🚀 "Control LLM with Guidance: Microsoft's alternative to LangChain" Microsoft launched a new language to control large language models simply and intuitively. Could this be the React of LLM? I think it's the strongest contender yet. 📚Blog: seeai.hashnode.dev/control-llm-wi…

I love the development experience you get with Guidance. How it adds structure and reusability to prompt template is so nice. --- How to make Custom Search Tool using Microsoft Guidance and OpenAI Blog: seeai.hashnode.dev/how-to-make-cu… OpenAI Microsoft Azure

Did you know there are two ways of accessing ChatGPT API? OpenAI API and also Microsoft Azure OpenAI API service. "What are the differences?" "Which one should I use for production?" 🚀Check out my new blog! seeai.hashnode.dev/azure-openai-s… #microsoft #azure #openai #chatgpt

Did you know the new Function API by OpenAI is a powerful tool for transforming unstructured data into a structured format? 🤖 🚀 Read more about it in my new blog! seeai.hashnode.dev/openai-functio… #OpenAI #gpt #langchain

Find myself often repeating 'delimited with triple backticks' when writing prompts. It started just as a good practice, but I find it that it helps me structure my thoughts effectively, and remind me that LLMs aren't mind readers. To get what we want, we need to ask clearly.

Choosing the right LLM for app development isn't one-size-fits-all. The choice depends on trade-offs like speed, cost, data security, and fine-tuneability. If you're just starting, GPT4 is a great choice.

On the hunt for recommendations for prompt management tools. Even tiny tweaks can make big differences, but it is hard to keep track of the changes history and compare results! Has anybody used tools like comet, mlflow or w&b for tracking? and have any preference?

Working with LLMs, sometimes there's no clear right answer to generate, making evaluations challenging. Validation through another model could help, but at increased costs and latency. For non time-critical applications, patience is a small price to pay for reliable results.

Looking forward to attending the Applied AI event. If you are curious to know how you can use AI to unlock capabilities within your business or simply how to use AI in your day-to-day. This is the perfect event for you! eventbrite.co.uk/e/applied-ai-t…

LLMS has the pink elephant problem. When you're told NOT to think of pink elephants, they become the first thing on your mind. So when you are writing prompts use assertions instead of negation statements e.g., replace 'do not forget your keys’ with ‘remember to bring your keys’.

Zero-shot prompting can work well, but in my opinion it is better to always provide an example in the prompt.

It is impressive how sometimes one example can help the model understand how to perform a task. Even if 1 example is helpful, it is good to consider that when the task is quite complex, it is better to increase the number of examples provided.

Azure has their own version of ChatGPT, this could be a good alternative for companies or teams that are very worried about their privacy of their information github.com/microsoft/azur…

It is so hard to measure if the changes of your prompts are making any effect sometimes. When I want to make small tweaks without modifying my original prompt in my code, it is nice to use openai playground platform.openai.com/playground

It's that exciting time of the year again - Leeds Digital Festival! 🎉 We'd like to introduce our event: "ChatGPT in Action" You'll get insights on building LLM powered applications and how to build your own! Register now: bit.ly/seeai-chatgpt-… #LDF2023

Save your space at our GenAI panel event as part of Leeds Digital Festival 🤖 🕛 Tuesday 26th September 12pm 🏢 Codurance office, Platform New Station Street Leeds LS1 4JB Sign up here: bit.ly/45GzUrm #leedsdigitalfest #genai #paneldiscussion