B station video: www.bilibili.com/video/BV1xp…
Project download address: gitee.com/drinkjava2/…
Open Source project author
Zhu Yong: Guilin Institute of Electronic Industry mechanical undergraduate course, in the national computer tide, on the programming this thief ship, after learning machine, PC, 286, 386, 486, pentium… , witnessed the development of personal computer. Basic, Delphi, C#, Lisp, Java. The main business is mechanical design, energy-saving lamp, air conditioning and oil tool design, sideline IT, network management, MRP system and Java website. Emigrated to Canada in 2006. Currently spending most of my spare time on open source projects, I have recently decided to focus on the exploration of scientific method ———— artificial intelligence, also known as the Gitee “Artificial life” project.
Artificial Life – an open source ai project that simulates the evolution of life in code
Artificial life: This is an artificial life experiment with the ultimate goal of creating a “self-conscious expression” of simulated life, based on an artificial brain model proposed in 2002. The project never ends, starting with the simulation of a simple living organism, followed by frogs, dogs…… “Will end with artificial brains with” self-conscious manifestations, “or the day humans are replaced by robots.
Interview the body
Your open source project “Artificial life” is an artificial intelligence research project, the basic idea is to simulate the evolution of life to create artificial intelligence.
2. Background of the project and why did you decide to develop this project? The reason why I do this artificial life project is also because the appearance of AlphaGo has touched me, which reminds me that the development of computer hardware has been rapid, and under the push of Moore’s Law, it has become possible to “make people”. And the existing neural network research, I feel that has not been the point, the focus on “making people”, I think this is very short-sighted and dangerous.
Artificial life is not a new concept, and I have added some of my own ideas into this project, which may inspire others a little. Artificial intelligence is not only a technical problem, but also an outlook on life. What is “consciousness” itself? If you can’t answer that, you can’t go into neural networks.
In fact, I started thinking about questions like artificial intelligence and the nature of consciousness in 1996, and wrote to the Chinese Academy of Sciences proposing research to enhance “humanoid intelligence” (typical civil science behavior, of course, no response). In 2001, he launched a “baby-making forum” site, which also received little response. Now think about it, or because academia attaches great importance to practice, to this kind of unrestrained “soft text” despise eye is very normal.
This project is trying to create a kind of intelligent single-handedly (actually is not likely), rather than a plan to use some simple techniques to explore to remind people, “made man” must be the focus in the study of neural network, it must be all the time, if you happen to have any results on technical exploration, more persuasive, It will attract more people’s attention to the field of “making babies”, which is the fundamental purpose of this project. “Made man” in addition, this project also can remind some neural network projects don’t detour, for example, I believe that can completely replace human automatic driving, automatic translation is never before “made man” this goal, this article is the technical line, don’t understand this red line will be doomed to the bankruptcy of enterprises.
At present, the project “useless” is its biggest characteristic, its basic principle is to copy the process of biological evolution, from paramecia, ants, frogs, dogs…… Step by step, until it reaches the level of humanoid intelligence.
At the beginning of the project, it implemented very rudimentary functions, so the focus was on fun, to attract more people to this area. In the later stages, of course, “useless” will become “useful,” just as an ant is useless and a frog seems useless, but a dog is somewhat useful and a child is more useful, and the reasoning is the same.
Today’s artificial intelligence research aims to be “useful” at the very beginning. For example, for tasks such as face recognition and automatic driving, many fixed algorithms may be artificially added. However, when the environment changes, these “useful” algorithms become factors restricting it to solve new problems.
4, project is suitable for the scene, the crowd and the project involves the function, technology is the study of the general artificial intelligence, rather than what it is suitable for the scene, as it represents a research direction, general purpose is to solve the problem of all the smart of artificial intelligence, including instead of human thinking, which is applicable to all need smart scenarios. Artificial life is a way to study general intelligence, and I think it is a relatively smooth and stable way, because there are so many samples of life in nature, from single-celled organisms to human beings, which are the gift of God to human beings.
