Reading as a god

Chapter 242
“…the curriculum itself is not enough, but the whole environment.

Therefore, some senior students and graduate students can further understand the relevant research culture during the macro design process.

In China, at least in terms of computer science research, our experience tells us that it is very difficult to attract Chinese professors working in the West to come back to work in China, and it may be more difficult than retaining physics and biology research talents.So, I later decided to build a very strong group, which is the postdoctoral supervisor.This group is composed of postdoctoral supervisors working in the West, including more than 20 well-known researchers.Some are also good friends of mine who have made a personal commitment to provide relevant support and mentorship to students.

Recently, I have taken a new step. As Professor Alferov once pointed out, sometimes it is never too early to start such an educational project even in high school. The earlier you start doing such an educational project, the better.

We’ve also launched some innovative new projects…”

~~~~
~~~
"...we've also launched some innovative new projects..."

Listening to Yao Lao's words, Zhang Shan now regrets asking such a question!
What Zhang Shan wants to know is Yao Lao's growth experience, but for Yao Lao who is such a selfless person.

What he said is all about the experience of cultivating talents~
It is true that Yao Lao's life is closely related to education.

During Yao's more than [-] years in the United States, he has taught at top universities such as Harvard University, University of Illinois, Massachusetts Institute of Technology, Stanford University, University of California, Berkeley, and Princeton University, and has accumulated considerable teaching experience.

The most notable achievement in domestic education is naturally Yaoban~
Another identity that makes Yao have a strong connection with artificial intelligence is "teacher".In the more than ten years after returning to China, the Institute of Interdisciplinary Information of Tsinghua University led by him has filled the gaps in related technical fields in China, which has the significance of opening up new territories.

He insisted on teaching undergraduates himself. The "Theoretical Computer Science" he lectured became the first computer theory course offered in China. From then on, "Yao Ban" was passed down as a good story.

But Zhang Shan is not interested in these~
Now if you want to know a person's non-confidential information, you can rely on Jarvis' in-depth search~
Soon Zhang Shan received in-depth information feedback from Jarvis~
I don't know, I was shocked when I saw it.

It's an exaggeration to say that I was shocked, but Zhang Shan was really shocked and said~
Who would have thought that Mr. Yao, the big crocodile in the contemporary Internet world, actually studied physics at the earliest time~
While it sounds weird, it's true:
Yao was born in Shanghai, moved to Taiwan with his parents when he was young, and obtained a bachelor's degree in physics from National Taiwan University; he went to the United States for further study, and obtained a doctorate in physics from Harvard University, and then turned to the emerging computer field, and obtained a doctorate in computer science from the University of Illinois in the United States . At the age of 57, Yao resigned from his tenured teaching position at Princeton University and was invited back to China to serve as a full-time professor at Shuimu.

Sure enough, the boss is a daily cross-border ah.

Think again about what Mr. Yao said just now:

"In China, at least in terms of computer science research, our experience tells us that it is very difficult to attract Chinese professors who work in the West to come back to work in China, and it may be more difficult than the retention of physics and biology research talents. .”

The reason why the co-author Professor Yao said that it is more difficult to retain talents in computer research than in physics.

It is entirely because I have been engaged in related fields, rather than relying on subjective speculation.

These are not all of Yao Lao.

Not the best yet, Zhang Shan saw the news from Jarvis that Mr. Yao once worked part-time as a professor of mathematics.

Zhang Shan regrets again~
Chatting with such a giant-level person, I actually talked about his life, not about academics!
You must know that a small question raised casually by a person of Yao Lao's level may be a big topic if it is taken out.

Its great academic value is immeasurable.

This is not an exaggeration~
At present, all kinds of online payments are coming one after another, the root of which lies in the breakthrough in the issue of "secure multi-party computing"~
The research on secure multi-party computation is mainly aimed at the problem of how to safely compute an agreed function without a trusted third party.

Secure multi-party computation is the cryptographic basis for the implementation of many applications such as electronic elections, threshold signatures, and electronic auctions.

