GitHub has launched a new GPT-4-based open source application project AutoGPT, which has exploded in the developer circle with more than 42k Stars.
AutoGPT can perform tasks independently according to user needs without the user's intervention at all, including daily event analysis, marketing plan writing, code programming, mathematical operations and other tasks. For example, a foreign tester asked AutoGPT to help him create a website, and AutoGPT directly called the React framework to generate a front-end page.
Because ChatGPT, which is also based on GPT-4, relies heavily on people to input prompt words to complete tasks, AutoGPT will hand over the intermediary rounds of questioning to machines, so some people say that AutoGPT will replace ChatGPT.
"AutoGPT is the next frontier of hint engineering," Karpassi said. He believes that a GPT call is like an instruction on a computer that can be chained into a program that uses hints to define I/O devices and tool specifications. Define cognitive loops that paginate data input and output in context and then run.
2. Complete the setup in 30 minutes to create your own AI assistant
Before starting, users need to ensure that they have a basic understanding of terminal commands, and need to complete the following steps: 1. Set up Git; 2. Install Python; 3. Download Docker desktop; 4. Get the OpenAI API key and use the provided link To visit, the link is as follows:
Docker：
https://www.docker.com/products/docker-desktop/
OpenAI API：
https://platform.openai.com/account/api-keys
Python：
https://www.python.org/downloads/
The specific steps of installation are as follows:
1. The user first needs to clone the AutoGPT repository from GitHub
“git clone https://github.com/Torantulino/Auto-GPT”
Navigate to the newly created folder, which contains:
3. Run the following command to install the required Python packages;
4. Make sure Docker is running, no need to download any containers, just make sure the program is active;
5. Before running AutoGPT, execute the command: Python scripts/main.py
For continuous mode, use: Python scripts/main.py—Continuous
6. When setting goals, start testing with simple goals. Too complex goals may cause AI to fail to input anything;
7. AutoGPT can save files, and the user only needs to guide it in the right direction to save and analyze. The biggest feature of AutoGPT is that it can automatically analyze and execute according to task instructions, ask yourself questions and answer them, and the middle link No user involvement is required. If ChatGPT obeys user instructions and does what the user asks it to do, then AutoGPT means "do whatever you think you should do". It is like a human being who can go to Baidu on his own when he encounters a problem he doesn't understand. Users can set overall goals for AutoGPT, and then let it take steps to achieve the goals one by one. This is the origin of the concept of "AI agent", which performs operations completely automatically. Currently AutoGPT has been equipped with functions including: collecting information online; storing information; generating GPT-4 instances for text generation; using GPT-3.5 to summarize information. People are currently exploring various ways to play AutoGPT. A user on Twitter demonstrated a simple example, let AutoGPT develop e-commerce business, it can replace people to automatically browse the Internet to find business ideas, and save their findings to files for future reference.
Materials + 60G introductory to advanced AI resource package (classic papers/books/project courses/learning route outline) + with papers/study/employment/competition guidance + Daniel technical questions and answers
Follow the public H: Gupao AI, reply 211 to get
Contains: deep learning neural network + CV computer vision learning (two major frameworks pytorch/tensorflow + source code courseware notes) + NLP
Applicable people ①Students who are preparing for graduation thesis ②AI algorithm engineers who are planning to change jobs and are looking for a job ③Those who are self-study and ready to switch to the AI field ④Those who want to consolidate core AI knowledge and fill in gaps