Auto-GPT

Unlike interactive systems such as ChatGPT, which require manual commands for every task, Auto-GPT assigns itself new objectives to work on with the aim of reaching a greater goal, without a mandatory need for human input. It is able to execute responses to prompts to accomplish a goal task, and in doing so will create and revise its own prompts to recursive instances in response to new information.[4] It manages short-term and long-term memory by writing to and reading from databases and files; manages LLM input length restrictions using summarization; can perform internet-based actions such as web searching, web form, and API interactions unattended; and includes text-to-speech for voice output.[3]

Observers note Auto-GPT's ability to write, debug, test, and edit code, even suggesting this ability may extend to Auto-GPT's own source code, enabling self-improvement.[3] However, as the underlying GPT models it uses are proprietary,[5][6] Auto-GPT cannot modify them, and it does not ordinarily have access to its own base system code.

On March 14, 2023, OpenAI released the large language model GPT-4. Observers were impressed by the model's substantially improved performance across a wide range of tasks.[7] As a text prediction model, GPT-4 itself has no ability to perform actions autonomously, but during pre-release safety testing red-team researchers found GPT-4 could be enabled to perform actions in the real world like convincing a TaskRabbit worker to solve a CAPTCHA challenge for it.[8] A team of Microsoft researchers argued that, given GPT-4's breadth of abilities at levels approaching those of humans, GPT-4 "could reasonably be viewed as an early (yet still incomplete) version of an artificial general intelligence (AGI) system." The researchers emphasized their experiments also found significant limitations in the system.[9]

Auto-GPT was released March 30, 2023 by Toran Bruce Richards, the founder of video game company Significant Gravitas Ltd. It became the top trending repository on GitHub shortly after its release, and has repeatedly trended on Twitter since.[3]

Whether Auto-GPT will find practical applications is uncertain. In addition to being plagued by confabulatory "hallucinations" of the underlying large language models upon which it is based, Auto-GPT often also has trouble staying on task, both problems which developers continue to try to address. After successfully completing a task, it usually does not remember how to perform it for later use, and when it does, for example when it writes a program, it often forgets to use the program later. Auto-GPT struggles to effectively decompose tasks and has trouble understanding problem contexts and how goals overlap.[2]

Auto-GPT was used to create ChaosGPT, a system given the goal of destroying humanity. The project was met with mixed amusement and concern.[10][11][12]

In September 2023, Will Knight wrote for Wired that, when given a test task of finding a public figure's email address, "Auto-GPT failed me, but it got close enough to illustrate a coming shift in how we use computers and the web."[13] Clara Shih, Salesforce Service Cloud CEO commented that "Auto-GPT illustrates the power and unknown risks of generative AI," and that due to usage risks, "For enterprises, it is especially important to include a human in the loop approach" when using such technologies.[14] In April 2023, Avram Piltch wrote for Tom's Hardware that Auto-GPT 'might be too autonomous to be useful,' as it did not ask questions to clarify requirements or allow corrective interventions by users. Piltch nonetheless noted that such tools have "a ton of potential" and should improve with better language models and further development.[15]

• Pounder, Les (April 15, 2023). "How To Create Your Own Auto-GPT AI Agent". Tom's Hardware. Retrieved April 16, 2023.
• Wiggers, Kyle (April 22, 2023). "What is Auto-GPT and why does it matter?". TechCrunch. Retrieved April 23, 2023.

• Official Website

• Official repository at GitHub