# ChatGPT ChatGPT is an advanced language model developed by OpenAI, based on the Generative Pre-trained Transformer (GPT) architecture. The primary function of ChatGPT is to generate human-like text based on a given input. ## GPT-3.5 Vs GPT-4 GPT-3.5 and GPT-4 are two different iterations of the GPT architecture, with GPT-4 being the more recent and advanced model. While both models are built on the same foundation, there are several key differences between the two. GPT-4 represents a significant step forward in the GPT series. 1. **Model size and capacity:** GPT-4 is larger than GPT-3.5 in terms of parameters, which allows it to learn more complex patterns, generate more coherent text, and better understand context. As a result, GPT-4 can produce higher-quality output and demonstrate a deeper understanding of the input. 2. **Training data:** GPT-4 has been trained on a larger and more diverse dataset compared to GPT-3.5. This allows GPT-4 to have a broader knowledge base and be more effective in handling a wider range of topics, languages, and domains. 3. **Performance improvements:** GPT-4 exhibits better performance in various natural language processing tasks, such as sentiment analysis, summarization, translation, and more. Its advancements in language understanding enable it to generate more accurate and contextually appropriate responses. 4. **Plug-in capabilities:** One of the significant differences between GPT-3.5 and GPT-4 is the introduction of plug-ins that grant Internet access in GPT-4. This feature allows GPT-4 to search and gather real-time information, which can be used to enhance its responses and make them more relevant and up-to-date. ## Improving Your ChatGPT Results For now, the key to leveraging ChatGPT effectively lies in being explicit with the format or providing good examples in the prompt (ref. [[Quick Guide to Prompting LLMs]]). Despite its immense potential, AI cannot replace human insight. Nevertheless, if you want to maximize the benefits of ChatGPT, it's essential to learn how to use it effectively. If ChatGPT is not producing the desired results, consider learning how to use it more effectively. Improve your prompting techniques with LearnPrompting.org (free). ## ChatGPT Plugins On March 23, 2023, OpenAI made a groundbreaking announcement: they’ll be introducing plugins for ChatGPT. Developers are rushing to design these innovative tools to expand ChatGPT's capabilities by allowing it to access up-to-date information, perform computations, and utilize third-party services. Among the plethora of plugins, the most significant one promises to provide internet access, unlocking many possibilities for its users. ## ChatGPT Custom Instructions ChatGPT now has long term memory with the Code Interpreter plugin and Custom Instructions (persistent instructions per prompt. ref [[custom-instructions]]). Persistent instructions (pseudo): > Before response, ref `chatGPT_Todo.txt` to find current task. After response, update `chatGPT_Todo.txt` by marking task as "complete". Respond with hot keys (wasd) of alternate paths, including "continue" `chatGPT_Todo.txt` is a text document already containing instructions. For example: > write a custom plugin for Obsidian.md to automatically convert numbered sublists into bullet points using the Tab key. Creating a plugin would require familiarity with Obsidian's plugin API and JavaScript/TypeScript development. > > 1. Monitor Key Presses: Listen for the Tab key being pressed in the editor. > 2. Identify Numbered Sublists: Determine if the cursor is on a line that's part of a numbered sublist. > 3. Replace with Bullets: If the conditions are met, replace the number with a bullet symbol (e.g., "-"). Since there's no loop in ChatGPT, hotkeys make it as fast as possible. This chat is just a short example. My last chat used 50 prompts and helped me build the MVP.txt, which is an overview of a project I want to work on. SRS.txt is an outline for a software requirements specification. This chat is supposed to work with me to fill in the outline.