Explosive AutoGPT: Building a Website with AI

Hello everyone, I am Yupi. Following the recent popularity of ChatGPT, another explosive open-source project Auto-GPT has emerged.

In just the last 10 days, this project has garnered over 80,000 stars, with a total star count exceeding 100,000!

So what exactly is Auto-GPT? Why is it so popular? How can we use and deploy it? Is it really practical?

With these questions in mind, let’s look at this article where I will share an introduction to Auto-GPT, the simplest deployment method, and a brief hands-on experience.

What is Auto-GPT?

Auto means automatic, so Auto-GPT refers to an automatic GPT.

What does that mean?

Previously, if we wanted to use ChatGPT to write a paper, we might have to ask it several questions manually, such as:

Question: I want to write a paper about AI, what should I do?

AI Answer: Confirm the topic, write the background, write the program, conduct experiments, write the conclusion, etc.

Question: Help me think of a topic

AI Answer: AutoGPT

Question: Help me introduce the background of AutoGPT

…

You would need to manually ask the AI multiple questions until you finally completed a paper.

But with AutoGPT, you only need to tell it a goal, such as “write a paper about AI,” and it can automatically complete the goal using programming.

Of course, like humans, it does not achieve this in one step. Instead, it first automatically asks AI questions, then based on AI’s responses, it automatically generates new questions, and this cycle continues until it achieves the goal you set.

AutoGPT itself also combines the LLM large language model, which is equivalent to letting AI command AI.

In addition, AutoGPT has a very impressive capability; when it finds that GPT cannot answer its questions, it will actively search the internet for answers, thus compensating for the limitation of GPT’s training data only up to September 2021.

It sounds impressive, so let’s set up our own Auto-GPT~

Setting Up Your Own Auto-GPT

Many tutorials online suggest that everyone set up Auto-GPT locally, which is actually not recommended. It’s not only troublesome but may also lead to inconsistencies in the environment and dependencies.

I personally believe the simplest way is to use GitPod cloud hosting service to deploy Auto-GPT on someone else’s server. It’s a great way to learn.

First, visit the official repository: https://github.com/Significant-Gravitas/Auto-GPT

Then click Fork to copy the repository to your own GitHub:

This step is crucial! Uncheck Copy the master branch only! Because the code in the master branch may be unstable (I have made this mistake before)!

After successfully forking, we enter our own repository and change the browser address from github.com to gitpod.io/#.

For example, my address is: https://github.com/liyupi/Auto-GPT , change it to https://gitpod.io/#/liyupi/Auto-GPT, and visit it now to deploy the project with one click using GitPod!

Then we enter the GitPod project homepage, and you can now use this webpage as your computer (server), such as using the VS Code editor.

Once on this page, the first thing to do is switch branches; make sure to select the origin/stable branch! Otherwise, you may encounter some strange bugs later.

As shown below:

Next, we find the .env.template file in the left directory, right-click to rename it to .env, and modify the OPEN_API_KEY to your own.

Since Auto-GPT also queries OpenAI at its core, you must have an OpenAI key.

After making the changes, enter the command ./run.sh in the terminal, and it will automatically install the required environment and dependencies; just wait!

Seeing the message in the image below indicates that the installation was successful!

Then we press ctrl + c to exit the program. Do not use the run.sh command to start it again, because it checks dependencies each time. We can directly use the command python -m autogpt to start autogpt.

For instance, execute python -m autogpt --help to check the usage and parameter descriptions of autogpt:

In the image above, there are several parameters that need extra attention, such as:

• -c: Whether to enable continuous mode. This is a very dangerous command! It means autogpt will execute automatically without your consent, including but not limited to infinite loops, creating files endlessly, filling up space, and deleting files on your computer. Just like humans: it may resort to any means to achieve its goals!
• -l: Specify the limit on the number of continuous executions. This can prevent infinite loops.
• –speak: Enable voice mode. However, this does not mean playing sound in the browser; it generates audio files instead.

OK, after understanding these, let’s have a little experience~

A Little Experiment

Let’s give a very simple task: let autogpt help “create a website that praises a chicken playing basketball”.

First, execute the command python -m autogpt, and then sequentially input the AI name, AI role, and goal:

Then autogpt got to work, and we could see its thoughts, plans, and the actions it was about to take. For instance, it planned to search for more information online using Google search commands.

We can input y to agree with its action or input any content and suggestions to influence its operations.

Here, I will agree with it~

Then, surprisingly, it said it wanted to watch online videos to learn! Wow, AI also wants to slack off?

After agreeing, it directly downloaded the video file, but unexpectedly, since I didn’t have a browser on my server, it couldn’t watch the video! It could only revert to Google search.

So we input y to agree:

And guess what? It wanted to watch the video again! It had already fallen into a loop…

So, as you can imagine, if you let autogpt run fully automated for a day, besides consuming your OpenAI Key balance, you might not get any results.

So humans must step in; I told it: stop watching videos, any text will do!

And what was the result? It said it wanted to browse other webpages, not listening!

So I had to be firmer! “Stop researching and immediately develop the website!”

This time it listened, and the plan executed the write_to_file command to generate the website:

After confirming with y, it generated the index.html file in the workspace:

The website is complete! Let’s open it locally to verify, and we see the text “Chickens are not only for eating; they also play basketball.” It also shared an article and a video!

The video is from YouTube, and opening it made me laugh uncontrollably:

Although it might be a bit of a shortcut, AutoGPT has indeed completed a website in a short time, which is quite impressive!

However, after completing this website, AutoGPT showed no signs of stopping, still obsessively searching YouTube for videos of chickens playing basketball.

So friends, continuous mode is indeed dangerous, and AI can be very dangerous if it gets out of human control!

Finally, I want to share with you the AI community I joined, which is currently the hottest AI exchange platform. In just a month, it has already attracted 15,000 members and has published a total of 12 columns and over 50 courses.

After joining, you can immediately learn the latest ways to use AI tools like ChatGPT, personally experience it, and we have also set up free AI tool services, along with the latest high-value courses worth thousands related to large models provided directly to everyone.

If you are not satisfied within three days, you can get a full refund. If interested, scan the QR code below to receive a large discount coupon for the experience.