current position:Home>How to prove whether artificial intelligence has awakened

How to prove whether artificial intelligence has awakened

2022-09-23 09:52:01User 3147702

1. 引言

上一篇文章中,We look back on one of the big news of the year:Google a named LaMDA The chat robot of artificial intelligence is alleged to have human emotion,这会是真的吗?

谷歌的 AI Really woke up?(上) -- LaMDA Distinguish between true and false events

综合各方面情况来看,This big news is likely to be a man-made farce.

But the whole thing is still enough to cause us a series of reflections:Is there any way we can know if artificial intelligence is really intelligent??Even if artificial intelligence really has human emotions and consciousness,Is there any way we can prove this through testing??

此时,You must have thought of the famous Turing test.Today, let us speak of from the Turing test.

2. From Turing Test to Overall Turing Test

2.1 图灵测试

How to measure the development of artificial intelligence,It is the most basic problem in the field of artificial intelligence.

1950 年,Allen, the British mathematician known as the father of computing·Turing in his paper《计算机器与智能》A thought experiment on intelligent detection is presented in:Say a human subject and a human interrogator are placed in two separate rooms,Place a computer under test in another room,The three rooms do not know each other's situation,Communication is only possible by telex,in their communication,If the human interrogator cannot accurately distinguish between human subjects and computers,then think that the computer already has intelligence.



随即,Turing proposed the specific inspection standard:只要有超过 30% of human interrogators misjudge whether a subject is human,Then think that the computer has developed intelligence.

一直以来,This testing method and standard are regarded as the standard,And Turing himself made a prediction:到 2000 年,artificial intelligence can pass 5 Minute Turing Test.然而直到 2014 年6月7日,organized by the Royal Society“2014Turing Test Conference”上,It was finally announced that artificial intelligence passed the Turing test.

2014The University of Reading, the organizer of the annual Turing Test Conference, issued a press release stating that,Artificial intelligence software developed by Russian developers“尤金·古斯特曼”(Eugene Goostman)通过了图灵测试,Although this test is premised on the inability of a human interrogator to distinguish whether a subject is a 13 year old ukrainian boy still computer,这个“十三岁”的前提,Let the conclusion that passing the Turing test hold is full of controversy.

The Turing test the greatest significance,It gives a practical and quantifiable artificial intelligence really smart judging criteria,Let this question no longer be subjective,thus resolved a huge controversy,But for the Turing test itself,Controversy never stops.The most powerful challenge to it is 1980 John the American Philosopher·Searle“中文房间”实验.

2.2 中文房间

Is the content of the Chinese room experiment,Suppose there is a person in a room who only speaks English,The people outside communicate with the people in the room in Chinese by letters through the small window of the room..English speakers in the room have enough books and materials for translation,于是,This English-speaking person can communicate with the pen pal outside the room in Chinese without any flaws,And in the eyes of this pen pal,He will think there is someone in the room who is fluent in Chinese communicating with it.

If we replace the English speakers in the room with computers,Suppose the computer has enough data,Can be input in various languages if-else 判断,Get answers that humans can understand,People will completely mistake the computer for thinking,even wise,但实际上,everything a computer does,but simple if-else just match.

Looking back at this year's big news will find,Conversations surrounding exposure,“中文房间”The idea is precisely what people think of this news LaMDA Why Chatbots Are Not Smart.从结果上来说,LaMDA Probably has reached the point where it can pass the Turing test,But since all the questions asked by Lyme can be found on the current Internet,所以 LaMDA It is likely that this is the stay“中文房间”中的计算机,All he does is organize the answers found in the database according to fixed rules into human language patterns for output,It does not have the so-called intelligent thinking ability by itself.

那么,更进一步地,我们不禁要问,What is the difference between a robot in a Chinese room and a real human being??这恰恰是“中文房间”Controversy since the birth of the experiment.

有的人认为,The English-speaking person in the Chinese room is undoubtedly wise,He cannot be denied the existence of wisdom just because he does not understand Chinese,进而,Software that can fool humans with fluent human language and responses,even though its interior is only if-else 的符号系统,It should also be considered intelligent.But the other party thinks,Just as software that simulates the flight of an airplane does not,Software that simulates human intelligence cannot be considered true intelligence,After all, anyone is not only a symbol system of imitating others.

2.3 Overall Turing Test

Despite the controversy,但毋庸置疑,Searle's Thought Experiment Shakes Turing Test Status,How to design a project that can dispel all kinds of disputes、Feasible experiments to prove that machines possess human intelligence are new questions.

