Autonomous GPT-4: From ChatGPT to AutoGPT, AgentGPT, BabyAGI, HuggingGPT, and Beyond

Emerging task automation and AI agents with GPT-4 after LangChain and LlamaIndex integration trend

Large Language Models on Fire (Photo courtesy by author, taken at Sedona on 4/9/2023)

The emergence of ChatGPT and LLM technology is revolutionary. These state-of-the-art language models have taken the world by storm, inspiring developers, enthusiasts, and organizations to explore innovative ways of integrating and building upon these cutting-edge models. As a result, platforms like LangChain and LlamaIndex have sprung up to streamline integration and foster the development of new applications.

Building AI is Central, albeit Tiny

As we continue to integrate ChatGPT and LLMs, we are seeing an increasing number of autonomous tasks and agents harnessing the power of GPT-4. These developments are not only enhancing the ability to handle complex tasks integrating different systems but also pushing the boundaries of what we can achieve with autonomous AI.

Here will touch upon a notable open-source but incomplete list of autonomous AI (or GPT-4). Please leave a note in the Respond if missing.

• Auto-GPT
• AgentGPT
• BabyAGI
• Microsoft’s JARVIS, aka HuggingGPT
• Web LLM
• God Mode
• CAMEL

These tools and applications can be broadly classified into Command Line Interface (CLI) and browser-based solutions, though HuggingGPT can support both.

CLI: AutoGPT, BabyAGI

Browser: AgentGPT, God Mode, CAMEL, Web LLM

These innovative platforms are making it easier than ever to access and utilize the power of LLMs, reinventing the way we interact with LLMs. They are like starships growing exponentially.

GitHub Star History (by author)

Auto-GPT

Auto-GPT is phenomenal per the above GitHub star growth, though it is an experimental open-source application showcasing the capabilities of the GPT-4 language model. This program, driven by GPT-4, chains together LLM “thoughts” to autonomously achieve whatever goal you set. As one of the first examples of GPT-4 running fully autonomously, Auto-GPT pushes the boundaries of what is possible with AI.

Please refer to its GitHub and usage below:

GitHub: https://github.com/Significant-Gravitas/Auto-GPT

It is easy to set up and run, and please see a few snapshots below.

Install Auto-GPT (Snapshot courtesy by author)

Run Auto-GPT (Snapshot courtesy by author)

AgentGPT

AgentGPT is an autonomous AI solution on the web. It allows for configuring and deploying autonomous AI agents. Name your own custom AI and have it embark on any goal imaginable. It will attempt to reach the goal by thinking of tasks to do, executing them, and learning from the results.

This platform is currently in beta with progressing features below:

• Long-term memory via a vector DB
• Web browsing capabilities via LangChain
• Interaction with websites and people
• Writing capabilities via a document API
• Saving agent runs
• Users and authentication
• Stripe integration for a lower-limit paid version

GitHub: https://github.com/reworkd/AgentGPT

Website: https://agentgpt.reworkd.ai/

Twitter: https://twitter.com/asimdotshrestha/status/1644883727707959296

BabyAGI

BabyAGI is a pared-down version of the original Task-Driven Autonomous Agent, as posted on Twitter below.

Its main idea is to create tasks based on the result of previous tasks and a predefined objective. The script then uses OpenAI’s language model capabilities to create new tasks based on the objective and Pinecone to store and retrieve task results for context.

GitHub: https://github.com/yoheinakajima/babyagi

Website: http://babyagi.org/

Twitter: https://twitter.com/babyAGI_

HuggingGPT

Microsoft HuggingGPT, aka JARVIS, is a collaborative system, including an LLM as the controller and numerous expert models as collaborative executors (from HuggingFace Hub). The workflow consists of four stages:

• Task Planning: Use ChatGPT to analyze the requests to understand the intention and disassemble it into possible solvable tasks.
• Model Selection: Use ChatGPT to select expert models on Hugging Face based on descriptions.
• Task Execution: Invokes and executes each selected model and returns the results to ChatGPT.
• Response Generation: Finally, use ChatGPT to integrate the prediction of all models and generate responses.

GitHub: https://github.com/microsoft/JARVIS

HF: https://huggingface.co/spaces/microsoft/HuggingGPT

Web LLM

Web LLM brought an LLM and LLM-based chatbot to web browsers, running inside the browser with no server support and accelerated with WebGPU. Technically, Web LLM is not a solution for autonomous AI but a lightweight web chatbot.

GitHub: https://github.com/mlc-ai/web-llm

Website: https://mlc.ai/web-llm/

GodMode

GodMode is another impressive web-based autonomous AI agent. Please see its website below.

Godmode.space

CAMEL

CAMEL stands for Communicative Agents for “Mind” Exploration of Large Scale Language Model Society. It proposes a novel communicative agent framework, role-playing, as an alternative to AutoGPT and AgentGPT.

GitHub: https://github.com/lightaime/camel

Website: http://agents.camel-ai.org/

Project Website: https://www.camel-ai.org/

TL;DR

Integrating ChatGPT and LLMs into a diverse range of applications is just the beginning of an exciting journey into the future of language models. As we continue to explore new ways to harness the power of GPT-4, we can expect even more groundbreaking innovations and advancements in the AI field.