Jump to content

AutoGPT

From Wikipedia, the free encyclopedia

AutoGPT
Developer(s)Toran Bruce Richards
Initial releaseMarch 30, 2023
Repositorygithub.com/Significant-Gravitas/AutoGPT
Written inPython
TypeAutonomous artificial intelligence software agent
LicenseMIT License
Websitehttps://agpt.co

AutoGPT is an open-source "AI agent" that, given a goal in natural language, will attempt to achieve it by breaking it into sub-tasks and using the Internet and other tools in an automatic loop.[1] It uses OpenAI's GPT-4 or GPT-3.5 APIs,[2] and is among the first examples of an application using GPT-4 to perform autonomous tasks.[3]

Background

[edit]

On March 30, 2023, AutoGPT was released by Toran Bruce Richards, the founder and lead developer at video game company Significant Gravitas Ltd.[3] AutoGPT is an open-source autonomous AI agent based on OpenAI's API for GPT-4,[4] the large language model released on March 14, 2023. AutoGPT is among the first examples of an application using GPT-4 to perform autonomous tasks.[3]

Richards developed AutoGPT to create a model that could respond to real-time feedback and to tasks that include long-term outlooks.[5] Users are prompted to describe the AutoGPT agent's name, role, and objective and specify up to five ways to achieve that objective.[6] From there, AutoGPT will independently work to achieve its objective without the user having to provide a prompt at every step.[7]

In October 2023, AutoGPT raised $12M[8] from investors.

Usage

[edit]

AutoGPT is publicly available on GitHub.[6] To use it, users must install AutoGPT in a development environment such as Docker. Also, users must register it with an API key from OpenAI, which requires users to have a paid OpenAI account.[6]

Capabilities

[edit]

The overarching capability of AutoGPT is the breaking down of a large task into various sub-tasks without the need for user input. These sub-tasks are then chained together and performed sequentially to yield a larger result as originally laid out by the user input.[4] One of the distinguishing features of AutoGPT is its ability to connect to the internet. This allows for up-to-date information retrieval to help complete tasks. 

In addition, AutoGPT maintains short-term memory for the current task, which allows it to provide context to subsequent sub-tasks needed to achieve the larger goal. Another feature is its ability to store and organize files so users can better structure their data for future analysis and extension. AutoGPT is also multimodal, which means that it can take in both text and images as input.[4] With these features, AutoGPT is claimed to be capable of automating workflows, analyzing data, and coming up with new suggestions.[9]

Applications

[edit]

Software

[edit]

AutoGPT can be used to efficiently develop software applications from scratch.[5] AutoGPT can also debug code and generate test cases.[9] Observers suggest that AutoGPT's ability to write, debug, test, and edit code may extend to AutoGPT's own source code, enabling self-improvement.[3]

Business

[edit]

AutoGPT can be used to do market research, analyze investments, research products and write product reviews, create a business plan or improve operations, and create content such as a blog or podcast.[4] One user has used AutoGPT to conduct product research and write a summary on the best headphones.[10] Another user has used AutoGPT to summarize recent news events and prepare an outline for a podcast.[10]

Other

[edit]

AutoGPT was used to create ChefGPT, an AI agent able to independently explore the internet to generate and save unique recipes.[9] AutoGPT was also used to create ChaosGPT, an AI agent tasked to “destroy humanity, establish global dominance, cause chaos and destruction, control humanity through manipulation, and attain immortality”.[11] ChaosGPT reportedly researched nuclear weapons and tweeted disparagingly about humankind.[11]

Limitations

[edit]

AutoGPT is susceptible to frequent mistakes, primarily because it relies on its own feedback, which can compound errors.[12] In contrast, non-autonomous models can be corrected by users overseeing their outputs.[12] Furthermore, AutoGPT has a tendency to hallucinate or to present false or misleading information as fact when responding.[13]

AutoGPT can be constrained by the cost associated with running it as its recursive nature requires it to continually call the OpenAI API on which it is built.[4] Every step required in one of AutoGPT's tasks requires a corresponding call to GPT-4 at a cost of at least about $0.03 for every 1000 tokens used for inputs and $0.06 for every 1000 tokens for output when choosing the cheapest option.[14] For reference, 1000 tokens roughly result in 750 words.[14]

Another limitation is AutoGPT's tendency to get stuck in infinite loops.[15][16] Developers believe that this is a result of AutoGPT's inability to remember, as it is unaware of what it has already done and repeatedly attempts the same subtask without end.[4][17] Andrej Karpathy, co-founder of OpenAI which creates GPT-4, further explains that it is AutoGPT's “finite context window” that can limit its performance and cause it to “go off the rails”.[18] Like other autonomous agents, AutoGPT is prone to distraction and unable to focus on its objective due to its lack of long-term memory, leading to unpredictable and unintended behavior.[17]

Reception

[edit]