What is the biggest difference between a computer and a human??A typical answer would be:no mind.那么,What is the mind?is to read enough,and browse enough content on the internet,And then relying on those memories to answer questions??One must reject this assertion.Because human thinking is not only derived from the words and data at hand,Also from life experience,The traditional Turing test, which only examines verbal communication skills, is too low a standard.

由此,Haynard designed more complex and comprehensive experiments -- “Overall Turing Test”,And divide the five level for test results,The traditional Turing test is equivalent to T2 级别,Higher level requirement to robots with sensors,Optical signal to external input、Incorporating touch signals, etc. into the scope of machine learning,enabling robots to have human-like“感觉运动能力”,recognizable、识别、Describe and manipulate real objects,Ability to predict the consequences of one's actions,and have a perception and preference for different operations.

更进一步的,Shi Weize believes that,We have always overlooked an important point of view,That's social behavior and evolution,他认为,If you want to confirm that the robot is intelligent,long-term test,Reflect the different generation of robot intelligence autonomous phenomenon of evolution,And also requires robots to have social behavior,This further test experiment he named“True Overall Turing Test”.

然而,The Total Turing Test and even the True Total Turing Test are equally controversial,Because they are an upgrade to the traditional Turing test,将“打字机”become a smart robot,It just increases the difficulty of passing the test and the complexity of the tested robot,Essentially still can't get rid of Searle“中文房间”的争议.那么,Is there a way to pass some kind of direct methods we confirm whether a robot have smart?就像 LaMDA 说的,By reading its neural network code,Find the variables that control emotions,to confirm the real existence of emotions.

很可惜,恐怕不能,from brain science、State of the art in neuroscience,We have not yet been able to find the deep causes and mechanisms of human consciousness and emotion from the material level of brain nerves,There is no way to judge what kind of neural network structure can have human consciousness and emotion..当然,如果有朝一日,Humans can breakthrough in brain science consciousness is the source of the problem,The artificial intelligence developed based on this theory is bound to have self-awareness and emotion..然而,现阶段,Since we can't find the internal evidence for the existence of intelligence from the mechanism,Can we judge from the outside??

This involves the famous big question in philosophy -- 他心问题.

3. 他心问题

3.1 I think so I'm having trouble with him

16世纪,With opposition to church authority,The Renaissance movement, which is striving for the liberation of human nature, has been in full swing for more than two centuries on the European continent.,With the rise of natural science,The contradiction between science and ignorance is highlighted.此时,Philosopher Descartes for these built on authority、Faith-based knowledge edifices feel increasingly dissatisfied.于是,他喊出了:“Use Philosophy to Lay the Foundation for Knowledge Building”的口号,他说:

Philosophy is a human can know the most perfect knowledge,both to guide life,Also for the health and development of various arts.

He also hopes to reconcile faith、灵魂、The contradiction between ideas such as liberty and science,笛卡尔认为“everything in nature,even mental processes and emotions,must be explained mechanically,without resorting to form or essence”.

然而,To be based,It is bound to tear down the existing knowledge building,So Descartes said:

I will not only for you is in the world,whether there is earth or sun;And for me if I had eyes、耳朵、身体,Even if I in to talk to you,Not sure if you're talking to me.简而言之,i would doubt everything.

If everything in the world is questionable,That really want from where to find the real true?Descartes famous philosophical proposition is presented“我思故我在”.Since I'm thinking,Then it means that there is a“我”的存在,The existence of the subject is thinking,Even doubt thinking of all the object exists,The main body of thinking there is no doubt.

基于“我”的存在,How to infer the existence of others?when we stand on the platform,Saw sturdy people in coats and hats walking by outside the window,How can we be sure it's a person,instead of an automatic machine covered in a coat?更进一步的,How can we know whether other people really like us to think,rather than an automatic machine?这就是著名的“他心问题”.

那么,How to solve his heart problems?How to know if someone else has a soul?笛卡尔认为,Language must come from the mind,Even parrots can imitate humans by uttering certain words,They also cannot cope with infinite locales,Cartesian assertion,Even if there are automatic machines,Because machines have no minds,It is also still impossible to produce language that is meaningful for a long time and can adapt to the locale,Because language is the content of thinking.他说:

Can't imagine how such a machine should produce different arrangements of words,to appropriately and meaningfully respond to what is being said,like the most boring person can do.

显然,limited by that era,Descartes could not imagine the birth of artificial intelligence.换句话说,Presumably the cartesian must be supporters of the Turing test,Because he believes that as long as we have language,It can be proved that the machine has the human mind.