A secure multi-party computation protocol is called information-theoretic security or unconditional security if it is safe for an attacker with infinite computing power; if it is safe for an attacker with polynomial computing power, it is called a cipher Learning to be safe or conditionally safe.

The existing results prove that under the unconditional security model, if and only when the number of malicious participants is less than 1/3 of the total number, a secure scheme exists.

However, under the conditional security model, if and only if the number of malicious participants is less than half of the total number, a secure scheme exists.

The reason for mentioning this is that Zhang Shan learned through Jarvis that the so-called secure multi-party computing is closely related to Yao Lao.

The earliest topic of secure multi-party computation was due to Yao's millionaire problem in 1982.

Yao's millionaire problem was first proposed by Chinese computer scientist and Turing Award winner Andrew C.Yao in 1982: There are two competitive millionaires, Alice and Bob, how can they compare who they are without revealing their wealth? richer?Later, the problem evolved into Secure Multi-party Computation (SMC).

Zhang Shan knows that Yao's millionaire problem has many practical applications~
Here is an example: Suppose Alice wants to buy some goods from Bob, but the maximum amount she is willing to pay is x yuan, and the minimum selling price Bob wants is y yuan. Alice and Bob want to know which is bigger, x or y?
If x > y, the two parties can continue to bargain, otherwise there is no need to waste words, but neither of them wants to tell the other party their bid, so as not to be at a disadvantage in the bargaining.

This looks like a primary school application problem, but in fact, after it is linked to finance, programming and security at the same time, this problem is really not so easy~
However, the fact is that in 1982, Professor Yao Qizhi gave a solution to the millionaire problem after he raised it.

This scheme is used to compare two numbers to determine which is greater. Alice knows an integer i; Bob knows an integer j, Alice and Bob want to know whose number is the largest, but neither wants the other party to know their own number.For simplicity, it is assumed that the range of j and i is [1, 100]. Bob has a public key Eb and a private key Db.

(1) Alice chooses a large random number x, and encrypts it with Bob's public key c=Eb(x);

(2) Alice calculates ci and sends the result to Bob:

(3) Bob calculates the following 100 numbers: Yu=Db(c-i+u), (u=1, 2,..., 100) where Db is Bob's private decryption key. Bob chooses a large prime number p (p should be slightly smaller than x, Bob does not know x, but Alice can easily tell him the size of x), and then calculates the following 100 numbers: Zu = (Yu mod p),.Then verify that for all u≠v, |Zu-Zv|>=2, and verify for all u: 0
(4) Bob sends the following sequence and p to Alice: [Z1, Z2, . , Zj + 1, Zj+1 +1, . operate;

(5) Alice verifies whether the i-th number of this sequence is congruent with x modulo p.If it is congruent, she concludes that i<=j; if it is not redundant, she concludes that i>j;
(6) Alice tells Bob this conclusion.

For a while, Zhang Shan couldn't help feeling a lot~
Feelings are emotions, how to let Yao Lao continue to talk about some academic things~
Before Zhang Shan could think too much, Yao Lao changed the subject and started asking Zhang Shan~
"Student Zhang Shan, I also heard that you plan to continue to invest an additional 20 billion US dollars in the research of artificial intelligence. I don't mean to pry into commercial secrets, and I have no intention of doing so. It's just that what you want to do is obviously not research. Ordinary AI, but work on superintelligence.”

"Current super artificial intelligence has many constraints..."

Zhang Shan listened carefully, and Yao only mentioned the two major technical bottlenecks of "robustness" and "interpretation", as well as the ultimate challenge of future super artificial intelligence.

Zhang Shan pays a lot of attention to artificial intelligence, so she naturally understands what Professor Yao means~
Regarding robustness issues, it can be interpreted as fragility, non-security, and non-reliability issues.

The design of artificial intelligence systems has always focused on functional design and neglected reliability and safety design; or functional first, followed by safety and reliability.

This phenomenon is particularly serious in early products.In the field of self-driving cars, the ultimate constraints are safety and reliability issues. In the future, the research and development of driverless cars will become a topic of "always on the road" due to reliability and safety issues.

Regarding interpretability, it can be understood as open source in a broad sense.

