Currently,ChatGPT has achieved remarkable results,But it is far from reaching a level close to the human mind。A rough look,Artificial bet365 champions league technology can be said to be a technology in which humans use mathematical calculations to simulate themselves and other bet365 champions league,Machine bet365 champions league (such as expert systems) originally built using rule-based mathematical models,Then realize machine learning and classification with the help of data processing based on statistical probability,The next step is to try to use supervised/unsupervised learning、Sample pre-training、Fine-tuned alignment、Human-machine calibration and other methods to implement context-aware intelligent systems。The development trends of these three types of artificial bet365 champions league technologies,Continuing the transition from humans to machines and then to humans and machines、Research path of human-computer environment system。From the perspective of human-machine environment system bet365 champions league,The explicit description part of ChatGPT is still at the shallow stage of human interaction with big data,The implicit part has not been reflected yet。Dark knowledge、Analogy、Metaphors, etc. seem loose、The tacit and implicit part that lacks mathematical logic,It is difficult to reason through rational thinking,But these things are precisely an important part of human bet365 champions league。In the field of bet365 champions league research,It is very important to analyze and understand basic concepts,The same should be said for the analysis of ChatGPT。The following is from the data、Reasoning、Reference、Consciousness aspects will be explained separately。
From a data perspective,ChatGPT does not have the essential characteristics of bet365 champions league。Small samples and small data solve big problems,It is the expression of bet365 champions league。In many scenes,The intentions of both parties in the interaction are often in the incomplete dynamic small data with uncertainty,It appears with a small probability and gradually evolves。Human bet365 champions league can take advantage of these small data,From different dimensions、Guess the opponent’s intentions from different angles and different granularities,Thus realizing a confidant (seeing signs)、Interesting time (seize the opportunity)、Flexible (adaptive)。This is completely different from what machine bet365 champions league is good at extracting repeatability from big data、Extraction of verifiable rules。Human bet365 champions league is also good at using singular data beyond statistical probability,And be able to comprehensively extract possible needs intentions from valuable small data,Especially the ability to break the rules,Facts or counterfactuals to achieve cross-domain connections、Speculation of value or anti-value。ChatGPT is a program model of big data + machine learning + fine-tuning transformation + human-machine alignment,Act based on factual generalizations from data,But the action value formed by generalization is completely unknown。Sometimes the behavioral results of this kind of generalization are wrong or even dangerous,For example, various "nonsense" phenomena that appear in conversations。
From the perspective of reasoning logic,ChatGPT does not have the essential characteristics of bet365 champions league。It is a misunderstanding to regard bet365 champions league only as calculation or logic。In fact,Real bet365 champions league not only includes the rational and logical part,Also includes the non/super logical perceptual part。Mathematical tools as the basis of artificial bet365 champions league,Only the part of the logical system based on axioms。The core of ChatGPT is computational bet365 champions league、Data bet365 champions league,His perception、Cognitive functions are on-demand matching combinations of pre-trained text,There are issues such as the property rights of intellectual sources and the risk liability of content generation results。Although ethical threshold constraints are set in the ChatGPT algorithm,However, the misleading harm it may bring cannot be underestimated (especially the multi-source causal explanation of unknown knowledge、Explanation of non-causal correlation)。and,There is a big difference between the "autonomy" of the ChatGPT system and the "autonomy" of humans。The "autonomy" of ChatGPT is the rule or statistical reasoning process of big data in text symbol space and time,is an individual sequential process based on mathematical calculation algorithms;Human “autonomy” is the process of causal inference or inference based on small or no data in a mixed physical/cognitive/information (symbol)/social space and time,This is a group process。Further details,ChatGPT’s calculation is causal reductionist,The knowledge is equivalent to explicit factual knowledge。But the key to bet365 champions league is not just computing power,It is more about the calculation ability with reflection。Calculation is a system theory of co-existence,The knowledge is equivalent implicit value knowledge。Calculation is more powerful than calculation than counterfactual、Anti-value ability。For example,Human “autonomy” often includes the ability to reflect (fact feedback + value feedback)。
From the point of view of reference,ChatGPT does not have the essential characteristics of bet365 champions league。Generative artificial bet365 champions league such as ChatGPT,Unlike most artificial bet365 champions league in the past, it can only analyze existing data,And can create completely new content (such as text、Picture、Video、Music, etc.)。But compared to human bet365 champions league,ChatGPT limitations include: limited common sense and causal reasoning (favoring knowledge over bet365 champions league)、Limited natural language and logical reasoning、Lack of grounding in the real world (no visual input or physical interaction)、Unreliable and unpredictable performance, etc.。One of the important shortcomings,It is the "reference" that cannot realize human beings。Wittgenstein raised the issue of "reference" outside logic in "Philosophical Investigations",Things that are “ineffable” and “should remain silent”。Although Wittgenstein was aware of the shortcomings of human bet365 champions league,But no appropriate solution was proposed。His student Turing thought of a way: if human rational logic and perceptual reference are separated,Then human bet365 champions league can be simulated through a mathematical formal system。In this way,There are rules、Logical fields (such as Go games、Shallow text processing, etc.),Artificial bet365 champions league and machines can reach or surpass human levels。Of course,This kind of simulation will lose a lot of things (such as sensibility、Intuition, etc.)。“Reference” is not only a matter of natural language and mathematical language,It also involves thinking (such as intuition、Cognition) and groups and other "language" issues。
From the perspective of consciousness,ChatGPT does not have the essential characteristics of bet365 champions league。ChatGPT is based on word marking,Realized chapter context、A lot of logical marking,And human-machine marking。The core is the Transformer transformation model that relies on the self-attention mechanism to calculate input and output representations,Achieved faster、Stronger calculations。But it is far from touching the more central issue in bet365 champions league: how to generate consciousness?As Marx pointed out,Language is the same as consciousness,Just because of need,Occurred due to the urgent need to interact with others。Maybe,In a sense,Consciousness is an “interaction”。No matter the inner interaction,It’s still external interaction,They are all calculations of situational awareness facts and values of human-machine environmental systems,Not just a simple fact marking calculation。Knowledge has its scope and context,Leave these contexts and contexts,The connotation and denotation of knowledge will change greatly (may even be the opposite)。We not only need to discover the unknown in the known,Sometimes it is necessary to find the known in the unknown。In these aspects,ChatGPT is basically powerless at present。Go further,Artificial bet365 champions league can perceive the objective fact environment through various sensors、Identification、Reaction etc.,But generating subjective value is a difficult problem that is far from being solved。bet365 champions league and language、Same consciousness,Existing rules、The rational side,There are also irregular rules、The irrational side。Can the finite reflect the infinite、Reflecting the inevitable with what should be、Reflect subjectivity with objectivity,Perhaps it is an important sign to test the bet365 champions league of a system。bet365 champions league cannot be achieved by relying solely on calculation,But you need to join the calculation,It is possible to form an intelligent system of human-machine environment systems with deep situational awareness。
(The author is the director of the bet365 champions league-Computer Interaction and Cognition Laboratory of Beijing University of Posts and Telecommunications、Researcher)
Friendly links:
Website registration number: Beijing Public Network Security bet365 champions league. 11010502030146 Ministry of Industry and Information Technology:
All rights reserved by China Social Sciences Magazine. bet365 champions league reproduction or use without permission is allowed
Chief editor’s email: zzszbj@126.com Contact information of this website: 010-85886809 Address: Floor 11-12, Building 1, bet365 champions league. 15 Guanghua Road, Chaoyang District, Beijing Postal Code: 100026
>