Original link: https://mp.weixin.qq.com/s/ui-O4CnT_W51_zqW4krtcQ
The development of artificial intelligence has reached a new stage, in this stage, artificial intelligence can interact with us in depth like human beings. Among them, OpenAI's ChatGPT is a representative model. However, in order to interact happily and effectively with ChatGPT, we need to understand the differences between human-computer interaction and human-human interaction, and then take effective measures against these differences to deal with those special challenges. Only in this way can we happily interact with ChatGPT ChatGPT chat.
First, let's look at the interaction between people, that is, human interaction. The purpose and outcome of human interaction are varied and may be information exchange, relationship building, problem solving, etc. Interpersonal interaction is a complex and in-depth process, including various verbal and non-verbal elements such as language, facial expressions, body language, and emotional colors. Human beings are able to continuously learn and adapt during the interaction process, and can make changes to the environment, situation and human response. In the process of human interaction, emotions are also a very important part. People can perceive and express various complex emotions. At the same time, in interpersonal interactions, if there are misunderstandings or misunderstandings in understanding, people can usually quickly solve them through further communication.
Secondly, let's understand the interaction between humans and machines, that is, human-computer interaction. The purpose of human-computer interaction is usually more specific and goal-oriented, such as executing commands, obtaining information, completing tasks, etc. Human-computer interaction usually involves communication between humans and machines through some kind of interface, which may include text, images, voice and other forms. For ChatGPT, only text is currently available. Machine understanding has so far been limited. They can understand and respond to specific commands or queries, but often fail to understand vague, ambiguous, or unpredictable human input. Machines require explicit programming or powerful machine learning algorithms to learn and adapt within a limited range. In human-computer interaction, although there are some advanced systems that can recognize and simulate simple emotional states, this ability is usually extremely limited. In the process of human-computer interaction, the ability to deal with errors or ambiguities may be limited, which needs to be considered at design time.
Therefore, although ChatGPT has performed very well in understanding human language and generating human natural language, it is not omnipotent. There are more than 7,000 kinds of human natural languages, which are not only very complex, but also full of ambiguity and ambiguity, which is a huge challenge for the text-based interactive system ChatGPT. For example, dealing with ambiguous or multi-meaning sentences may cause some troubles for ChatGPT. When communicating with ChatGPT, we should use clear and unambiguous sentences as much as possible and avoid words and expressions that may cause confusion.
In addition, the context understanding ability of ChatGPT is also very limited. While it can remember previous input and generate responses based on it, ChatGPT may struggle if a conversation is too long or complex. In order to solve this problem, we should try to keep the conversation coherent, concise and focused on the problem points. Don't expect ChatGPT to fully understand a long conversation or remember all the details.
And, we also have to admit that the content generated by ChatGPT may sometimes deviate from our expectations. Since ChatGPT is a text generated based on statistical methods, it may sometimes produce some inaccurate or irrelevant content. When confronted with this situation, we need to examine it with critical thinking and not simply or blindly accept machine-generated content.
Also, we need to be aware that the interaction with ChatGPT is entirely text-based, it cannot process graphical or visual information. This means that if the problem we need to deal with involves visual information, ChatGPT may not be of much help.
Finally, due to the limitations of the system implementation, the total number of tokens that ChatGPT can input and output is limited. This means that ChatGPT may not be able to give a complete answer to some long-winded questions. In this case, we can try to break down the problem into smaller parts and ask it step by step.
According to our experience in the process of using ChatGPT for software development, using a structured method to describe problems to ChatGPT can give full play to the role of ChatGPT, so as to obtain more accurate, more efficient, more suitable and more valuable solutions ,See below:
To describe the problem structurally, we should follow the following seven steps:
①Determine the core of the problem (core):
First, clarify the key points of the problem, including the specific problem you want to solve and the desired goal.
② Break down the problem (detailed):
Split the problem into smaller, more manageable parts. This helps to gain a clearer understanding of the various aspects of the problem, and how they are related.
③Provide background information (background):
Give background information and context related to the problem, which helps ChatGPT better understand the actual environment and needs of the problem.
④Set priority (priority):
Determine the priority of each part of the question so that ChatGPT can provide targeted answers according to your needs and concerns.
⑤ Raise specific questions (specific):
In the process of describing the problem, try to use clear and specific language. Avoid vague or ambiguous terms to reduce the possibility of ambiguity and misunderstanding.
⑥ State assumptions or constraints (restrictions):
If the question involves specific assumptions or constraints, please express them clearly. This will help ChatGPT provide solutions that are closer to actual needs.
⑦ Specify the expected output format (output):
Clearly express the form of the answer you want, such as lists, paragraphs, diagrams, etc. This can help ChatGPT better meet your expectations.
In general, although there are some challenges in the process of interacting with ChatGPT, as long as we understand these challenges and adopt appropriate strategies to deal with them, then we can effectively use and even control this powerful tool . Hope these experiences can help you chat with ChatGPT more happily.
Orange is sending books!
Like and comment on this blog, five winners will be randomly selected in the comment area, and technical books will be delivered to home for free. To know more stories about payment, please read "Understanding Payment in a Book"
"Understanding Payment with a Book" makes you the first group of people who understand payment thoroughly! A landmark work in the field of payment, a summary of the 30-year experience of payment leaders in China, the United States, Japan and other four countries, strongly recommended by the executive vice president of China UnionPay, with a 360° interpretation of payment. Preferential book purchase link (35% off): Click to learn
