Reinventing the Millennium.

Chapter 1452 Weak (4k)

Yi Ke's "Taibai" was born and broke through the human Go camp at a lightning speed, but even in a 1V5 match, even among the top human players, many people were still dissatisfied with Yi Ke's claim that AI has surpassed human intelligence in the field of Go.

There are many reasonable reasons for this, for example, AI appeared too suddenly, and no one had time to carefully study its chess moves; for example, the time was too short, and the chess players had not fully adjusted to their best state...

However, it is too late to feel sorry for the human chess players now, as the one arriving on the battlefield is "AlphaGo" from Google.

Following AI's unrivaled victory over humans, Yike quickly determined the time for its technical competition with Google, directly setting it for July 7, and the entire process will still be live broadcast.

Fang Zhuo doesn't understand Go, but he is very interested in the verification and communication of this technology.

This time it was not just a game between two AIs. Yike also held an exchange event with Google's Deep Learning in Shanghai, and R&D personnel and startups in this field from NVIDIA and Silicon Valley also participated.

Yi Ke is really willing to compete with others, and Google and Silicon Valley don’t have much intention of competing with each other, because there is really not much hope for breakthroughs in this field, and it is still in the accumulation stage.

It was precisely based on this situation that many developers and companies who were not invited saw the interaction between the two companies and actively signed up for the event. After discussing with Google, Yike simply expanded the scale and postponed the time to July 7.

This is not an AI interaction in a Go game, but a discussion on deep learning (DL) with Go as an appetizer.

On July 7, Yike, Google, Nvidia and other companies held a meeting at the Yike Center in Shanghai. At the same time, Yike also invited previously defeated Ke Jie, Lee Sedol, Shin Jin-seo and others as guests to witness the duel between "Taibai" and "Alpha".

Both of these AIs have a track record of defeating top human chess players. The multi-faceted attack and public hype that Taibai created last month also successfully aroused the public's interest in AI. In addition, there was a live broadcast, so there were many viewers.

At nine o'clock in the morning, two robots sat in the center of the stage. A large screen was set up on the scene. Fang Zhuo, Larry and others sat in the first row. Chess players Ke Jie, Shin Jin-seo and other guests commented in the live broadcast room.

Different from the last 1V5 live broadcast, both "Taibai" and "Alpha" made their moves very quickly. There were occasional hesitations, but these moments were extremely short compared to humans.

The game followed traditional rules, with each player having three hours to play. However, in just thirty-two minutes, this much-anticipated game ended with Taibai winning by one point.

These 32 minutes are 32 minutes that most people cannot understand. However, they can see the performance of the chess players and the process in which there are disagreements in the live broadcast room at the beginning, disputes in the middle, and finally silence.

When Ke Jie was invited to the stage to comment on the game as a professional, he looked at the camera blankly. After a while, he said, "AI in Go may have completely exceeded human imagination. Last month, I seemed to see the shadows of Gu Li, Lee Chang-ho, and Wu Qingyuan in AI's chess. At the same time, I seemed to see my own shadow."

Ke Jie's expression was filled with struggle and pain. He put his hands over his face and said, "Today, I can't see anything. I can't see anything at all. I don't seem to understand what Go is like..."

Seeing that Ke Jie had lost his composure, the host quickly asked this top human chess player to step down from the stage and tried to smooth things over, but this scene undoubtedly left a deep impression on those watching the live broadcast.

The appetizer of AI game was over, and Fang Zhuo took the microphone and made a brief speech.

"Go is a masterpiece of human intelligence, but so is AI."

"My expectation for AI is that it can greatly free human hands. This day may be far away, but today's 'Taibai' has made progress compared to last month."

“AI will evolve at an amazing iterative speed, and we are gathered here today to find the right direction for development.”

"AI has defeated humans in the field of Go, but this is not the end of human wisdom. Instead, it is an extension of human wisdom, another advancement in science and technology, and another exploration of the future."

Fang Zhuo's expression of this view won a lot of applause and praise in the live broadcast room.

For many people, the excitement ends here, but for practitioners and R&D personnel, the real part has just begun. Both Yike and Google have deep research in the field of deep learning DL. This kind of Go game is just a superficial manifestation, and the internal operation and thinking are more important.

As one of the leaders of Yike's "Venus" project, Andrew Ng exchanged views with Silva of Google on the model logic of DL.

Whether it is Taibai or Alpha, they are all based on the development of convolutional neural networks. The foundation is similar, and its breakthrough originated from the paper on AlexNet deep convolutional neural network published by Alex, Ilya and Hinton in 2012. It was after that that related research experienced explosive growth.