Explainability is how to give people insight into AI systems.Taking cars as an example, in the era of the industrial revolution, car drivers knew the principles and structure of cars at a glance;
Now, the car is just a black box for the driver, with only an application interface such as steering wheel, accelerator, and brake.

In contrast, cars in the era of the industrial revolution are highly interpretable, and emerging cars with artificial intelligence cannot be explained.

Similarly, mobile phones, digital cameras, and televisions are inexplicable compared to telephones, cameras, and electronic tube televisions. They are black boxes for all users, and no one disassembles, repairs, or understands their internal structures. , which is a very realistic and advanced fool-like application model of artificial intelligence products, which completely separates knowledge innovation from innovative knowledge application.

However, for the field of artificial intelligence innovation, inexplicability is a huge obstacle to technological innovation.

The transparency of hardware and the open source of software have solved the interpretability barriers of technological innovation to a certain extent.

At present, explainability obstacles are prominently manifested in the field of algorithms (one of the three foundations of artificial intelligence), which hinders the promotion, evaluation and marketization of algorithms, and the fragmentation of algorithms will seriously hinder the development of artificial intelligence.

It has to be said that Yao is a visionary with unique insights into artificial intelligence.

"Professor Yao, students remember that in a speech at Shuimu University in the early years, you proposed the concept of "big science" in the era of artificial intelligence.

I remember you said at that time: The era of artificial intelligence has entered the era of cross-integrated development of many powerful sciences, and each strong discipline will interpret artificial intelligence from its own perspective (self-righteously)..."

Professor Yao really didn't expect Zhang Shan to remember this, how could he not remember~
In fact, at that time, Yao mentioned super artificial intelligence and proposed three expectations for super artificial intelligence.

In the past, people have always divided artificial intelligence into weak artificial intelligence and strong artificial intelligence.

Weak artificial intelligence begins with the birth of semiconductor microprocessors, which is an artificial simulation of human intelligence; strong artificial intelligence begins with brain engineering, which is an era of intellectual innovation in artificial intelligence.

Due to the crazy development of artificial intelligence at an ultra-high speed, people cannot foresee the future of artificial intelligence, so there is the concept of super artificial intelligence with strong artificial intelligence.Yao believes that super artificial intelligence has great uncertainty, and human beings will face many potential threats.

For the future of super artificial intelligence, Yao gave three principles that will restrict intelligent machines in the future: one is altruism, that is, the interests of people are above machines; the other is humility, that is, machines cannot be self-righteous; the third is dedication, that is, machines can learn to understand people's preferences.Each principle must be implemented with rigorous algorithms so that artificial intelligence can be effectively harnessed.This is a rational expectation for the unknown era, hoping that artificial intelligence will eventually move towards a rational path.

Those who are one step ahead of the times are pioneers, and those who are two steps ahead of the times are martyrs.

Zhang Shan knew that the reason why Professor Yao said these things to him sincerely was because he was afraid that he would be too radical in the industry and suffer heavy losses.

This is an elder's love for the younger.

But Zhang Shan has no choice at this time:

"Yao Lao, AI has entered the deep water zone of industrial implementation: judging from the technology maturity curve, artificial intelligence is in the bubble period of the "Valley of Death", and AI that does not bring real value will be eliminated."

"In the past 20 years, from the Internet to the mobile Internet, the next natural extension is the era of the Internet of Things. AI is a core technical capability in the Internet of Things. In the short term, it will continue to be the main axis of future technological innovation, but it is not the core point of industrial landing. .

Realizing the real commercial value of AI, most of the companies that can truly scale and commercialize are offline-oriented and combined with the real industry. AI and IoT will be more and more connected in the future.

If artificial intelligence is compared to the "brain", the Internet of Things is the "body". The combination of the two can make the world more intelligent.

Algorithms are the core variables in the AI ​​and IoT era.More visually, it is an algorithm based on deep learning.

AI innovation is an infinite game: it started decades ago, and it will continue beyond.Finite games are usually limited in time, with winners and losers, and clear rules; but infinite games continue indefinitely, with changing rules and players. "

(End of this chapter)