However, a lot of what happens with AI chatbots depends on a person's capacity to think of questions that the AI will answer. The new application Auto-GPT, which enables AI to behave autonomously and "self-prompt," is fundamentally altering the way we view this technology.
What is Auto-GPT?
On March 30, 2023, a programmer by the name of Significant Gravitas published the open-source Python programme Auto-GPT on GitHub. The application allows the AI to perform "autonomously" without requiring human input by using GPT-4 as its foundation. You can give Auto-GPT a broad objective, and it will take steps to achieve that objective step by step. It is from this that the idea of "AI agents" is derived, which are capable of browsing the internet and carrying out tasks on a PC entirely on their own without the need for constant prompting.
Auto-GPT is used in the straightforward example provided in the original GitHub with the task of searching the web for distinctive and original recipes for "the next upcoming event," such as Easter. The web search engine Chef-GPT then begins looking for a solution. The second objective was to store the recipe on the user's computer as a file.
That might not sound particularly inventive on its own. However, Auto-GPT goes far beyond a straightforward chatbot thanks to its capacity to conduct online searches on the user's behalf and carry out tasks like storing files.
How does Auto-GPT work?
The way the AI's steps are broken down in Auto-GPT is fascinating, and this is where GPT's top-notch text production shines. They are referred to as "thoughts," "reasoning," and "criticism" by Auto-GPT, which makes it clear what the AI is doing and why. Chef-GPT's initial "thought" in the aforementioned scenario was, "I will search for upcoming events to find a suitable one for creating a unique recipe." In other words, the "reasoning" is that "Finding an upcoming event will help me come up with a relevant and exciting recipe."
Auto-GPT's "criticism" will outline some of the issues and restrictions with what it is accomplishing. As you can see, Auto-GPT is acting entirely independently to carry out the user-specified objectives.
The integration of text-to-speech via ElevenLabs and long/short-term memory are a couple of other cool features of Auto-GPT. The combination of all these characteristics gives Auto-GPT the impression that it was designed to interact with people.
Use cases for Auto-GPT
Auto-GPT has a wide range of potential applications, and we're just getting started. Anyone can use the tool because it is totally open-source. The purpose of "Ecommerce-GPT," a straightforward example that was put on Twitter, was to independently create and manage an online store with the intention of boosting net worth.
In the field of coding, there was another intriguing example. One Twitter user suggested "Robo-GPT," which is tasked with examining, modifying, and saving code.
There are countless further instances, and it's simple to see how this could develop into bots building websites, managing social media campaigns, and much more.
Additionally, competing systems that carry out comparable tasks have been developed. These include BabyAGI and Microsoft Jarvis, both of which let GPT "self-prompt" and behave independently.
How to use Auto-GPT
Like many GitHub projects, setting up Auto-GPT requires more than just downloading a file or visiting a website. Before you begin, there are a few necessary prerequisites, including Python 3.8 (or later), an OpenAI API key, and a Pinecone API key. Additionally, if you want the optional text-to-speech feature, you'll require an ElevenLabs API.
Links to those and other crucial details can be found on the Auto-GPT GitHub page. Once you've completed those three steps, click "Code" to get the Zip file. As an alternative, you can use the Git programme to access the files.
To clone the repository, first launch a command-line application like PowerShell by typing "git clone https://github.com/Torantulino/Auto-GPT.git".
To access the project directory, put "cd 'Auto-GPT'" into PowerShell in step two. To install the necessary dependencies, type "pip install -r requirements.txt" after that. Finally, you must add your OpenAI API key and rename the file ".env.template" to ".env".
It's really easy to use Auto-GPT once it's installed. It will first ask you to give the bot a name before asking you to give it a purpose. Even illustrations of both are provided to point you in the proper direction.
Has Auto-GPT achieved AGI?
The first example of AGI (Artificial General Intelligence) cited by many AI enthusiasts is Auto-GPT. Naturally, this is due to the fact that Auto-GPT exhibits the capacity for thought and executes several independent actions in the direction of achieving objectives. Auto-GPT has permanency thanks to the incorporation of long- and short-term memory, which enables it to pick up new information.
Many will argue that a set of interconnected prompts doesn't qualify as "intelligent," while others contend that much of human intelligence and behaviour behaves similarly.
Auto-GPT undoubtedly raises some philosophical issues concerning the future of the "intelligent beings" living and functioning on the internet, regardless of whether it represents the first steps towards AGI or simply a highly helpful next step in ordinary AI.