I had an amazing day at #MSBuildAIDay and the opportunity speak with many people about AI. A keynote opened the event and demonstrated a number of different uses of AI.
At the event I created some artwork using Azure OpenAI DALL-E lab #InventwithAI . There were a few questions to answer via voice input that resulted in your own artwork being created. The questions were: what is you favourite colour; favourite landscape; where do you live; artwork style; favourite technological innovation. The image created is below.
I had the opportunity to be a connection hub expert and proctor in the 2 Prompt Engineering workshops run by Amy Kate Boyd and Henk Boelman on Learn how to use OpenAI models (e.g. ChatGPT) using Azure OpenAI.
It is strange to be talking about traditional AI, of a few years ago, as historic as we move to Generative AI. Responsible AI was described as, about reducing the amount of time people spend doing certain tasks, and not replacing them as people are always required to validate the output.
A few points learned in the workshop
- Generative AI models can generate humanlike text, images, and code
- Generative AI models are stateless: they do not learn, and are constrained by their training data which is frozen at a fixed point in time
- Azure OpenAI Service is a managed service that provides access to state-of-the-art natural language generative AI models, including ChatGPT and GPT-4 from OpenAI with the security and enterprise promise of Azure.
- Azure OpenAI Service provides a simple REST API for accessing these models
- Prompt engineering is a technique for "grounding" generative AI models, and can be used to influence the style of their output, provide factual information, and constrain their behaviour.
With Generative AI systems beginning to disrupt the world, it comes with challenges. Guidelines are put in place in prompts to govern the context of asking questions and receiving relevant results. The term 'Jailbreaking' has sprung up to break the guidelines or restrictions set down in the AI program. Thus there is the need to consider how to make these AI models safe and secure for use.
In the workshops AzureOpenAIService-Workshop Understanding-LLMs the sizes of the models were shown to see the rapid changes over such a short period of time.
The prompt engineering lab was great to be able to get started with immediately.
It is important to think about the business use cases that need to be solved and have a high level understanding of what Generative AI can do to help. A business should look at the benefits of using AI, the ethics of using AI and how that impacts employees and customers.
Terms explained
Generative AI is a type of artificial intelligence technology that can produce various types of content, including text, imagery, audio and synthetic data.
A large language model (LLM) is a deep learning algorithm that can perform a variety of natural language processing (NLP) tasks
Prompt engineering is the practice of designing inputs for generative AI tools that will produce optimal outputs.
Microsoft Tools
A couple of tools that are worth looking at:
Responsible AI Toolbox
Responsible AI is an approach to assessing, developing, and deploying AI systems in a safe, trustworthy and ethical manner. Learn how Microsoft thinks about Responsible AI here.
Hands-on tools for building effective human-AI experiences
The HAX Toolkit is for teams building user-facing AI products. It helps you conceptualize what the AI system will do and how it will behave.
Following the event Microsoft Published
AI Innovations Shine at Microsoft Build: AI Day London