Andrew Ng and Silva talked about the architectural innovation after AlexNet, which is the effective integration of traditional search algorithms and deep learning models, as well as the efforts made by the entire team in local receptive field, parameter sharing and sparse connections, and translation invariance.

This kind of communication between Yike, Google and the guests off the stage was extremely pleasant and made Fang Zhuo quite satisfied. Although he did not understand it, he felt that the knowledge was stuffed into his brain when he saw such a scene.

However, the next day, when Andrew Ng raised the team’s confusion in R&D, a heated debate ensued.

Yike has voice assistants like "Siri" as a practice of artificial intelligence, while Andrew Ng's team is not only conducting research on convolutional neural networks CNN, but also recurrent neural networks RNN. They believe that the latter is more suitable for combination with voice assistants, but the effect is not very good and does not achieve the desired results at all.

where is the problem?

Andrew Ng expressed his confusion and also talked about the solution direction within Yike.

Some of the participants agreed with Yi Ke’s solution, but different opinions emerged from Google.

"Why do we have to use recurrent neural networks?" Google's Uskert was on vacation, but signed up because he was interested in DL exchanges. "Why not try self-attention? I think it will bring better changes to the field of NLP."

"Self-attention can perform better parallel computing capabilities instead of sequential processing like RNN. It can also directly compare the vector representations of any two positions in the sequence, so that it can more effectively capture and utilize long-distance dependencies, but RNN cannot!"

“Although RNN can theoretically capture long-distance dependencies, it is often difficult to achieve in practice due to the vanishing or exploding gradient problem!”

Uskert studies ways to improve Google's machine translation. His father is a professor of computational linguistics. Although he disliked the language translation work when he first joined Google, he eventually focused on research in this field. What he is currently thinking about is the improvement of "self-attention" in related fields.

Andrew Ng quickly understood what the Google researcher meant, and after some thought, he refuted it: "Self-attention does not explicitly encode position information, which means that a model based on it cannot distinguish the difference in meaning of the same word at different positions in a sequence. In natural language processing, the semantics of words are closely related to their positions."

“Moreover, the self-attention model necessarily has a large number of parameters due to the attention weights calculated for each pair of elements in the sequence, which is likely to lead to overfitting.”

As soon as he spoke, Silva from Google's own DL also refuted the new approach proposed by Uskert. One of the important reasons is that the cyclic structure of RNN is too consistent with everyone's understanding of sequence data processing, that is, the current state depends on past information, and the global dependency of self-attention is not as intuitive as RNN.

Both Ike and Google's two leaders criticized self-attention, but Uskert was not convinced. He went on stage to explain more of his ideas.

Moreover, some solutions are given to the shortcomings criticized by Andrew Ng and Silva, such as introducing position encoding and conducting research on multi-head attention.

Some people found it eye-opening, some thought it was fantastic, and some conducted quick analysis and calculations on the spot.

Fang Zhuo in the first row was extremely confused. He turned his head and asked Nvidia CEO Huang Renxun who was thinking beside him: "What are they discussing?"

"Uskert said that GPU is the most suitable hardware for deep learning technology." Huang Renxun gave a summary.

Fang Zhuo: "???"

He wondered, "Why haven't I heard any similar statements?"

“Because Self-Attention emphasizes parallel processing, which is what GPU is better at.” Huang Renxun smiled, “As for the others, it’s not important. We just need to extract what is beneficial to us.”

Fang Zhuo observed the atmosphere at the scene. This was no longer an exchange of knowledge, but more like a clash of knowledge.

He silently put away his already limited presence in such occasions.

However, in the evening, Fang Zhuo still asked Andrew Ng, who was participating in the debate, in person, wanting to know what the heated discussion was about.

It was really difficult for Andrew Ng to explain to Mr. Fang what had happened.

"Mr. Fang, let's think about it before writing a report." He said sincerely after thinking for a long time, "Maybe we should have more exchanges like this. His self-attention, now I think, is a shortcoming that can be solved."

Fang Zhuo asked patiently: "Then what should we do?"

Andrew Ng recalled today's debate, thinking about the different ideas proposed by different people, and murmured: "Maybe we need to try a new type of software first, so that it can talk to computers."

Fang Zhuo hummed, expressing his usual support: "Okay, let's get started."

Andrew Ng didn't know whether to laugh or cry, and said sincerely, "Mr. Fang, the support you give us is really generous."

