Microsoft Envision and Microsoft Build AI Day Review

Date published: October 20, 2023

I attended the Microsoft Envision and Microsoft Build AI Day on October 18th-19th, 2021, which was held at ExCeL London. During the Microsoft Envision event, the keynote discussed how businesses can start their AI journey and how to use AI to transform their business. I had the opportunity to see Satya Nadella, CEO of Microsoft, and other speakers discuss the future of technology and its impact on our lives. I also had the chance to meet some of my Kainos colleagues and other professionals from the industry.

It was a great opportunity to network and learn about the latest technology trends. However, this event was more focused on business and technology leaders, which gave me a chance to learn more about the business side of technology.

Microsoft Build AI Day 2023

On the day, Scott Hanselman, at Microsoft, gave a keynote speech on the future of AI and how it will impact our lives. He also discussed the ethical use of AI and how it can be used to improve our lives. He addressed the common fears surrounding this evolving technology, drawing on an analogy from his own life.

The discussion then turned to the ethical use of AI, with Scott explaining AI as a soft puppet sock, capable of great help or potential harm depending on how it is used.

• Higher temperatures lead to more random and creative outputs, while lower temperatures will have more predictable outcomes.
• Tokens are the building blocks of AI understanding, and AI processes numbers, not words.
• Metaprompt and grounding at the application layer ensure user-friendly experiences, with Bing serving as an example with its three-section approach for reference filtering and interrogation.
• OpenAI is incredibly potent: Azure OpenAI Service provides REST API access to OpenAI’s powerful language models including the GPT-4, GPT-3.5-Turbo, and Embeddings model series.
• Azure Copilot & GitHub Copilot - Enhance your productivity including script/test case writing, code review assistance, code improvement suggestions, code explanations, language translation, etc.

MitigationLayers

Model: LLM (Large Language Model)

• Safety system prompt: safety check to help stop hate and violence and stay polite
• Metaprompt & Grounding: the end-user prompt
• User experience: interaction with the end-user
• Jailbreaking and Responsible AI - The concept of “jailbreaking” was introduced, reminding users to respect the rules set in place.

Sessions and Workshops

• Getting Started with Generative AI using Azure OpenAI Service

This session talks about how language models work with natural language, tokens, models, probability distribution, and the natural language output. It also shows how the language model works and how you can use Azure OpenAI for function calling by giving it a meta prompt, function, and a prompt. We were also shown a demo on how easily you can add data to your OpenAI from various sources and some information on vector-based-retrieval.

• Building and using AI Models responsibly

This is second session I went to that talks about the importance of AI ethics and how to use AI in a responsible way such as the Microsoft Principle AI and Azure AI Content Safety that is currently on preview, whilst recaping the standard goals and the our responsibile AI journey. A demo was shown on how to use Azure AI Content Safety to detect and remove inappropriate content from your application.

• Build and Maintain your company Copilot with Azure ML and GPT 4

This session talks about how to build a company copilot using Azure ML and GPT 4. It utilizes a language calculator behind the scenes. Azure Machine Learning Prompt Flow serves as a development tool specifically crafted to simplify the entire AI application development cycle, particularly for applications powered by Large Language Models (LLMs). This session also includes a demo that shows us how to use Azure ML Prompt Flow to improve your productivity when building OpenAI.

Tips

• Prompt - thin declarative wrapper for your calls
• Groundedness Test evaluation - searching on simulates - chuck data into small pieces
• Don’t treat it like a database
• Deploy by turning on data monitor
• Test by using Prompt Flow
• Prompt flow extension on Visual Studio Code

Workshop

• Build flexible, intelligent apps and OSS on Microsoft Azure and Github Copilot

This is the only workshop I attended during the event. It was a hands-on workshop that allowed me to build a flexible, intelligent app using Azure and Github Copilot. It was a great opportunity to learn more about Azure and Github Copilot. However, I did not manage to complete the workshop. I will try to complete it in my own time. The labs can be found here: https://github.com/Azure-Samples/PetSpotR. I felt like the workshop was too short as it was only for an hour, but it was a great experience to learn more about Azure and Github Copilot.

Conclusion

The rest of the day after the workshop, I was catching up with some people and managed to get a Microsoft Build Socks and some cool stickers. I would recommend attending this event if you are interested in learning more about AI and how it can be used to improve our lives by making us more productive. I will definitely attend this event again next year if I got the chance again.