After ChatGPT came into the public eye, the AIGC industry ushered in an explosion, especially last month, as if every day could be a "historic" day.
Now there are many excellent creators summarizing and analyzing on major websites, and they are all good articles worth reading. Today, this article also shared my views on ChatGPT. I talked about technology and chatted about ChatGPT . After reading this article, I believe you will have a deeper understanding of ChatGPT.
What is GPT?
The GPT model is a generative pre-training model, the full name is Generative Pre-Training , GPT uses specific downstream tasks for fine-tuning (classification tasks, natural language reasoning, semantic similarity, question answering and common sense reasoning), in the pre-training stage, first based on the huge Train a generative language model on the raw corpus; in the fine-tuning stage, continue to train the model using labeled data.
With this divergence, there are actually many tasks that GPT can perform, such as: text generation, code generation, question and answer, etc. If you want to involve other types of data, it is not impossible, such as image generation, video generation, etc. Who would Aren't they high-dimensional vectors?
ChatGPT Chat
This time ChatGPT directly released a dialog box for everyone to play, men, women and children can participate, and the effect is not bad, so it will become popular all of a sudden. First of all, ChatGPT is not an algorithm, but more like a set of solutions, that is, an organic system that integrates multiple methods. The model traceability is basically built on the basis of previous research.
The journey from GPT-3 to ChatGPT
Last month, Fudan University released the news of the first "conversational large-scale language model MOSS" in China. A large number of visits once caused the server to be overloaded. Recently, the entire technology circle has been involved.
-
Stanford releases LLaMA model
-
Tsinghua released the ChatGLM-6B model
-
OpenAI releases GPT4 model
-
Google gives PaLM API
-
Microsoft Releases Microsoft 365 Copilot
-
Baidu releases Wen Xin Yi Yan
Many people who eat melons will ask: Why is it that OpenAI is popular and not the first-tier domestic manufacturers?
Every step that OpenAI is taking now is because of every step it took before.
Since the GPT model was proposed, OpenAI has been continuously optimizing the model. The GPT3 model is the cornerstone of ChatGPT. It has 175 billion parameters and uses 45TB of data for training. The training cost is as high as more than 12 million US dollars .
The essence of GPT3 is to learn a large amount of data through a large number of parameters. It has three very important capabilities: language generation, context learning, and world knowledge. Guaranteeing the coherence of a long article or a book generated, even the CEO of OpenAI said that GPT3 will make stupid mistakes, and GPT3 is just a very early glimpse of the field of artificial intelligence.
Google released the "generalist" model PaLM-E, which can not only understand images, but also understand and generate language, and execute various complex robot instructions without retraining.
OpenAI teamed up with Microsoft to release GPT4, which can perform a large number of tasks and perform at a human level on various professional and academic exams. In the simulated bar exam, GPT4 achieved good results in the top 10%. For the SAT test questions of the US college entrance examination, GPT-4 also scored 710 points in reading and writing and 700 points in mathematics.
Immediately afterwards, Microsoft came with the family bucket that had been opened by GPT4, and GPT4 was too good. Microsoft CEO Nadella said that today is a milestone, which means that the way we interact with computers has entered a new stage. From then on, the way we work will change forever and start a new round of productivity explosion.
But ChatGPT still has some criticisms
1) Data problem
The data crawled by ChatGPT is usually collected without consent, which brings a series of responsibility attribution issues.
2) Cost issues
However, GPT3 has injected a boost to the AI field. Code generation is a very profitable market. It can be seen that when the GPT3 model was born, the OpenAI team used it for code generation. However, due to its high cost Computational costs have also sparked some discussion of monopoly:
Small and medium-sized enterprises have no money to do so, and AI giants have formed a technological monopoly on algorithms that require high computing power.
Customers have privatized deployment requirements, but these models are very large and require high resources, and it is currently impossible to achieve privatized deployment.
For enterprises and developers, the cost of trial and error in technology is too high. Both native technology and low-code are tools in the hands of developers. The significance of tool change and reform is very different. Tool change means improvement of production methods and increase of production efficiency. You can also focus on another track - low code, the core logic is to use the code base to quickly copy the existing development samples, and the labor cost in the entire development process is close to zero.
JNPF, based on low-code development technology, adopts two mainstream technologies Java/.Net development, focusing on low-code development, with drag-and-drop code generator, flexible permission configuration, SaaS service, powerful interface docking, you can do whatever you want Changing workflow engine. It supports multi-terminal collaborative operation, provides 100% source code, and supports multiple cloud environment deployments and local deployments.
Based on the code generator, it can develop multi-terminal applications for Web, Android, IOS, and WeChat applets in one stop. After the code is automatically generated, it can be downloaded locally for secondary development, effectively improving the overall development efficiency.
Open source entry: https://www.yinmaisoft.com/?from=csdn
It has covered mainstream industries such as retail, medical care, manufacturing, banking, construction, education, and social governance. One-stop construction: production management system, project management system, invoicing management system, OA office system, personnel and finance, etc. It can save 80% of the time and cost of developers, and has the functions needed to build business processes, logic and data models.
Finally, the people who eat melons are anxious about whether they will be replaced by artificial intelligence. LeCun, one of the deep learning giants, is anxious: Calm down, it’s still too early, let the bullets fly for a while.