The pioneer of artificial intelligence pioneers such as Turing has proposed that computers can think and act like people，Will artificial intelligence be built with intelligence -like intelligence and even surpass the latter，It has always been a question that is in the exploration and controversy。In the new wave of data -driven artificial intelligence in recent years，The focus of this problem gradually disputes between theoretical possibilities，Converted to how to cope with the possible subversive innovation that may occur in technology。Fighter human Go players from deep learning -based artificial intelligence，To the recent generation artificial intelligence represented by ChatGPT, it has achieved amazing success，Especially ChatGPT is clearly made for natural language issues and prompts、Answers with correct grammar，Like a giant magic performance, people's imagination of the future of technology。Facing the tsunami -like impact brought by this step -by -step "creative destruction"，People not only see the potential risks of sudden evolution of general artificial intelligence and even super intelligence，And more and more strongly realized，It is necessary to seriously think and take the open social risk and value ethical challenges that may be triggered by it。

　bet365 live casino games

ChatGPT's emergence has exceeded the cognitive framework of many artificial intelligence suspects。ChatGPT presents a super powerful human -machine natural language dialogue ability，Not only breaks the limits of many pessimists to a fairly extent，At the same time, it also exceeded the expectations of many technical optimists。This is once again，No matter what is intelligent、What is the theoretical idea of ​​artificial intelligence，Still reflection on theoretical theoretical of artificial intelligence，There is bet365 best casino games often a large cognitive gap between bet365 best casino games The implementation of artificial intelligence in technology and engineering。The cause of this gap，The tension between theoretical thinking and engineering thinking in the development of artificial intelligence。If you can handle unlimited tasks、"key requirements" such as autonomous and valuable systems as the standard for general artificial intelligence，ChatGPT is obviously not enough。But the question is，How to test these "key requirements" in engineering。

Many intelligent theoretical cognitive frameworks often lack necessary humility，Failure to take the position as one of the perspectives of inquiry，It is easy to fall into the closed -end negative thinking that adheres to the "priority" standard。and engineering thinking is mainly reflected in the accumulated innovation and emergence of innovation in engineering practice，It is an openness of the "post -test" iteration based on the evolution of the technical industry，Often choice of paranoia for certain technical paths，and can give these options in the evolution of technical evolution。Affected by a cognitive gap，Theoretical researchers and critics cannot foresee major breakthroughs that may emerge in the practice of artificial intelligence engineering，Engineering practitioners and applauders are difficult to predict the revolutionary effects caused by technical breakthroughs in the ethical level of social value。The overall cognitive state formed here，Obviously unable to deal with open ethical risks including super intelligence。

　Human -machine interaction with human -machine alignment

Whether it is the theoretical reflection or the engineer，In the relevant discussion, artificial intelligence and human intelligence are preset as a smart body that is separated from each other，and more focused on the difference between the two and the one who eliminates the long，and fewer interactions between the two as the starting point of thinking，Understand the essence of artificial intelligence from the dimension of the technical social system and intelligent ecosystem。bet365 best casino games Actually，From deep learning based on big data to large model -based generative artificial intelligence，Its innovative applications have occurred in data、Algorithm、In the giant technical social system built by computing power，It is formed in the highly social human -computer interaction intelligent ecosystem。The reason why they can achieve functional breakthroughs，Although the function of the data volume and model parameters reaches a certain scale。But at the same time, you must also realize，One of the more important maybe it is，Human feedback fine -tuning and human -computer intelligent interaction in use effect on its performance improvement。

During the bet365 best casino games research and development process of ChatGPT，Openai's technical route is in a highly social human -computer interaction intelligent system。New Formula for Artificial Intelligence used in ChatGPT，Learning based on the internal homogeneous form and structure of natural language。Where，There are a large amount of text data set，It also includes a large number of human -machine dialogue data during operation。This content generation idea is the combination of learning and pre -training: first automatically extract relevant content and aggregate，Then use the human -machine alignment project to correct the goals and value。According to Openai's statement，The background of this implementation human -machine alignment project is，OpenAI's big model of GPT series and ChatGPT's technical positioning is to explore common artificial intelligence。To prevent the influence of subversive social ethics that may bring，Developers pass through human beings、Feedback、Audit and other engineering methods，Calibrate the value conflict and ethical disputes in the generated natural language，Continuous supervision and continuous optimization of generating content and language expression strategies。This makes the output of ChatGPT relatively cautious about value sensitivity、Holding，Active avoidance and even refuse to answer questions about disputes。

bet365 best casino games The implementation of the human machine alignment project，Due to the existence of huge social ethical risks that may cause human survival risks due to the existence of super artificial intelligence，Technical development and value ethical regulation of generating artificial intelligence from the beginning is synchronized。From this，You can get two important revelations。On the one hand，The implementation of the human machine alignment project，Value ethics correction and prevention of vicious consequences on the generated artificial intelligence is feasible in engineering，This provides a reference experience for the bottom line and the red line for it in the innovative application。Of course，It must be clearly pointed out，ChatGPT as a language model does not really understand the connotation of various values。On the other hand，Human -machine alignment project is implemented on the basis of human -computer interaction，Whether in training data or human feedback links such as artificial calibration，Both the interests of the stakeholders and the good and evil，It will be affected by various value presets and cultural selection，There must be a certain bias。