Of course, this project is suitable for laymen like me, because it is more visual and intuitive, and there is no algorithm that requires advanced mathematics introduced at present. However, it should be studied by the academy in the end. After all, its target is too big, and the investment of hardware and time is bottomless. Computer is a new field, unlike high-energy physics for ordinary people has been very difficult to find anything new, and now neural network is a new field, for ordinary people, there may be technological loopholes.
The function of the artificial life project implementation is to construct a virtual environment, and demonstrate the evolution of virtual life in this virtual environment, except for demonstration purposes, it is the virtual life evolved function is also can solve some practical problems in reality, such as multivariate selection optimization, can in large sample + randomized trial and error that the genetic algorithm to implement.
Technical implementation, the project is still in groping, water amorphous, but the general idea is to use a computer to simulate a serial or parallel hardware chip based on parallel neural network, and the input and output of the brain as a whole to consider, the environment and the training itself as it the a key factor for the formation of brain structure growth, for within, but external cause is necessary, If you grow up as a child, you are not wise without education.
If some progress was made in this project, and then consider to parallel computer or hardware, now the bottleneck is not in the study of the neural network hardware, but on the software or way of thinking, so you can use the computer to construct a virtual model of high speed to speed up the trial and error and error process, of course for personal, A good computer, or even a supercomputer, can facilitate this quest because you don’t need to spend time using special programming tricks to optimize memory and speed things up.
5. Have you encountered any unexpected difficulties or gains during the initial construction of the brain model and virtual environment (the completed short-term goal)? There is no unexpected difficulty in this project, because the problem to be solved is too big, just like an ant trying to chew a mountain. This is not called difficulty, but the goal is too far away. If it must be called difficulty, the lack of time and algorithm knowledge can be counted as difficulties, and we can only do it in our spare time. This is an amateur project done by an amateur (my job is mechanical design) in his spare time.
When it comes to unexpected gains, programming usually doesn’t mention unexpected gains, because unexpected gains usually result in a Bug. But this project actually surprised me, not that it got more attention than the ORM tool project that I had worked on ten times as hard, but that when the second version of the Birth of the Eye program ran, I didn’t realize that frog behavior was much more complicated than I had expected, and it surprised me. Because it proved that eyes can be formed by randomly generated neural networks in nature, I was surprised and excited to see this phenomenon on my computer.
Images: uploader. Shimo. Im/f/j…
6. Do you expect it will take a long time to accomplish all your short-term goals? It’s hard to say. It’s not something one person can accomplish. I am no more intelligent than the others, but I have a more diffuse mind, poking here and there like a needle. I may happen to Pierce a layer of window paper, but I cannot open the window with this needle. The value of the project may lie in attracting more people to the field of artificial life research. The more people there are, the faster progress will be made.
For me personally, short-term goal is to build a combined pattern recognition and conditioned reflex function of 3 d model of brain, because I don’t know anything about basic algorithm, it is very difficult goal, if only one person, individual estimates it takes three to five years of time, may also don not to come out, this is really bad to predict, if can be predicted, We’ll have it in a couple of months.
7. In your project introduction, it was said that the project will end one day when human beings are replaced by robots. The topic of “robot replacing human beings” is still very controversial at present. This is a big topic, no one can predict the future, can only guess, and it is impossible to convince others on the basis of their own guesses. However, I still want to say that my guess will be a little more accurate than others, because I began to think about neural networks as early as 1996, predicted that Go would surpass human beings in 2001, and began to conceive the human brain model in 2002. What I think may have exceeded most ordinary people, and I am especially proud that I think I have seen through the nature of consciousness. It seems to me that information technology is advancing so fast that it will soon pass the singularity, and within 20 years there may be universal machine intelligence equal to human intelligence, that is, passing the Turing Test.
