TWIL: March 6, 2023
Artificial Intelligence hype is at an all-time high, with a lot of people writing about it (even if they don’t always understand it) and great repositories of samples and accelerators to get you started. Check out the AI Solution Accelerators repo and the Prompt Engineering Guide, because they are great. But there are also really interesting news on OpenAI’s ChatGPT and Whisper APIs, as well as Microsoft’s new Kosmos-1 multimodal large language model. Have fun!
The Azure Podcast
Episode 453: In the Real World – How Azure Networking Support uses Azure to support Azure customers
Ryan Bostelmann joins the Azure Podcast team to share how the Azure Support function use Azure and Power Platform to be more efficient and deliver a better customer service.
Episode 454: Sustainability and Azure
Our guest this week, Paul Henwood joins Kendall and Russell to talk about many different aspects of sustainability, and how Azure embraces and supports customers with their sustainability goals.
AI Samples & Accelerators
Sample Repository for the Microsoft Cognitive Services Speech SDK
This repository hosts samples that help you to get started with several features of the SDK. In addition more complex scenarios are included to give you a head-start on using speech technology in your application.
Business Process Automation Accelerator
This accelerator provides a no code Studio for users to quickly build complex, multi-stage AI pipelines across multiple Azure AI and ML Services. Users can select, and stack, AI/ML Services from across Azure Cognitive Services (Speech, Language, Form Recognizer, ReadAPI), Azure Machine Learning, and even Hugging Face state-of-the-art models, into a single, fully integrated pipeline. Integration between services is automated by BPA, and once deployed, a web app is created. This customizable UI* provides and drag-n-drop interface for end users to build multi service pipelines. Finally, the user-created pipeline is triggered as soon as the first input file(s) are uploaded, storing the results in a CosmosDB.
The purpose of this repo is to accelerate the deployment of a Python-based Knowledge Mining solution with OpenAI that will ingest a Knowledge Base, generate embeddings using the contents extracted, store them in a vector search engine (Redis), and use that engine to answer queries / questions specific to that Knowledge Base.
AI Solution Accelerators
Developed by the Microsoft AI Rangers Team, the AI Solution Accelerators are repeatable IP meant to provide developers with all the resources needed to quickly build an initial solution. The objective is to jump-start the development efforts and to learn the used technologies in as little time as possible. The AI Solution Accelerators should be considered as templates that are fully customizable to the user’s unique business case.
Image, Text, Video and Audio Search Using Azure Cognitive Search (Vector Search)
The goal of this is to enable search over Text, Images, Videos and Audio using Azure Cognitive Search. The technique was inspired by the following research article, which converts vectors (embeddings) to text which allows the Cognitive Search service to leverage the inverted index to quickly find the most relevant items. For this reason, any model that will convert an object to a vector can be leveraged as long as the number of dimensions in the resulting vector is less than 3000. It also allows users to leverage existing pretrained or fine-tuned models.
Prompt Engineering Guide
Prompt engineering is a relatively new discipline for developing and optimizing prompts to efficiently use language models (LMs) for a wide variety of applications and research topics. Prompt engineering skills help to better understand the capabilities and limitations of large language models (LLMs). Researchers use prompt engineering to improve the capacity of LLMs on a wide range of common and complex tasks such as question answering and arithmetic reasoning. Developers use prompt engineering to design robust and effective prompting techniques that interface with LLMs and other tools.
Explainit is a modern, enterprise-ready business intelligence web application that re-uses existing frameworks to manage and serve dashboard features to machine learning project lifecycle. Explainit helps ML platform teams with DevOps experience monitor productionized batch data. Explainit can also help these teams build towards a explainability/monitoring platform that improves collaboration between engineers and data scientists.
Language Is Not All You Need: Aligning Perception with Language Models
A big convergence of language, multimodal perception, action, and world modeling is a key step toward artificial general intelligence. In this work, we introduce Kosmos-1, a Multimodal Large Language Model (MLLM) that can perceive general modalities, learn in context (i.e., few-shot), and follow instructions (i.e., zero-shot). Specifically, we train Kosmos-1 from scratch on web-scale multimodal corpora, including arbitrarily interleaved text and images, image-caption pairs, and text data. We evaluate various settings, including zero-shot, few-shot, and multimodal chain-of-thought prompting, on a wide range of tasks without any gradient updates or finetuning.
Microsoft’s UniLM GitHub repo has a lot of information about Foundation Models, aka Large-scale Self-supervised Pre-trained models.
Backup and restore in Azure Synapse Dedicated SQL pool
Learn how to use backup and restore in Azure Synapse Dedicated SQL pool. Use dedicated SQL pool restore points to recover or copy your data warehouse to a previous state in the primary region. Use data warehouse geo-redundant backups to restore to a different geographical region.
Azure Storage blob inventory
The Azure Storage blob inventory feature provides an overview of your containers, blobs, snapshots, and blob versions within a storage account. Use the inventory report to understand various attributes of blobs and containers such as your total data size, age, encryption status, immutability policy, and legal hold and so on. The report provides an overview of your data for business and compliance requirements.
ChatGPT Hype is Proof Nobody Really Understands AI
What most people call “Artificial Intelligence” these days is a specific brand of tech called Machine Learning. You basically show a computer a bunch of examples of what you want to achieve and ask it to find a way to produce the same result.
AI-Generated Content is Dumb (and How to Fix It)
It started as a random shower thought that quickly made itself home in my brain and would not let go. As I stood there, letting the water cascade over me, I couldn’t help but wonder: what happens when we rely on AI models exclusively to mass-produce our content? Are we heading into an era of generic, dumb content?
Introducing ChatGPT and Whisper APIs
ChatGPT and Whisper models are now available on our API, giving developers access to cutting-edge language (not just chat!) and speech-to-text capabilities. Through a series of system-wide optimizations, we’ve achieved 90% cost reduction for ChatGPT since December; we’re now passing through those savings to API users. Developers can now use our open-source Whisper large-v2 model in the API with much faster and cost-effective results. ChatGPT API users can expect continuous model improvements and the option to choose dedicated capacity for deeper control over the models. We’ve also listened closely to feedback from our developers and refined our API terms of service to better meet their needs.
Now Microsoft has a new AI model – Kosmos-1
Microsoft has unveiled Kosmos-1, which it describes as a multimodal large language model (MLLM) that can not only respond to language prompts but also visual cues, which can be used for an array of tasks, including image captioning, visual question answering, and more.
Have an awesome week!