　Knowledge Authority and Turing Trap

Value Ethics Correction performed by the human -machine alignment project，Although it helps the legal regulation and ethical governance of the content of artificial intelligence。But more importantly,，generated artificial intelligence represented by ChatGPT is a new combination of machine intelligence and human intelligence，We are opening an unprecedented era of automatic generation of knowledge and comprehensively reshaping life。From the perspective of the paradigm of the knowledge production method，If big data analysis brings new analysis，Then the large model opened by ChatGPT plus human feedback automated knowledge generation method is a new comprehensive for intelligent future。Hugging for this new comprehensive，It will force us to face a series of new ethical challenges。In addition to the current hot discussion of learning and research bet365 Play online games integrity、Instead of infringing intellectual property rights and other issues，Especially worthy of attention is the following two open social ethics challenges。

First, it is the authoritative hallucination of knowledge that is regarded as "God"。Hugging of knowledge generation automation will face the paradox is，The generated artificial intelligence system can bring improvement of knowledge generation efficiency，But it is not the understanding of knowledge and knowledge。Put the answer to the language form generating system from massive training data，It does not really understand the meaning of the input and output data，No own goals，I don’t know what research and learning and why study and study。But people often have an illusion that intellectual producers that regard them as automated，and easy to ignore，Although they can produce a coherent text，But its significance and coherence are formed in the interaction between humans and machines，They didn't try to express anything。If you can't know this illusion，It is easy to trigger the risk that treats generating artificial intelligence as the authority of knowledge and moral authority。further development of ChatGPT，It is expected to evolve into an artificial intelligence assistant in ordinary people's daily life，Become answering knowledge、An important tool for distinguishing between right and wrong and even distinguish between good and evil。Given that ChatGPT does not really understand the knowledge content it generates and judgments on right, wrong, good and evil，and sometimes ridiculous errors or randomly stacked and fabricated content，In the absence of critical thinking，regard ChatGPT as education、Medical、Psychological、Spiritual confusion or instructors，It is inevitable to amplify the hazards caused by knowledge generation errors and inaccurate。

Second is the Turing trap caused by blind intelligence and automation。If you can't recognize the genetic artificial intelligence built on the intelligent ecosystem of human -computer interaction, and even bet365 live casino games above the intelligent technology social system that spreads all over the earth，The foundation and premise of the automation of knowledge generation are not the extraction of human intelligence，Its operation process is both a redefine of knowledge and intelligence，It is also an ecological environment for the earth、Reconstruction of human society and individuals。If you lack reflection on this process，It may be caught in various Turing traps: use automation to generate knowledge in education and research，Use automation to replace human intelligence with automation at work，Excessive automation knowledge generation will also make the ecological environment of the earth unsatisfactory ... The reason why there is a Turing trap，Because the intelligent and automated system does not achieve people -oriented，In the deployment of artificial intelligence, people often force people to passively adapt to intelligence and automation，The operational preset of the intelligent system is not the sexualization of the robot，but make people more and more machine，Make people's autonomy inadvertently deprived。

In order to overcome artificial intelligence, the authoritative illusion，Beyond Tu Ling Trap，Need to start further discussion in the whole society，To build a reasonable imagination of our acceptable deep intelligence future。In terms of concept，The question that must be directly faced is: Can human beings maintain the status of the main creators and leaders in terms of knowledge and intelligence? Does the main performance of a person stand the challenge of artificial intelligence?，We seem to persist: on the one hand，People should become the final knowledge authority; on the other hand，Artificial intelligence should be more intelligent assistant as a person，instead of blindly replacing people's work and alternative skills with intelligence and automation。From the perspective of long -term risks，The emergence of powerful functions of ChatGPT indicates，In the bet365 best casino games research and development of large models，Must be alert to the possibility of consciously universal artificial intelligence，The safety risks that artificial intelligence may threaten human survival in time will be prevented as a prerequisite for development。

　(The author is a researcher at the Institute of Philosophy of the Chinese Academy of Social Sciences)