"I can only give what I can." Fang Zhuo said with a smile, "Give it a try. If you don't have enough people, then you can recruit more people from China. I have a simple understanding. If it can cause controversy, then the new things must have their merits."

Andrew Ng gave a thumbs up, liking the support from his boss.

However, Yike’s current situation is not so good.

He raised a small concern in this regard.

Fang Zhuo said nonchalantly, "How much money have you invested? It just so happens that a batch of GPUs from Nvidia are about to arrive. Let's try them first. I see Huang Renxun is smiling very happily today."

If Andrew Ng was not bad at business, he would have been eager to help Yi Ke sell products shirtless.

However, although he cannot sell goods, the heated exchanges between Yike and Google and Silicon Valley companies have directly further stabilized the company's stock price.

Yike seems to have solid research and development. Although it is unknown when it will have commercial value, AI defeating humans in the field of Go is undoubtedly a real breakthrough.

Yike’s stock price has fallen badly enough, isn’t it enough now?

Russell of RC Hedge Fund was quite annoyed by such voices in the media. He and a group of short sellers were also paying attention to YiKe's recent exchanges with Google and other companies. However, they did not agree that this area would boost YiKe's stock price.

Not only that, Russell also invited Cohn, a senior executive of the American manufacturer DuPont, to let professionals judge the current status of Yike and Bingxin.

"The photoresist in the ice core must have been used up! This is determined by the material itself. The ice core cannot violate the laws of reality!"

"China's photoresist is very backward. It will be impossible to provide usable raw materials for ice cores in at least five years!"

The senior executives from DuPont gave their conclusions confidently in front of the camera.

This kind of professional judgment in the monopoly field has indeed cast a shadow on the future of Yike and Bingxin for some people. As time enters July, the market performance of Yike mobile phones seems to have verified the possibility of discontinuation of Bingxin's cutting-edge technology.

However, Russell's private communication with Comborn was slightly reduced in time.

"In two years at the shortest and three years at the longest, Bingxin will have usable photoresist." Kang Boen is still confident in his judgment, "We can't overestimate the research and development of Yike and Bingxin in this area, but we can't underestimate it either."

He exaggerated the expectations on TV by double the amount, which was a genuine exaggeration, but now he is also genuinely helping Russell analyze the situation.

"Two or three years is enough, absolutely enough." Russell was very satisfied with the boost this move brought to Yike's stock price. "This is enough to cause Yike to suffer an irreversible decline."

He commented on Yi Ke: "Kang Boen, Yi Ke has a shortcoming. Do you know what it is?"

Camborn listened to the short seller's opinion: "What is it?"

Russell smiled and said, "Yi Ke has only one shortcoming, it's not strong enough."

Kang Boen has reservations. There is no company in the world that can do everything. Yi Ke is already very strong, but he also respects the short seller who is attacking Yi Ke’s stock.

Moreover, he also shared his communication with this person on Twitter.

——Yike has only one shortcoming, it is not strong enough.

On July 7, YiKe, which was not strong enough, released its Q19 financial report, reporting to the market the company's current financial situation and the changes during this period.

Throughout the second quarter, Yike's revenue fell again from Q1, from US$293.44 billion to US$181.4 billion, a drop of 38.18%, while key data such as operating profit and net profit also showed a relatively consistent decline.

Generally speaking, the second quarter is a low season, and Yike's second quarter figures in previous years were lower than those in the first quarter. But putting aside the usual influences, this 38.18% decline is still very ugly.

Although there were different opinions on Yike's Q2 financial report on Wall Street, believing that "a significant increase in advertising revenue", "a significant increase in technology licensing fees" and "unexpected growth in cloud business" jointly supported Yike's transformation, Yike's market value, which has recently remained stable at US$2800 billion, still fell sharply.

As of July 7, Yike’s market value fell to US$26 billion within a week.

Yike is known as the two poles of YA in China, and with such a decline, it is very close to Alibaba's current market value of 2289 billion US dollars, with a difference of only one and a half degrees.

Russell of RC Hedge Fund was the first to laugh at Yike's decline.

"Hahahaha, after all these years, Mr. Fang's company's market value is only as high as Alibaba's!"

“Concepts are only temporary. Without any real breakthroughs in AI, its popularity will soon fade away. Yike is returning to where it should be!”

"I've said it before, Yi Ke has no more magic power, and Mr. Fang has no more magic power!"

"He messed up everything! Yi Ke has been messed up by him!" (End of this chapter)