You Can Try Auto-GPT, the Next Generation of ChatGPT, Right Now
If you’re trying to keep up with all the advances in artificial intelligence lately… good luck. Ever since OpenAI ChatGPT exploded at the end of last year , it seems like every minute there are new developments in the field of artificial intelligence. The latest craze is something called Auto-GPT, which uses ChatGPT’s underlying technology ( GPT-4 ) to do things that ChatGPT simply can’t. You can even create your own Auto-GPT AI agent, or try it online yourself.
What is Auto GPT?
Auto-GPT dramatically changes the relationship between AI and the end user (i.e. you). ChatGPT relies on feedback between the AI and the end user: you ask the AI for a request, it returns a result, and you respond with a new request, perhaps based on what the AI gave you. However, Auto-GPT only requires one prompt from you; from there, the AI agent will then generate a list of tasks it thinks it will need to do what you asked it to do without any further input or prompting. According to the developer of Significant Gravitas (Toran Bruce Richards), it essentially ties together the “thoughts” of the LLM (Large Language Model).
Auto-GPT is a complex system made up of several components. It connects to the internet to get certain information and data (which the free version of ChatGPT cannot do), has long-term and short-term memory management features, uses GPT-4 for the most advanced OpenAI text generation, and GPT-3.5 for file storage and summarization. There are a lot of moving parts here, but all of them together give impressive results.
How do people use Auto-GPT
The first example is from the Auto-GPT GitHub site : you can’t fully see all the targets that Auto-GPT demonstrates, but the bottom line is that someone is asking an AI agent to explore and learn more about themselves. It follows suit: opens Google, finds its own GitHub repository, parses it, and compiles a summary of the data into a text file for the demonstrator to view.
Here’s a more practical example : A user wants to find out which headphones are the best on the market. Instead of doing the research themselves, they turn to Auto-GPT and set the AI agent to the following four goals:
- Do some market research on the various headphones on the market today.
- Get the five best headphones and list their pros and cons.
- Include a price for each and save the analysis.
- Once you’re done, stop.
After thinking for a while, the AI agent goes into action, searching the Internet for information and reviews about the headphones. It then produces an easy-to-read text file ranking the best headphones, listing their prices, and highlighting their pros and cons.
A similar challenge was given to Auto-GPT by Twitter user Sally Omar : Omar posed as a shoe company and asked an AI agent to research waterproof shoes and report five top competitors. The bot even had a means to validate reviews, which it took into account in its analysis, as some reviews could be “biased or fake”.
You can find similar examples all over Twitter by searching for “Auto-GPT” or “AGI” (artificial general intelligence). But before we get carried away, it’s worth remembering that Auto-GPT is super new : Significant Gravitas released the project to GitHub at the end of March, and although the hype around this tool has grown significantly since then, it is still completely experimental.
NVIDIA AI scientist Jim Phan emphasizes that while Auto-GPT is a “fun experiment” and something that allows the general public to tinker with it, it’s still a prototype that shouldn’t be mistaken for a revolution.
Even the developer knows that Auto-GPT is still in its early stages and lists the following issues on their GitHub site :
- Not a polished app or product, just an experiment
- May not perform well in complex real-world business scenarios. In fact, if this is indeed the case, please share your results!
- Quite expensive to run, so set and control API key limits with OpenAI!
But I think what makes Auto-GPT cool (or at least the promise of Auto-GPT) is the idea of being able to ask the AI to take on most of the responsibility for any given task. You don’t need to know the right questions or the best hints to get the AI to do what you want. As long as your initial goals are clear, the AI can think of the next steps for you and create things for you that you might not be able to think of on your own. While we may not have reached that point yet, the fact that Auto-GPT was launched almost two weeks after GPT-4 means we have no idea how fast this type of technology will evolve in the future.
How to try Auto-GPT right now
You don’t need to know how to code to create your own AI agent with Auto-GPT, but it helps. You’ll need a computer, an OpenAI API key ( pay-as-you-go plan highly recommended ), a text editor (like Notepad++), Git (or the latest stable release of Auto-GPT) and Python, but there are plenty of other requirements if you want to extend Auto’s capabilities -GPT such as speech integration or alternative memory areas such as Pinecone .
The Auto-GPT GitHub page has an extensive list of instructions for setting up the tool as well as adding these add-ons. Tom’s Hardware also has a great easy setup guide if all you want to do is try the AI agent with Auto-GPT. If you’re building it yourself, keep an eye on token usage – we’re discussing setting limits in our part of the OpenAI API so you don’t accidentally let Auto-GPT burn your credit card balance.
However, you don’t need to create an AI agent yourself if all you want to do is try out Auto-GPT. Some developers have created interfaces for Auto-GPT that are easy to access from your web browser with no programming experience required. Cognosys was free to use until high demand forced developers to require an OpenAI API key for access. AgentGPT is an interesting example where you don’t need an API key, but it limits the number of tasks the AI will generate for itself. However, this will give you an idea of how the process works and you can increase these limits by providing an API key.