It seems like every other day, new developments are being made in artificial intelligence (AI). Last year, OpenAI released ChatGPT, an AI-based chatbot platform that has since taken the industry by storm. The latest AI advancement is Auto-GPT, which utilizes the same fundamental tech behind ChatGPT, namely GPT-4. With Auto-GPT, users can build their own AI agents or even try it out through the web.
Auto-GPT completely flips the dynamic between AI and users. Unlike ChatGPT, which is a two-way exchange between AI and the end user, Auto-GPT can work based on just one prompt from the user. Then, the AI agent will develop a list of tasks to complete, in order to fulfil the user’s request – without extra input. This AI counterpart has access to the internet, short-term and long-term memory, and is able to generate OpenAI’s most advanced text with GPT-4 and store and summarize data with GPT-3.5.
On Auto-GPT’s GitHub, an example is given of an AI agent researching and learning more about itself. It opened Google, obtained information and reviews from the web, and created a summary of the data stored in a text file.
In another research project by Twitter user Sully Omar, Auto-GPT was tasked to provide the top five competitors for waterproof shoes. It was able to go beyond just research and vet reviews that may be “biased or fake.” Auto-GPT has been making waves in the AI industry with its impressive results, leading developers to believe that AI being able to take full responsibility for any task without prompting from the user is on the horizon.
Despite the buzz surrounding this technology, Auto-GPT is still an experimental prototype. The developer has listed some issues to be sorted out, such as the inability to plan, remembering tasks, and the lack of selecting actions with multiple predictions available.
Setting up Auto-GPT requires a Windows PC, an OpenAI API key, a text editor (such as Notepad++), Git, and Python. Although users may have to put in some manual effort, there are also some projects that provide easy-to-access set-ups and don’t even require an API key, such as AgentGPT.
Significant Gravitas is the name behind Auto-GPT. After having worked in development of artificial general intelligence, Toran Bruce Richards established Significant Gravitas, an organization focused on creating systems that can be leveraged on a global scale. NVIDIA AI scientist Jim Fan highlighted the importance of Auto-GPT in the development of AI, but also the need for caution. While Auto-GPT may be an entertaining experiment and make it easy for the public to get hands-on understanding, it is still in its early stages.
