AI Agent 为什么是AIGC最后的杀手锏？

Author Hu Xiaomeng Chen Chuyi Tencent Research Institute is undoubtedly the most exciting development thread of the current big model. It is called the big model, the next war, the last killer product, and the opening of the new industrial revolution era. The first developer conference detonated and released the initial products and launched the corresponding production tools. Users can generate exclusive products only by describing the desired functions through chatting, which is more suitable for daily life, specific tasks, work or family. For this reason, a large number of new products have been opened. Visual images, voice and the newly launched Bill Gates, which allows developers to develop their own exclusive products more conveniently, recently published an article clearly stating that every user will have an exclusive user within the year. He only needs to tell him what he wants to do in everyday language, and after the release, it has accumulated more than one in a week. So what is so important that the industry has such a high degree of attention and even scholars. It is asserted that the development of American agent stores will continuously widen the gap between China and the United States. What is it in the field of computer artificial intelligence? It is generally translated as an agent. It is defined as a software or hardware entity that embodies one or more intelligent characteristics in a certain environment, such as autonomy, responsiveness, social anticipation, speculation, critical thinking and cognition. It is defined as a large language model driven by the brain, with the ability to independently understand, perceive, plan, remember and use tools, and can automatically perform complex tasks. The basic framework of the service system is as follows. Based on the basic framework of driving, it has four main modules: memory, planning, action and using tools. The memory module is responsible for storing information, including knowledge learned from past interactions and even temporary task information. For an agent, an effective memory mechanism can ensure that it can call past experience and knowledge when facing new or complex situations. For example, a chat robot with memory function can remember the user's preferences or previous conversations. So as to provide a more personalized and coherent communication experience, which is divided into short-term memory and long-term memory. All context learning uses short-term memory to learn long-term memory, which provides agents with the ability to retain and recall infinite information for a long time, usually by using external vector databases and fast retrieval, such as a large amount of data and knowledge deposited in a certain industry field. With long-term memory, a lot of data can be accumulated, making the usability of agents more powerful and deeper in the industry. The planning module has two stages: pre-planning and post-reflection. In the pre-planning stage, it involves the prediction and decision-making of future actions. For example, when performing complex tasks, the agent breaks down the big goal into smaller manageable sub-goals, so that it can efficiently plan a series of steps or actions to achieve the expected results. In the post-reflection stage, the agent has the ability to check and improve the shortcomings in the planning, reflect on the mistakes and learn lessons. After forming and adding long-term memory, it helps agents avoid mistakes and update their cognition of the world. Tools using modules means that agents can use external resources or tools to perform tasks, such as learning to call the outside to obtain additional information missing from the model weight, including the current information code execution ability and access to proprietary information sources, so as to make up for their weaknesses. For example, if the training data is not updated in real time, tools can be used to access the Internet to obtain the latest information. Or use specific software to analyze a large amount of data. Now there are a large number of digital and intelligent tools in the market. Agents use tools more conveniently and efficiently than humans. This way of using tools also represents an important feature and advantage of agents. The action module is the part that agents actually implement decisions or respond to. For different tasks, agent systems have a complete set of action strategies when making decisions. You can choose the actions you need to perform, such as the well-known memory retrieval, reasoning, learning and programming. Generally speaking, these four modules cooperate with each other to enable agents to take actions and make decisions in a wider range of situations, and to perform complex tasks in a smarter and more efficient way will bring about a wider range of man-machine integration. Based on the large model, the exclusive intelligent assistant will not only enhance everyone's ability, but also change the mode of man-machine cooperation and bring about a wider range of man-machine integration and generative intelligent revolution. Up to now, there have been three modes of human-computer collaboration, namely, the embedded mode, in which users set goals by communicating with language and using prompts, and then assist users to accomplish these goals, such as ordinary users inputting prompts to create novels and music works, etc. In this mode, the role is equivalent to the tool of executing commands, while human beings play the role of decision makers and commanders. In this mode, human beings and partners participate in the workflow together and play their respective roles. In the workflow, from providing suggestions to assisting in the completion of the process, for example, in software development, it can help programmers to write code, detect errors or optimize performance. Human beings and their ability to work together in this process complement each other is more like a knowledgeable partner than a simple tool. In fact, last year, Microsoft introduced the concept of co-pilot for the first time, which is a service to assist developers to write code. With the blessing of a big model, Microsoft ushered in a comprehensive upgrade and launch. And put forward the idea that it is a brand-new way of working, so life also needs the founder of Mobvoi, Li Zhifei, who thinks that the best work of the big model is to be an agent model of human beings. Humans set goals and provide necessary resources, such as computing power, and then undertake most of the work independently. Finally, human beings supervise the process and evaluate the final results. This model fully embodies the interactive autonomy and adaptability of agents, which is close to independent actors, while human beings play more supervisory roles. The role of the evaluator and the role of human beings and collaboration: From the functional analysis of the four main modules of agent memory, planning, action and using tools, it can be seen that the agent mode is undoubtedly more efficient than the embedded mode co-pilot mode, or it will become the main mode of man-machine collaboration in the future. Every ordinary individual may become a super-individual. The super-individual has his own team and automated task workflow. Based on the establishment of more intelligent and automated collaboration with other super-individuals, there are many people in the industry. There are some automatic exploration platforms based on super-individuals. 比特币今日价格行情网_okx交易所app_永续合约_比特币怎么买卖交易_虚拟币交易所平台

注册有任何问题请添加 微信：MVIP619 拉你进入群

文章转载注明 网址：https://netpsp.com/?id=63133