Top of the big era

Zhou Buqi previously communicated with Zhou Shaoning that Asda's current market valuation can reach US$35 billion, while Nokia's current market value is only US$32.7 billion.

If you do not count companies such as Microsoft and Samsung that do not focus on mobile phone manufacturing, then Asda is currently the second largest mobile phone manufacturer in the world after Apple.

This is very particular.

Asda has lost its reference object. As long as its valuation does not exceed Apple, Microsoft, and Samsung, how much difference can it make if the valuation is US$35 billion, US$45 billion, or US$55 billion?

If Asda's valuation is high, then in the Motorola mobile phone business deal, Google can get "a lot of money" and feel proud.

Google's stock price soared, and Larry Page was in high spirits.

At the same time, Asda will also benefit.

Even Google recognizes Asda’s ultra-high valuation. Why don’t other venture capital institutions recognize it? So when Asda launches a new round of financing, it can use Google as a springboard to make a higher offer.

This allows Asda to raise more money while diluting the same number of shares.

It is definitely a mutually beneficial and win-win transaction.

Blowing bubbles is indeed a powerful tool in the capital market!

The other party's proposal is really worth considering.

Next, Larry Page mentioned the choice of the technical direction of cloud computing. This is confidential. Zhou Buqi cannot just talk about it casually and just pretend that he is a technical layman and knows nothing.

Larry Page then suggested, "I have a product in my hand that is not doing very well, but I think this direction is in the right direction and is consistent with many of Ziweixing's products. You can take it away if you need it; If not, I will close the business within a few months."

"What product?"

Zhou Buqi would not listen to what he said so nicely. Google is a listed company, and the company's assets cannot be given away casually. If Ziweixing wants this product, it must spend money to buy it.

Like unsaleable milk.

want to?

You have to spend money to buy it.

Not buying?

I poured it directly into the smelly ditch.

Larry Page said: "Knol, you know? A product designed to compete with Wikipedia. It was developed in 2007 and launched in 2008. However, after several years of use, the results were not very good. Google's core strategy now is to do Social networking and some products that are distracting and perform poorly will be shut down one after another.”

"A competitor to Wikipedia?" Zhou Buqi was really interested, "Tell me more about it."

In China, Ziweixing has Weidian Encyclopedia, which provides services similar to Wikipedia.

There is no such product in overseas markets.

Wikipedia is so powerful.

Fortunately, Zhou Buqi approved a deal two years ago. Ziweixing International acquired Quora, a content product platform, which is the legendary "American version of Zhihu".

Producing content has always been a direction that Ziweixing is striving to develop in domestic and foreign markets.

In the future Internet market, content will ultimately be king.

Whoever has the content has hegemony.

This includes future artificial intelligence large language models.

For AI products such as ChatGPT and Wen Xinyiyan, if they want to achieve amazing intelligent performance, they must be trained with countless data contents.

However, data cannot be used indiscriminately. Only public data or open source data can be freely used by developers around the world.

There is a very large free database around the world.

There are all languages ​​available, more than 90% of them are English corpora, and the Chinese corpus data is only about 2%.

Almost all AI models in the world rely on these public data contents for training. Because the data is all in English, these AI models must be centered on English.

So when Baidu's Wen Xin Yi Yan is launched, many things that are difficult to understand will appear... In fact, the reason is very simple. Wen Xin Yi Yan uses English data, and there is too little Chinese corpus data.

Despite the large number of people in China, there is not much truly valuable content on the Internet. If it goes a little too far, these valuable corpus will be deleted according to laws and regulations.

What's left is a bunch of junk content that can't be used to train AI.

For example, what kind of book is "The Peak of the Great Era"?

If someone says it's good and someone says it's not, these contents are all valuable content. After some training and evaluation, the AI ​​model can give a more objective and fair evaluation.

If the author wants to maintain harmony in the comment area and leaves all the good content and deletes all the bad content, leaving only a chorus of praises, then even if these praises are correct, this is still spam.

Because there is a lack of diversified judgment for AI models.

There are 10,000 roads from the starting point to the end point. The training of the AI ​​model is to walk all these 10,000 roads and then select the most suitable road. This is the value of AI.

Just like biopharmaceuticals, there are 10,000 choices. AI can help give you the best choices, which will greatly reduce R\u0026D expenses and improve R\u0026D success rates.