Neural network research is not only a technical issue, but also an ethical and management issue. If accurate prediction is not made in advance and countries are left to their own devices, robots may appear at an unexpected time and may be first used in the military field, which will be a disaster for human beings. If not controlled, then human control of the robot from the first day, the fate of the human robot has been doomed. Therefore, the prediction, research and development of neural network technology and robot legislation should be put on the agenda in advance, and robots should be guided to develop in a direction beneficial to human beings, that is to say, to coexist peacefully with human beings, which can be guaranteed through the robot human rights Law. But who has the final say in these forecasts, regulations and legislation? The question goes back to where it started, and the person who has the most technology has the most say, not in technology, but in predicting the future, because people tend to believe that the person who has the most technology has the most accurate prediction.
8. Have you considered materializing/hardwareizing the project? This is inevitable and the ultimate goal of this project. But for now, only the project itself or the development of artificial life can be considered. Only when more people participate in the technology, software results come out, and then consider to hardware transformation. At present, the bottleneck of humanoid machine intelligence research personally thinks that there is no breakthrough in algorithm or model. One way to speed up development is to simulate parallel neural network models on serial machines such as personal computers or supercomputers. Hardware is always behind software, and no one has the luxury of doing algorithm research directly on hardware.
9. How do you recharge your batteries after work? As a post-70 generation, I have too little spare time to think about charging, but how to discharge and reflect residual value. If I can release my good ideas and inspire others, I will be very happy. In the field of neural network, there are too many and deep knowledge to learn, so it is impossible for me to learn everything. Instead, I can only learn relevant knowledge through projects and consider how to overcome specific difficulties, which is an efficient way to use my time. If a man devote all his time to studying only in one direction, he may soon find himself ahead of all others. There is a line in Alexandre Dumas’s The Count of Monte Cristo that I love: “When you are trying to accomplish something, you are no longer the opponent of others, or, to be more accurate, the opponent of others.”
10. I wonder if you have ever watched the American TV series Westworld. If you have, does the development level of AI in the show have any influence on your project concept? I haven’t seen it, but I think it’s about robots in an amusement park who become self-conscious and rebel against humans. From its story setting, it can only be said to be a “soft science fiction”. I think “hard science fiction” refers to a fictional technology that cannot be proven to be fake with the current level of science and technology, while “soft science fiction” is the opposite, with too big flaws and too much confusion, which focuses on the entertainment value of literature and does not inspire the development of science and technology.
As a hard science fiction fan, this science fiction drama is a little contrary to my understanding of the principle of neural network technology, so I don’t like it very much. Similarly, there are “three laws of Robotics” and so on, which also goes against my belief. Because I always believe that the basic characteristics of intelligence is the “nature”, not likely to be comparable to human intelligence but again not showing self-consciousness, simple said is that if a robot can walk, can chat, so it must have been “owns” the consciousness, on the other hand, if it doesn’t “own” sense of self, They can’t even walk, sweep the floor, or wash the dishes properly. Consciousness is a gradual process, not a sudden awakening as depicted in movies and TV shows. In fact, about science fiction, I have also had some ideas, but the writing style is not good, can not write a novel, if you are interested in you can search “zhangrex science fiction”, can find me many years ago to write some science fiction ideas about artificial intelligence, the first three of which were published in the second issue of “science fiction world” in 1996.
As for artificial intelligence’s resistance against human beings, human beings need to make preparations in terms of robot ethics and legislation. It is unpredictable after the singularity, and science and technology will make big explosive progress after the singularity. Even time has lost its meaning, and everything is unpredictable. With the attitude of welcome may have a bad ending in humans, this can’t, a 50-50 chance, we can’t instead of higher intelligence to think, but to make a slavery or hostile attitude to treat more advanced machine intelligence than human, is destined to have dire consequences, even threw away the chance of the half. Perhaps the only factor influencing robots in the future will be the emotion generated by the peaceful coexistence between humans and robots. Of course, this emotion may be increasingly weak, and the fate of human beings may be entirely determined by the morality and dignity of the robot society.