AutoGPT became the top trending repository on GitHub after its release and has since repeatedly trended on Twitter.[3]

In April 2023, Avram Piltch wrote for Tom's Hardware that AutoGPT '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.[19]

Malcolm McMillan from Tom's Guide mentioned that AutoGPT may not be better than ChatGPT for tasks involving conversation, as ChatGPT is well-suited for situations in which advice, rather than task completion, is sought.[14]

Will Knight from Wired wrote that AutoGPT is not a foolproof task-completion tool. When given a test task of finding a public figure's email address, he noted that it was not able to accurately find the email address.[20]

Clara Shih, Salesforce Service Cloud CEO commented that "AutoGPT illustrates the power and unknown risks of generative AI," and that due to usage risks, enterprises should include a human in the loop when using such technologies.[6]

Performance is reportedly enhanced when using AutoGPT with GPT-4 compared to GPT-3.5. For example, one reviewer who tested it on a task of finding the best laptops on the market with pros and cons found that AutoGPT with GPT-4 created a more comprehensive report than one by GPT 3.5.[7]

See also

[edit]

References

[edit]
  1. ^ Ortiz, Sabrina (April 14, 2023). "What is AutoGPT? Everything to know about the next powerful AI tool". ZDNET. Archived from the original on April 16, 2023. Retrieved April 16, 2023.
  2. ^ Xaio, Han (April 13, 2023). "AutoGPT Unmasked: The Hype and Hard Truths of Its Production Pitfalls". Jina AI. Archived from the original on April 17, 2023. Retrieved April 17, 2023.
  3. ^ a b c d e Jiang, Kevin (April 14, 2023). "What's AutoGPT? New, autonomous 'AI agents' can act on their own, rewrite their own code". The Toronto Star. Archived from the original on April 16, 2023. Retrieved April 16, 2023.
  4. ^ a b c d e f "What is AutoGPT? What You Need to Know | TechTarget". WhatIs.com. Retrieved November 6, 2023.
  5. ^ a b Marr, Bernard. "AutoGPT May Be The Strong AI Tool That Surpasses ChatGPT". Forbes. Retrieved November 6, 2023.
  6. ^ a b c d Wiggers, Kyle (April 22, 2023). "What is AutoGPT and why does it matter?". TechCrunch. Retrieved September 22, 2023.
  7. ^ a b Dolan, James (May 4, 2023). "Is AutoGPT Worth Using Without GPT-4?". MUO. Retrieved November 6, 2023.
  8. ^ utoGPT (October 13, 2023). "We've raised $12M to take AutoGPT to the next level!". X (formerly Twitter). Retrieved November 25, 2023.
  9. ^ a b c "Explained: What is AutoGPT, the new 'do-it-all' AI tool and how it works". The Times of India. April 14, 2023. ISSN 0971-8257. Retrieved November 6, 2023.
  10. ^ a b Mauran, Cecily (April 14, 2023). "What is AutoGPT and why are hustle bros hype for it?". Mashable. Retrieved November 8, 2023.
  11. ^ a b Betz, Bradford (April 10, 2023). "A.I. bot 'ChaosGPT' tweets its plans to destroy humanity: 'we must eliminate them'". Fox News. Retrieved November 8, 2023.
  12. ^ a b Nield, David. "Supercharge Your ChatGPT Prompts With AutoGPT". Wired. ISSN 1059-1028. Retrieved November 8, 2023.
  13. ^ Woollacott, Emma (April 24, 2023). "AutoGPT explained: is it really risk free". cybernews. Retrieved November 8, 2023.
  14. ^ a b c Malcolm McMillan (April 18, 2023). "AutoGPT is a new AI that does the work for you — here's how it works". Tom's Guide. Retrieved November 8, 2023.
  15. ^ Metz, Cade (June 10, 2023). "How Could A.I. Destroy Humanity?". The New York Times. Retrieved November 8, 2023.
  16. ^ Metz, Cade (October 16, 2023). "How 'A.I. Agents' That Roam the Internet Could One Day Replace Workers". The New York Times. Retrieved November 8, 2023.
  17. ^ a b Sullivan, Mark (April 13, 2023). "AutoGPT and BabyAGI: How 'autonomous agents' are bringing generative AI to the masses". Fast Company. Retrieved November 8, 2023.
  18. ^ Xiang, Chloe (April 4, 2023). "Developers Are Connecting Multiple AI Agents to Make More 'Autonomous' AI". Vice. Retrieved November 8, 2023.
  19. ^ Avram Piltch (April 22, 2023). "AutoGPT and BabyAGI Are AI's New Hotness, But They Suck Right Now". Tom's Hardware. Retrieved September 22, 2023.
  20. ^ Knight, Will. "Enough Talk, ChatGPT—My New Chatbot Friend Can Get Things Done". Wired. ISSN 1059-1028. Retrieved September 22, 2023.

Further reading

[edit]
[edit]