Did you think Chat GPT was too much? Or was GPT 4 making it all weird? Well, now you have Auto GPT to tackle with. GPT stands for – Generative Pretrained Transformer, a neural network designed to function like a human brain. And added an “automated” or “independent” part to the same, you get “Auto GPT.”
The world learned about “Auto GPT: The latest AI tool” on the 30th of March, 2023, with a user named Significant Gravitas posting a simple video on GitHub.
You have Auto GPT, an open-source application developed by “Toran Bruce Richards”. It uses OpenAI’s LLM (Large Language Models) GPT 4, automating the execution of multiple projects that otherwise require external prompts.
Give a task, and Auto GPT will care for the goal independently. Going by the Twitter results, till now, users have been able to use the same to analyze stocks, create podcasts and automate product reviews.
Auto GPT: The Latest AI Tool’s Functional Backdrop
The AI tool is publicly available on GitHub and would not need you to insert constant user input like Chat GPT and GPT 4.
While earlier, Microsoft’s researchers did explain “how GPT 4 requires an additional prompt, and cannot decide to use external tools on its own.” And how another research team from Oregon University highlighted “that Chat GPT is low on semantic reasoning.” On the other hand, Auto GPT is known for creating “AI agents” that can execute tasks on their own. Thus diving actions into reasoning, criticism, and thoughts.
To be more precise, the AI uses GPT 4 to write its code and then debug, self-improve, and develop the same. Also, it uses GPT 3.5 to store and summarize file types.
As you will notice in GitHub, the general problem described is
- To develop an out-of-the-box and original recipe for a recent event, Easter.
- Also, the same video showcases how the developer then asks the AI tool to enhance the piece of code, evaluate the same, and shut down.
- Moving on, the AI organizes all the goals into a to-do list, notifies the developer of its current position as “reading the code,” and then shuts itself down after writing the outputs and completing them individually.
According to internet sources, “Richard (developer of Auto GPT) explained how pre-existing AI models cannot execute goals independently. And that inspired him to develop Auto GPT, where he initially wanted to receive daily news of AI over email.”
How To Use Auto GPT: The Latest AI Tool
As per the system specifications required to use “Auto GPT: The latest AI tool” , you will have to have
- Python 3.8 or above
- An Open API Key to have access to GPT and
- Pinecode API key
Required links are available on the Auto GPT page of GitHub, and you can click on the “download Zip file” option after arranging the needful. Now, it’s time to proceed with the installation part. Where,
- Open a command line program.
- Create a clone of the repository within the command line. Type “git clone https://github.com/Torantulino/Auto-GPT.git”
- Then type “cd ‘Auto-GPT’ into the command line and navigate to the project directory.
- Once everything is in the correct directory, perform the necessary dependencies after typing “pip install -r requirements.txt” within the command line.
- Now, rename “.env.template” to “.env” and fill in their Open AI API Key.
After you are done with the installation, name your bot and make it do tasks as per need.
Auto GPT: The Latest AI Tool’s Use Cases & Limitations
Now, given all the prerequisites of a fully functional AI, Auto GPT showcases endless opportunities when it comes to the execution part. The inbuilt feature list of the same goes as follows:
- Real-time access to the internet
- Steady managing abilities of both long-term and short-term memories
- Power of GPT-4 for text generation
- GPT 3.5’s backup for file summarization and storage
- DALL-e access for image generation.
According to Twitter users, Auto GPT can develop apps, handle complex topics, and generate a new startup.
However, regarding the limitations, it is highlighted that AutoGPT’s results might not work adequately in real-world/complex business scenarios. Though one can be hopeful that the product is relatively new and, in the future, the world will see upgraded versions.
Another Co-existing Synonymous Tool Like Auto GPT
Yohei Nakajima (developer and venture capital partner at Untapped Capital) has come up with a “task-driven autonomous agent” that uses Pinecone, GPT-4, and a specific framework for developing apps powered by LangChain (LLM).
The system can develop new tasks, complete pre-existing ones and prioritize others in real time. There are separate agents within the program entitled to
- Execute a task
- And complete
All the models listed are still in their initial state and have immense scope for development and further experiments. So one can hope to experience added usability of the same.