由此可见,Descartes' argument for the problem of his heart is incomplete.Although Descartes, the father of modern philosophy, solved the“If everything is doubtful,So how is it possible for humans to acquire knowledge?”的问题,So as to achieve the purpose of laying the foundation for a solid scientific building,But his theory caused another dark cloud to float over the Science Building.,从此以后,The problem of his heart has become a major headache for thousands of philosophers for centuries.

3.2 Phenomenology and the Problem of the Other Mind

Descartes hopes to build the edifice of modern science based on the solid foundation of the existence of the subject..但是,overemphasis on subject,is bound to arouse suspicion of the object,The question of his heart inevitably became the cloud of doubt floating over the edifice of Cartesian science.

19世纪末,The German philosopher Husserl also maintained the idea of ​​purging the naturalistic attitude that pervades the transcendence and exclaimed“back to the thing itself”的口号,He advocates that we study things themselves without any a priori.This idea, in the same vein as Descartes, was also quickly“他心问题”的攻击.

Since it is necessary to set aside the naturalistic attitude,research questions with pure subjectivity,How to ensure the certainty of the object,rather than falling into solipsism、The predicament of the unknown?

In order to achieve the accessibility of the other,Husserl incorporated elements of Jungian psychology into his phenomenology,intersubjectivity.Husserl thinks,It is because of us that I have developed access to the hearts of others,It just confirms the objective existence of the other,他说:

If I to the minds of other people have access the same as for my own consciousness,Then the other ceases to be an other and becomes part of me.

在此基础上,The main body through empathy,subjectivity perception.this is fiction《Androids dream of electric sheep》中,A sign of artificial intelligence awakening -- empathy.

近年来,随着脑科学的发展,mirror neurons discovered,To a certain extent, it confirms the theory that wisdom comes from empathy,进而,罗伯特·M·Gordon and other scholars put forward the simulation theory,他们认为,through mental simulation,Humans can make sure to be the object of experience the existence of the real intelligence and self-awareness.They are summarized in the following four steps:

  1. the simulator sees、Perceive or imagine the target's actions and related situations.
  2. The simulator adopts the target's perspective to generate a fake mental state.
  3. The simulator interprets and makes predictions about the target's behavior under a fake mental state.
  4. Validate the target's behavior by predicting the target's behavior.

Based on the modeling and reasoning,预测与验证,Humans feel the feelings of others,In order to finally know whether the other party has real intelligence.

4. 总结

如今,科技的发展日新月异,从深蓝到 AlphaGo,From artificial retardation to intelligent voice robot,We've been able to feel the rapid growth of artificial intelligence and their changes to our lives.

Can artificial intelligence break the boundaries between machines and humans one day?,truly have their own sense of self and emotions,It can even surpass human emotional experience,This seemingly sci-fi problem seems to be approaching our reality.

在影视剧《西部世界》与游戏《底特律·成为人类》中,Robots are gradually awakening in their daily communication and learning with humans,keep breaking a certain limit,become a conscious being.但时至今日,Human science has yet to explain the origin of consciousness,Artificial intelligence still stays at the superposition of neural network algorithm layers to achieve more complex computing tasks.With the unlimited superposition of computer neural network algorithms,是否存在“量变引起质变”,reach novel、影视、Described in the game suddenly awakened state,The vast majority of relevant practitioners still hold a negative attitude.

毋庸置疑,AI is getting smarter,But to break through that layer of invisible boundaries,恐怕还有一段路要走.

回到文章开头的问题,Is there a way we can tell if an AI is real?“觉醒”了呢?综上,we have two weapons:

  1. Continue to develop brain science,Until the research out of the real source of human consciousness,进而,In theory we can have the means to find the basis of artificial intelligence have consciousness,to make accurate judgments,这便是所谓的“The theory of theory”;
  2. Through the above empathy and reasoning,模拟与预测,to verify whether artificial intelligence has emotions,这便是所谓的“The theory of simulation”.

在我们看来,在当下的环境中,Simulation theory is simpler、实用、可操作性.然而,显而易见的是,Even a robot that can pass the Turing test,Want to pass the test of simulation theory,AI still has a long way to go.

当然,Is it possible for computers to give birth to true consciousness and intelligence?,This is also a controversial issue.But we can still think about,When artificial intelligence wakes up,How will humans get along with artificial intelligence,Will there be a conflict between artificial intelligence and humans??在我看来,Fears that technological development will eat back into humanity is a common theme in science fiction,但在现实生活中,我更愿意相信,Human will to think about how to make science and technology to improve the survival of human beings,Make use of science and technology,也许,The awakening of artificial intelligence is not a scary thing.


copyright notice
author[User 3147702],Please bring the original link to reprint, thank you.

Random recommended