0509 Liu weipeng's toplanguage Party [1]

Zheng Yu @ playmates SR

Beijing. Wudaokou.

Saturday. September April 15. The temperature is suitable for going out.

@ Pongba recently left Nanda and was invited to Beijing by Microsoft. Following the previous month, he and Li xiaolai met each other. (The Jianghu once circulated a sentence: Since I read Li xiaolai, I did not write my feelings about life. Since I read Liu weipeng, I did not write about the learning and thinking methods) then he convened a toplanguage group.

Seven technicians attended the meeting (arranged in the order of appearance ):

• @ Pongba blog: http://mindhacks.cn Microsoft Asia Research Institute
• Wang leheng blog: http://joyfire.spaces.live.com Institute of computing, Chinese Emy of Sciences
• @ Googollee blog: http://googollee.blog.163.com/nthcode
• @ Zhengyun blog: http://blog.csdn.net/zhengyun_ustc/ play network
• @ Tinyfool blog: http://www.tinydust.net/dev yinxingtek
• Zhang Shen Peng blog: http://zsp.javaeye.com/Douban
• Sun Yong http://blogs.sun.com/yongsun/ sun

 

. Let's look at table 1 for the discussion topics. As we all know, there are more people (I .e. nodes) and more interactions (I .e. connections) between each person. It is difficult to organize a readable topic context. The following texts are basically from the oral expression of each person. I try to combine them as much as possible. First, I would like to discuss two topics first.

1. Artificial Intelligence and Evolution

It may be an offline continuation of the Self-replicating_machine topic on TL.

When joyfire from the biological information research group introduced that the research direction was "the complexity of genome information structure and the fundamental laws of genetic language", Everyone began to brainstorm.

DNA is a string

First, I feel that DNA is well designed. If God isProgramEach DNA is considered as a string array of four elements, with dozens of MBCodeAnd control all the growth and behavior of people. It is a great programmer.

Note: single-chain DNA can be considered as a string consisting of A, G, C, and T. We simplified it to string search and matching, so we can almost think that gene sequencing or determination is like a search engine for string operations. The corresponding application is the determination of the SARS gene sequence and the determination of melamineMass Spectrometer (this extends another topic: disruptive innovation, as shown below).

The topic is "Evolved" due to the formation of such sophisticated programming skills ".

Simulated Evolution

Everyone is devoted to the discussion:

It is not possible to save. simulateAlgorithmBuild an environment, put a piece of code that represents life (just an early life form that represents the same as a crawler), simulate genetic variations, the Darwin rule for survival of the fittest, and use high-speed concurrent computing, is it finally a high-level life form that evolved to a certain extent?

ZSP believes that algorithms may run, but because the simulated dimensions are designed by humans, they must be far less than the real world. Due to quantitative changes, qualitative changes can be caused, and too few variant dimensions and environmental dimensions are difficult, therefore, the core issue is how to extract "Features.

In general, humans have abstract capabilities. They can use various methodologies to find the "features" of a thing or transaction and tell the "features" to the machine. The machine has a strong computing capability and can be executed. At this level, self-replication robots are actually not difficult to create. After all, "Features" can be solidified.

However, if machines cannot have abstract capabilities and cannot automatically find external or internal "features" that allow evolution to continue, what stage can "Evolution" reach?

Evaluation of evolutionary results

Of course, the ultimate result of "Evolution" here is not the emergence of smart creatures like humans, as long as they are logical and can continue. Because at a certain point in time, you cannot use good or bad to evaluate a certain evolutionary life form, such as dinosaurs. You may think that they are the best life form in a certain period of time, but soon after they were all destroyed by random external environment changes, is it a good or a bad "Evolution "? For example, if we are the best smart creature, but if we are self-destroyed by war or destroyed by aliens soon, will our evolution be right or wrong?

Therefore, this program mainly checks whether the living body can evolve continuously to generate intelligence.

Of course, we will talk about the spore game. Its evolutionary logic is a bit simple, but it is still a very good game that integrates real-time strategy and evolution.

Brain: connection, node, address

@ Tinyfool mentioned that people used to think that the brain is divided by functional area and cannot be replaced by each other. However, the experiment later found that the functional area of the brain can be transferred and "rebuilt ". To put it bluntly, after some organizations in the left brain are destroyed, the functions (such as sports and language functions) that are originally loaded will be lost, but over time, the brain can quickly re-build these features in the right brain area.

@ Tinyfool mentioned that the Network was also a seemingly sluggish solution. At first, it was just a static webpage provided by every host. Later, the HTTP protocol appeared to be not as beautiful as it was, but the magic thing happened. Because only three elements are defined:

• Node
• Connection
• Address

At the beginning, it was not surprising. However, with the evolution, each node gradually had its own intelligence, such as dynamic web pages. nodes were found through addresses, and nodes were connected to each other. nodes were intelligent, therefore, the Internet is omnipotent. It is an organism with massive amounts of information and intelligence.

This seems to have some interesting commonalities with the brain mechanism.

Prior knowledge or instinct

Joyfire believes that there is a big problem in rebuilding the functions of the brain and self-evolutionary algorithms of the program, that is, "A prior knowledge ".

