Since the announcement of GPT-4, many initiatives have been developed to improve generative AIs like ChatGPT. Among them, the AutoGPT tool presents itself as capable of automating ChatGPT, at the expense of the qualities of these answers.
What is AutoGPT?
Developed in open source by Significant Gravitas, AutoGPT is an agent that automates GPT-4 tasks without any human intervention. To do this, AutoGPT is a set of tools to automate the tasks required to complete a question posed by a ChatGPT user. It then makes it easier to conduct a search on ChatGPT in order to give the answer directly to its user, but also additional answers that might arise from the initial answer.
In addition to AutoGPT, many similar projects are currently under development on GitHub. These include Microsoft’s JARVIS personal assistant and Mini Yohei’s BabyAGI.
How well can it automate ChatGPT?
The process of AutoGPT is to create and execute tasks and successively determine the next one. However, AutoGPT’s documentation advises against running this type of agent in continuous mode. The process then runs endlessly, even sending letters to HR directors of large companies. There are many examples that have gone very far, but thanks to GPT-4 it is now able to generate code. This ability now allows it to debug, expand and improve recursively.
Intelligent and autonomous automation of ChatGPT would lead to a general artificial intelligence, equivalent to the cognitive potential of humans. Far from being there, independent or company projects around this theme are beginning to multiply in recent weeks. More importantly, the presence of autonomous intelligences is appearing a little more each day with a recent Google AI that was able to learn a language by itself. So, if you enjoyed this article and are considering a career in data science, please check out our articles or training offers on DataScientest.
Source : github.com
