis must be clear. Accurate analysis of user needs and clear goals for summary applications will have an impact on development time and resource investment.

If the requirements analysis is not clear, it may take more time to refactor and modify the code.

Secondly, it involves the selection of data sets and algorithms:

Building generative text summarization applications requires choosing appropriate data sets and algorithms.

The data set and algorithm will determine the quality and efficiency of the summary.

Choosing the right data sets and algorithms will help avoid detours.

Once the data set is large or the algorithm is complex, model training and optimization are required, and the development time may be longer.

Especially algorithms related to machine learning.

Usually the performance is that there is an algorithm inside the algorithm.

In other words, a complex algorithm can be decomposed into multiple simpler algorithms.

These simple algorithms can be combined together to complete more complex computational tasks.

This is not just the case with natural language processing.

As far as Lin Hui knows, many machine learning algorithms will have similar situations.

For example, in the field of computer vision, image processing algorithms usually consist of multiple sub-algorithms, each sub-algorithm is responsible for different tasks, such as edge detection, feature extraction, target detection, etc. These sub-algorithms can be combined together to form a complete image processing algorithm.

Another example is in the field of machine learning, where it is usually necessary to decompose the original data set into multiple sub-data sets, and perform feature extraction and model training on each sub-data set. The algorithms used in these processes can also be viewed as sub-algorithms, which are combined to form a complete machine learning model.

There are many similar examples.

It can be said that this phenomenon is very common in large computing tasks:

——Complex algorithms are decomposed into multiple sub-algorithms, which can be implemented using different computing technologies and algorithms to collaboratively complete more complex tasks.

The main reason for adopting this split combination is to improve calculation efficiency, simplify the development process, and make the maintenance and expansion of the algorithm easier.

In addition to data sets and algorithms,

Furthermore, it also involves resource investment:

Building an efficient generative text summarization application requires investing sufficient human, material and financial resources.

This may include hiring developers, purchasing hardware and software equipment, etc.

Overall, the time required to build an efficient summarization application built around generative text summarization can range from weeks to months, depending on the factors mentioned above and other factors that may affect development time.

In fact, if there were not all the conveniences brought by rebirth,

Lin Hui also encountered a lot of trouble when developing Nanfeng APP.

Fortunately, all the conveniences brought by rebirth,

Only Lin Hui has a lot of room for display.

…

Speaking of the past life, from 2010 to 2015, it was actually a period of rapid development of artificial intelligence.

During this period, many machine learning algorithms and technologies were widely used and promoted during this period.

For example, deep learning, random forest, support vector machine, etc.

However, the progress made in these areas still has many shortcomings compared with more mature machine learning technology after 2020.

If a technician who is very good at machine learning takes you back to 2015 from 2020 with all your memories,

In fact, it is easy to make a difference in many aspects.

For example, in terms of machine learning applications:

In 2015, the application fields of machine learning were still relatively limited, and most applications were concentrated in areas such as speech recognition and recommendation systems.

Then a machine learning technician from the next few years can undoubtedly consider exploring more machine learning application scenarios.

For example, new fields such as intelligent customer service, financial risk control, and medical diagnosis will contribute to the development of these fields.

For another example, in terms of deep learning algorithms, deep learning algorithms had just begun to receive attention and application before 2015, and they were only used in a small number of fields.

If a future generation of technicians comes before 2015 with all their memories,

There is no doubt that he can promote and optimize deep learning algorithms during this period and apply them to a wider range of fields.

Including but not limited to natural language processing and recommendation systems, etc.

In addition, there are many aspects.

But the problem now is that the above are all things that can be done before 2015 under normal circumstances.

But the problem is that I don’t know what happened in this time and space.

Machine learning and even artificial intelligence research as a whole are facing a certain lag.

Even the most cutting-edge machine learning research results in country M still lag behind compared to the same period in the previous life.

Generally speaking, this lag is friendly to born-again people;

This means that the information advantage of the reborn after rebirth is further amplified.

But this is also bad,

If the basic content of a space-time study has been laid,

Then many results can be achieved naturally.

But what now?

Because many underlying technical achievements lag behind, if you want to launch mature application-level achievements, you often have to develop those backward underlying technologies.

To be honest, these jobs are just "moving bricks", boring and boring.

But it happens to be indispensable.

Let's put it this way, in this time and space, even if the big guys at the picture award level come, they have to "move bricks" honestly.

Start by completing the underlying technical achievements.

After all, there are no castles in the air in academic research.

Academically, we can only make steady progress.

…

After meeting Eve Carly,

When Lin Hui stepped into Mingpin International's study room again, it was already 9pm on June 30.

At this time, Lin Hui discovered that his overseas account had an additional income of US$15 million.

Lin Hui knew that this was the reward for completing the adaptation and function upgrade of the shortcut command software.

Pingchun's payment speed exceeded Lin Hui's expectations. It seems that Pingchun is very satisfied with the adaptation and upgrade of shortcut commands and function upgrades.

Speaking of which, after receiving the transfer of US$15 million, the funds in Lin Hui's overseas account had reached US$113.9 million (mainly from the income from the sales of "HILL CLIMB RACING" and shortcut command software)

There is still US$110 million in domestic accounts (mainly from AppStore revenue sharing)

In terms of basic attributes, there is actually no difference between the two amounts of money.

That is, there is no essential difference between the US dollars in domestic bank accounts and the US dollars in overseas bank accounts.

It's all dollars stored in bank accounts as currency.

But why distinguish between the two?

Different countries have different financial regulations and regulatory requirements, which will lead to some differences and restrictions.

Generally speaking, U.S. dollars in domestic bank accounts are stored in mainland my country.

This money is more regulated and restricted.

If Lin Hui has considered transferring the US$110 million in domestic accounts overseas.

You need to apply to the foreign exchange management department for fund transfer or use. In some cases, you may be subject to quotas or review restrictions.

The U.S. dollars in overseas bank accounts are stored overseas, so it is relatively convenient to invest overseas.

But if this money wants to flow into the country, it will be equally troublesome.

In fact, the cross-border transfer of large amounts of funds is a complex task, whether it is transferring domestically or abroad.

There are many factors to consider and ensure safety and legality.

Many times it is also necessary to hire professionals, such as foreign exchange consultants, lawyers or financial advisors, to assist with this task.

Lin Hui has always been a person who is more afraid of trouble.

There is no need to cause a lot of trouble for nothing.

Therefore, it was a simple decision, as there is currently no major direction that requires all funds to be concentrated in one place.

Simply spend the money in the domestic account only at home, and the money in the overseas account is only used for overseas deployment.

Speaking of which, Lin Hui's domestic and overseas accounts now total about 224 million US dollars. In addition, Lin Hui's domestic account also has 99.5 million yuan in cash.

These alone are close to 1.5 billion yuan.

This does not include the remaining assets including the upcoming BTC.

With these assets, although there is still a big gap between Lin Hui and his goals.

But it can also be said to be an advanced model of financial freedom.

It’s pretty good to be able to do this in less than a month.

At least it didn't embarrass the born-again people.