Humans or animals possess a lot of prior knowledge that is solidified in genes in the evolution of generations. For example, if it is hot, it will scale down, for example, the instinct does not like the bitter taste.

How does a program simulate these things?

Or, how can we restore this "prior knowledge" when the brain reconstructs the functional area, if these things are stored in a fixed area.

One Party believes that prior knowledge can be simulated. For example, if there are multiple variations, the genes of each mutator carry this variation. If other mutators are destroyed during evolution, then the surviving mutator naturally carries this variation, which explains the question of where "prior knowledge" comes from. Of course, this is just an ideal description. Because it does not involve any variations that will solidify into genes. For example, if a puppy who is born to lose a leg does not pass this variation during reproduction, that is, some variations do not inherit. Who will decide whether to inherit the variation?

For machines, "prior knowledge" is a bit as difficult as "features". The reason is as described above, it is difficult for machines to determine which ones should be solidified, I still cannot explain this problem with "Evolution.

In this way, we go back to the feature extraction problem.

2. Disruptive Innovation

The aforementioned joyfire Mass Spectrometer for melamine determination has extended another topic: university education, disruptive innovation and domestic environment.

We should often see such a phenomenon that our research institutions or companies, based on the wisdom of the Chinese, have almost approached or surpassed the world's first-class level in a certain field of research, at this time, everyone was overjoyed, but with the disruptive innovation of a foreign experimental group or small company, the previous scientific research results were destroyed and pulled away, and people continued to take the lead, of course, it was replaced by a group of people.

The problem lies here.

It is said that disruptive innovation (hereinafter referred to as "breakthrough") is basically one after another in the world.

However, in fact, disruptive innovation is hard to appear in mainland China.

The simplest example is that when our portal model is dominant, SNS were born. Twitter was born when our celebrity blog attracted billions of clicks every day.

Joyfire said that in the field of biological information, when we quickly approached and surpassed them, they suddenly experienced a leap growth, and we would not "jump ".

One party comes to the conclusion that China does not invest any money, but the investment area is always very narrow and cannot produce chain linkage effect. It is always stuck in the previous step or the next step.

Suppose that the country's key development of gene sequencing requires several microns of test tubes, We can't produce, we can only buy foreign, then assume this is a problem with precision instruments.

Okay, in the field of precision instruments, is also strongly encouraged in China.

However, the same process and production line cannot reach the foreign level. For example, the test tube mentioned above requires that the internal diameter should not be rough or thin, and the error should not be greater than a certain value, however, domestic production cannot meet the standards. As a result, if you do not pay for your next job, you have to purchase it abroad. If you cannot purchase it, you cannot do the research tasks of the next job. As a result, precision instrument manufacturers have closed down. So, a vicious circle. This is a bit like the history of our large aircraft for so many years.

Modification history:

0509

@ Tinyfool added:"

This is probably the case, that is, we used to think that all functional areas of the brain have distinct functions, but now many experiments have proved that if a functional area is cut down, other areas of the brain may take over these functions (although they are completely irrelevant in the past. In addition, physical stimulation and sufficient activity and cell regeneration of the brain are required ). This is exciting, that is, all functional areas of the brain are equal and can be dynamically deployed.
It is generally believed that intelligence comes from the connection between neurons and cells. For example, the ability to understand language is achieved by the connection of a series of neurons. If these cells and connections are cut off, but if the external sound stimulation of the brain does not disappear, neurons that are completely irrelevant to this ability may be stimulated and associated, then, these stimuli are processed to restore the ability to understand the language."

 

0510

Joyfire added:"

At that time, it probably means that the human body is such a complex and efficient system, but all the design blueprints, orSource codeIn fact, they are all stored in a few DNA strings. (And
In fact, only 5% of the DNA is translated, And the rest may be traces of evolution, which can be regarded as the historical code commented out.) as a programmer, God is very efficient.

Moving forward, we mean that our existing hardware and software structure is far from the limit of nature. Whether it is hardware storage and retrieval capabilities, or software coding efficiency.

After this sentence, we will discuss that DNA is a dynamic language that can be in turn influenced by the exe-protein compiled by it. For example, each person has only one nose and two nostrils,
High Fault tolerance. But on the other hand, code has a lot of flexibility: everyone has a lot of differences to facilitate diversity and evolution ."

Reference resources:

0424 xiaolai and pongba:

1: Sketch

"

Yu Cheng Liu weipeng

Huo Ju Li xiaolai

Sijo

My fat rabbit porridge

"

0509 TL party:

1: TL party, encountering several Titans:

"Funny content: from the commercial model of Internet advertising, to the credit and logistics system of e-commerce, to the recommendation and Evaluation of Semantic Network Technology and source mining; to artificial intelligence, genetic evolution, and brain mechanisms, to the needs of various features that must be well-developed in Douban, to Beijing food and a number of "Beijing offices", to housing prices and economic crises; to the state-owned monopoly industry and it entrepreneurship; finally to their respective work is facing algorithm and engineering difficulties."

2: IP address, IC, and iqcard. You can tell me the password:

"This is back to the things discussed at the TL party yesterday. The essence of intelligence is" ultra-large storage + ultra-fast retrieval "(in other words, there is enough prior knowledge ), or another computing model?"