Microsoft announced a series of new features at the Inspire 2023 conference. Among them are three new AI features worth exploring.
Prior to these features, the Redmond-based company released a large number of Azure-based AI solutions this year.
John Montgomery, Vice President of Microsoft's AI platform company, told ZDNET, "More and more companies are looking for tools to customize their AI to fit the needs and budgets of their organizations. Our goal is to make AI innovation accessible to any organization, such as building chatbots using organizational data, transforming search and document intelligence to gain valuable insights, and make better business decisions."
Azure AI Vector Search Preview
As we learn more about large language models (LLMs) like ChatGPT, one of the most amazing (and controversial) features is their ability to find and describe data based on queries or prompts. Analyzing large datasets requires powerful search capabilities.
Most people are familiar with traditional keyword search, which uses words or phrases as lookup keys for result data. However, AI engineers have been researching a different type of search called vector search.
Vector search can search not only text but also media types such as images, videos, and audio. This chart provides a quick overview of the differences between keyword search and vector search:
Vector search doesn't use keywords but creates a numerical representation of the data called vector embeddings.
Vector search is part of Azure Cognitive Search, which has many search mechanisms. Search applications can invoke a mixed set of these search options to perform large-scale data searches, providing rich search capabilities.
Today, at Microsoft's Inspire 2023 conference, Microsoft announced that vector search is now in public preview, and developers can start integrating this feature into their projects.
Azure AI Document Intelligence Preview
AI chatbots like ChatGPT do offer significant productivity benefits. However, once someone tries to use them in an enterprise environment, they quickly hit a wall.
While LLMs provide a wealth of knowledge analysis (along with some randomly generated filler), they lack the ability to handle company-specific information.
The lack of this capability is driven by three key factors:
1. Data in public LLMs, especially ChatGPT, has a time limit. In other words, ChatGPT knows nothing about events after 2021.
2. Public LLMs have no knowledge of your company's information, whether it's the content of user manuals and maintenance manuals or financial and strategic planning documents.
3. There is no way to insert deep company information into the public LLM's dataset because you certainly don't want all proprietary information to be public.
However, combining the capabilities of generative AI chatbots with deep enterprise data and documents has enormous potential. Imagine interacting with your proprietary user manuals or being able to ask questions based on meeting records from years ago. The potential for discovery and cross-departmental knowledge integration is enormous.
This is where document generation AI comes in, which is a collaboration between Azure AI Document Intelligence (formerly Azure Form Recognizer) and Azure OpenAI services. This feature was also just announced at Inspire 2023, supporting various document types, OCR (with AI-driven error correction), and information extraction.
Most importantly, Microsoft and Azure bring security to all cloud-based products, so you can upload your proprietary data and ensure that it remains locked and under your control.
In addition, the true key to Azure-based solutions is scalability. Individual documents can be small or large, and document libraries can be enterprise-level.
Whisper Model Preview
The third major AI announcement that caught our attention at Azure AI is the new Whisper Model, which can handle enterprise-scale audio transcription. This is also a collaboration with OpenAI, providing OpenAI services on top of Azure.
The Whisper Model can understand 57 languages. It supports what Microsoft calls "enhanced readability," which basically means that AI can understand what is said in context and create transcriptions with more colloquial language when appropriate.
One important feature here is scalability. Whisper mode on Azure can scale to transcribe hundreds or even thousands of documents. For example, it can transcribe thousands of customer service conversations and index them using some of the tools we discussed earlier.
This also ties into customization and Azure integration with the service. As we've seen, Azure AI offers a wide range of products, and the new Whisper Model is integrated into these products, allowing users to customize and build personalized applications according to their needs.
It's worth noting that IBM offers a speech-to-text solution for Watson, and Google has an AI-driven speech-to-text solution, as does Amazon.
The real difference with Azure/OpenAI products is their tight integration with other Azure solutions and inclusion of Azure security features, making it a very natural component to add alongside other aspects of custom applications.
The Whisper Model was announced at Inspire 2023. Once the preview version is available, users can apply for access to Azure OpenAI services, which will open the door to testing Whisper.
Inspire 2023 brings more AI
Microsoft's Montgomery said, "From our supercomputing infrastructure and trusted cloud capabilities to cutting-edge AI models and powerful responsible AI tools, we will continue to focus on making Azure the best place to train, deploy, and scale AI models, whether it's leading technology or openness."
In this article, we focused on three AI technologies announced at Inspire 2023. Microsoft has clearly invested a lot of time and resources in providing AI tools for its Azure clients.
Source: https://www.zdnet.com/article/microsoft-announces-azure-ai-trio-at-inspire-2023-ai-vector-search-al-document-intelligence-and-whisper-model/#ftag=RSSbaffb68
Alphabet Q1 paid subscribers reached 350M, fueled by YouTube & Google One. YouTube ad revenue slightly missed at $9.88B amid Premium growth. Total Q1 revenue hit $109.9B, beating forecasts.
Anthropic releases Claude connectors for Adobe, Blender, and Ableton, enabling AI workflow automation and app control. It also sponsors the Blender Development Foundation.
Apple introduces an App Store model splitting 12-month plans into monthly installments. On iOS 26.4+, it lowers upfront costs without reducing total price, adding billing tracking.
Google launches Aadhaar digital IDs in Wallet for India, enabling secure QR/NFC verification. A phased partner rollout targets key sectors, while passport-based credentials expand globally.
YouTube tests "Ask YouTube," an AI tool blending text and video for complex queries. Initially for US Premium users, it supports interactive follow-ups, expanding Google's AI search.
Google grants US DoD classified AI access, joining OpenAI and xAI in defense deals amid Anthropic's dispute. Contractual safeguards against surveillance and lethal weapons remain unenforceable.
Mistral AI launches Workflows, a Temporal-powered orchestration platform for secure, Python-driven enterprise AI automation featuring human-in-the-loop approvals.