GPT-4 is the next LLM in OpenAI's GPT series. It improves upon GPT-3 in a number of ways, including better reasoning and longer context length. Currently, we do not directly cover it in this course since 1) not everyone has access to it yet, and 2) most of the prompting strategies we cover in this website work for GPT-4 as well. We encourage you to read the OpenAI page on GPT-4 for more details.
Agents are AI systems that can take actions and perform tasks, which often include the use of tools.
Imagine having a personal assistant that can do anything you want with the power of AI. Auto-GPT is that, an open-source application that uses OpenAI’s GPT-4 to perform autonomous tasks. Created by Toran Bruce Richards, it works by breaking down a given goal into sub-tasks and then using the internet and other tools to complete those sub-tasks. For example, if you ask Auto-GPT to ''write a blog post about the latest news in artificial intelligence,” it will first find the latest news articles about artificial intelligence, read them, and then write a blog post about what it learned. You need to follow the setup instructions to use Auto-GPT. Once you have installed Auto-GPT, you can start using it by typing a goal into the prompt bar. Auto-GPT will then try to achieve that goal.
BabyAGI is an AI-powered task management system created by Yohei Nakajima. The system uses OpenAI and Pinecone APIs to create and execute tasks based on the priority and difficulty of previous tasks and a predefined objective. BabyAGI works by running a loop that does the following steps:
Reworkd’s AgentGPT is a platform that unleashes users to craft and deploy self-reliant AI agents, ranging from customer service to data analysis. It educates an AI agent on a distinct task and then endows it to accomplish a preordained goal. Agents can exploit diverse APIs to access and process real-world information, and they can also converse with humans and other agents. To use AgentGPT, users need to register and select a goal for their agent. Then, they can instruct their agent by supplying it with data and instructions.
Sander Schulhoff is the Founder of Learn Prompting and an ML Researcher at the University of Maryland. He created the first open-source Prompt Engineering guide, reaching 3M+ people and teaching them to use tools like ChatGPT. Sander also led a team behind Prompt Report, the most comprehensive study of prompting ever done, co-authored with researchers from the University of Maryland, OpenAI, Microsoft, Google, Princeton, Stanford, and other leading institutions. This 76-page survey analyzed 1,500+ academic papers and covered 200+ prompting techniques.
OpenAI. (2023). GPT-4 Technical Report. ↩
Significant-Gravitas. (2023). https://news.agpt.co/ ↩
Nakajima, Y. (2023). https://github.com/yoheinakajima/babyagi ↩
Reworkd.ai. (2023). https://github.com/reworkd/AgentGPT ↩