If you only provide one path from the beginning, it will be a waste of training. If you don’t give the AI ​​a chance to choose, judge, think and analyze, and just tell it a standard answer, the AI ​​will be meaningless.

The amount of data in German, French, Japanese, and Korean is too small, and the amount of data in Chinese is large. However, due to environmental limitations, there is also very little valuable data.

Therefore, if you want to train a world-class AI model, you can only use English corpus for training.

This requires some AI models for weak languages, which must have language conversion capabilities to convert them into English.

Like using Wen Xinyiyan and asking him to draw a picture of a crane.

The result was a crane.

This is very confusing.

In fact, it is very simple. The English word for crane is crane, and crane in English mainly refers to crane. So the AI ​​model drew a picture of a crane.

Another example is "Cola Chicken Wings". In the Chinese context, it is actually easy to understand that it is just a dish. However, there is a translation procedure to translate "Cola Chicken Wings" into English. During the translation process, the information is separated, resulting in the drawing of Coke and chicken wings.

ChatGPT will also encounter this dilemma.

If you ask questions to ChatGPT in English, the response speed will be very fast and the accuracy will be extremely high; if you ask questions in other languages ​​​​such as Chinese, Japanese, Korean, French, German, etc., the response speed will be very slow and the answers given will also be full of mistakes.

Because the amount of data in other languages ​​is too small, the trained models are not smart enough.

Only the English version is the smartest.

If Wen Xinyiyan wants to express it more intelligently, it cannot connect to Chinese corpus. It has to connect to English training data, and then add a set of translation algorithms and content review algorithms behind it.

Therefore, if you want to make a great AI product, in addition to the top algorithms in the field of artificial intelligence, there is also a very important factor, which is the data for training the AI ​​model.

Some of these data are open source public data and can be used all over the world. Some of it is private data that only you can use.

Silicon Valley giants like Microsoft, Google, and Facebook can use not only huge public data, but also huge and valuable English private data. The AI ​​models they produce will be better than those trained only on public data. AI models need to be more powerful.

Therefore, Zhou Buqi is very important for content, especially overseas English content.

These are all private data and may become Ziweixing’s most valuable asset in future competition in the world.

If the Knol mentioned by Larry Page is really a product that benchmarks Wikipedia, it would be very meaningful. Wikipedia is public data, and Knol is private data.

"Wikipedia is more like a dictionary, the content is relatively basic, and it is a product for the general public. Knol is a group of capable and authoritative users who work together to explain what a thing is, how to do it, the causes, consequences, and Its surroundings, and indicate the author’s real information, so as to make the content more accurate and authoritative.”

Larry Page made a simple comparison.

Zhou Buqi immediately understood the reason for the failure of the Knol product.

The accuracy and authenticity of information are actually not important at all. The public does not care about the truth at all, only the appearance. This will form a phenomenon of "the more popular, the more popular" and "the more authoritative, the more niche".

Google wants to build Knol into an authoritative content product. It has high intentions and great ambitions, but it is still difficult to gain enough users.

Just like Ziweixing’s two domestic products, Weizhi and Toutiao.

The content on Weizhi is much better than Toutiao.

However, Weizhi’s daily traffic is only over 65 million, while Toutiao’s daily traffic is as high as 280 million. It’s because the content of Weizhi is so good, and the headline content is very sensational and attracts the public’s attention.

The two had a brief exchange. Zhou Buqi became more and more interested in the Knol product and planned to go back and study it first. If it's suitable, you might as well buy it and modify it.

At the end of the meeting, the two talked about immortality.

Larry Page is very interested in this. "After the Twitter case is finalized, the next step is to get the board of directors' approval to try to do something in this field."

Zhou Buqi smiled half-heartedly, "How much money do you need to invest?"

Larry Page said loudly, "U.S.$2 billion. The funds are too small and there is no need for the board of directors to approve it. I plan to establish a subsidiary specializing in biological research for anti-aging related research, with the goal of realizing human Long life and even eternal life.”

Zhou Buqi asked: "Integrate it with IT?"

Larry Page said frankly: "Of course, it must be related to artificial intelligence. Now every time humans study the efficacy of a protein, it may take dozens of medical experts spending tens of millions of dollars to research it. If AI develops If we get up, we can decipher the functions of all known proteins in mankind at once."

Zhou Buqi gave a thumbs up and said, "There is a pursuit."

"You will also participate, right? AI is changing the world."

"certainly."