OpenAI’s ChatGPT has took on the world by storm with its popularity and effectiveness. Numerous applications have already integrated with ChatGPT to make use of the potential of underlying GPT-4 language model. The potential of autonomous AI using GPT-4 is supposedly realized during the pre-release testing by the researchers at OpenAI. They had realized GPT-4 could be modified in such a way that it could give commands to be performed on behalf of a human user. AutoGPT was released on 30-Mar, 2023 shortly after the launch of GPT-4 and it has got the world divided due to the pros and cons associated with a fully autonomous AI model in place.

What exactly is AutoGPT?

AutoGPT is a python-based application developed by Toran Bruce Richards and it is fully open source with its source code available in github. AutoGPT is basically a GPT-4 based AI model where it doesn’t need a prompt from an external source like human. All that it need is a goal and it will automatically create and execute the steps required to fulfil its goal.

It is fully autonomous i.e., it will be able to write code on its own and then execute it without any external intervention and this repeats itself until the original task is completed. The AI model will analyze the step done and using the feedback data, it will re-align itself for the next task in line.

How it works

AutoGPT as explained above works in recursion, as in it continuously evaluates its previous step and updates itself to do the next steps better.

Risks?

The model is ever self-evolving, and this feature can be extremely dangerous. If you take an example like ChatGPT, this AI model is completely bounded by safety protocols, and it reacts only based on a trigger which could be a question from a human. On the other hand, AutoGPT model works completely beyond any boundaries. The fact that it can continue to re-align itself, learn and upgrade its own base will mean that the model will undergo rigorous changes in quick time and could easily grow beyond the capacity of human control. Further, it is said to stray away from the original task assigned due to the autonomous nature of the model.

As with every AI model, AutoGPT also holds the threat or fear about replacing human jobs and in fact even companies. The technology is very new and even the founders are not very certain about the capabilities or the limitations of this model.

Future?

AutoGPT is yet another exciting technology that has come out from the Artificial Intelligence industry. The model is in its early stages, and we may not be able to judge how the model will turn out to be. However, there is no denying the fact that the underlying GPT-4 model is extremely powerful and the autonomous nature of AutoGPT must be looked upon with a keen interest than fear because of the potential it holds to make our lives better.

Getting Started

You need to have the following pre-requisites to get started:

• Computer with decent configuration
• OpenAI’s GPT-4 API key
• ElevenLabs API key

Once you have the pre-requisites, you can get started with the steps below:

• Set-up python
• Clone the AutoGPT code from the github repo
• Update keys (OpenAI and ElevenLabs)
• Configure AutoGPT (Documented in github readme)
• Run autogpt python module in command prompt.

Hope everyone will make use of AutoGPT and the potential it holds to create wonderful applications in a responsible way and not plan to create an AI that will work towards the  judgement day envisioned in the Terminator movie series 🙂