Time Travel: 2014

Understand the algorithm and you can exchange money.

In fact, it is not difficult to understand why data can be exchanged for money.

After all, data is the cornerstone on which many machine learning algorithms are built.

The emergence of machine learning algorithms often relies on labeled data.

And for a long time, machine learning algorithms have not only relied on labeled data.

And it relies on a large amount of labeled data.

When the amount of labeled data is small, it is often not enough to train a machine learning algorithm with excellent performance.

From this perspective, it is not difficult to understand why data can be exchanged for money.

In many cases, data can even be understood as a kind of tacit knowledge.

The process of data annotation is actually the process of structuring and labeling scattered discrete data.

Beyond algorithms and data, what is the so-called narrow knowledge?

Narrow knowledge generally refers to explicit knowledge defined manually through rules or dictionaries.

Narrow sense knowledge mainly includes three types:

——That is, language knowledge, common sense knowledge and world knowledge.

Among them, language knowledge refers to the definition or description of the morphology, syntax or semantics of language.

Its main feature is the definition of synonym sets. Each synonym set consists of words that have the same meaning.

Common sense knowledge refers to the basic knowledge that people acquire based on common experience.

World knowledge includes entities, entity attributes, relationships between entities, etc.

Maybe someone doesn't understand?

Why can this kind of knowledge be exchanged for money?

Aren't these things obvious?

Although this knowledge is essentially explicit knowledge that people can understand.

But explicit knowledge that is obvious to people.

It is not equally obvious to the machine.

This type of knowledge often needs to be processed through regularization or lexiconization so that the knowledge can be easily understood by the machine.

This knowledge that is easily understood by machines is called narrow knowledge, also known as expert knowledge.

Although now the main training model is to seek algorithms or even the data itself.

But narrow knowledge is quite marketable.

Anyway, judging from the previous exchanges with Eve Carly.

It’s already 2014, and Silicon Valley still has to cooperate with universities like Harvard and Oxford to develop machine learning.

The reason why these people rely on Harvard and Oxford is not only to expect these universities to label data.

The main reason should be to count on the blessing of these universities in narrow sense knowledge.

It's easy to understand why these people do this.

After all, when it first came to model data in natural language processing, people used narrow knowledge for training instead of relying on data and algorithms.

In Lin Hui's impression, even in the previous life, before the rapid rise of the Internet, the only way people could train natural language processing models was to use narrow-sense knowledge.

Lin Hui possesses quite a lot of narrow-sense knowledge, and the level should be much higher than what is currently used in the Western world.

Judging from the tens of millions of dollars spent every year in Silicon Valley to acquire narrow knowledge.

If some of the narrow-sense knowledge in Lin Hui's hands could be monetized, it would be more convenient than using annotated data to monetize it.

But this is only theoretically easy to realize.

Lin Hui does not have the absolute right to speak and the authority that comes with it.

How can Lin Hui tell potential audiences that the narrow knowledge materials he possesses are superior to the general materials currently used in Silicon Valley?

In fact, the potential buyer Lin Hui knew it very well.

You must know that even in the next few years, there will not be many buyers who are interested in a large amount of narrow knowledge and are not short of money.

Lin Hui estimates that the buyers who may be interested in large-scale narrow knowledge bases in this time and space are none other than super giants such as Microsoft and Google.

But even if he knew about these potential buyers, Lin Hui would not be able to take the initiative to find them.

Wouldn't that become peddling?

Touting is equivalent to a direct loss of initiative.

After all, according to the buyer’s thinking logic:

Touting is equivalent to the seller not being confident in the product.

That is, the product lacks authority.

Lack of authority over some products is tantamount to a direct death sentence.

Why should buyers pay for content that lacks authority?

Maybe this is just Lin Hui's conjecture, but Lin Hui feels that this kind of thing is a high probability event.

It's not authoritative, even if it's something of a higher level.

It's also very stupid to use it to exchange money.

But with absolute authority, things are different.

Many times it becomes a seller’s market.

The kind that buyers come to ask for.

Just like the algorithm teams in natural language processing described by Eve Carley, they are constantly rising and falling.

But universities such as Harvard and Oxford never have to worry about not having algorithm teams to cooperate with them.

After all, to a certain extent, these top universities are almost equal to authority in narrow sense knowledge, especially in some language knowledge that is inseparable from NLP development.

In this case, don’t say that these colleges and universities don’t have to worry about food.

Even many algorithm teams have to look at other people’s faces.

Have authority.

In fact, it is easy to attract money at a level that is not that high in terms of narrow knowledge mechanization.

These universities do not even need to produce knowledge themselves.

Many times third world countries produce shirts, pants and other clothing, and developed countries directly use them and put a trademark on them.

Developed countries almost always do the least in this process, but they easily reap the largest profits.

However, labeling does not only exist in the field of clothing.

In terms of machine learning, in many cases a team with strong authority can directly "label" "common knowledge".

Yes, that’s right, knowledge labeling.

This is the real meaning of making money while lying down.

And it's the kind that lies in the atmosphere.

It's beautiful, but far away.

But there is no need for Lin Hui to be too discouraged.

Because it involves the pursuit of the right to speak, Lin Hui is not alone.

Walking with Lin Hui is not someone.

It is an ancient oriental country with a long history of five thousand years.

Lin Hui believes that through constant exploration, he will one day achieve the ultimate pursuit involving the right to speak.

The ideal is beautiful, but the road is tortuous.

Now that we don’t have the right to speak, it’s difficult to rely on narrow knowledge to exchange money.

Unless you are looking for some colleges and universities on the same level as Harvard and Oxford, and ask them to endorse Lin Hui.

But wouldn’t this mean making money by looking at other people’s faces?

Uh... the most important thing is that 70% of it still belongs to others.

Lin Hui would not do this kind of thing like making wedding clothes for others.

Before there is enough voice, it is impossible to rely on narrow knowledge to stay in the atmosphere.

It seems that it is better to make money honestly through data annotation or other realistic means.

Even data annotation can be a huge asset to Lin Hui if used properly.

At the end of the 18th century, during the westward expansion of the United States, gold sand was discovered in the Sacramento River.

Driven by courage and greed, workers, farmers, sailors and missionaries came one after another to pan for gold.

This is the famous "gold rush".

However, in this vigorous westward expansion movement, not many people actually made big money through gold mining.

On the contrary, water sellers who found other ways during the gold rush made a lot of money.

The field of "data annotation" is to some extent the "water seller" during the rapid rise of artificial intelligence in the previous life.