Generative AI is no longer a buzzword or “tech for tech’s sake.” Today, small and large companies across industries are using generative AI to create value for their employees and consumers. This has led to new methods including quick engineering, retrieval augmented generation, and fine-tuning to help enterprises use generative AI for their particular use cases and data. Innovation occurs along the value chain, from new foundation models and GPUs to unique applications of extant capabilities like vector similarity search and MLOps for generative AI. These fast expanding methods and technology will help enterprises improve generative AI application efficiency, accuracy, and safety. That means everyone can be more productive and creative!
Generative AI inspires new audiences to work on AI initiatives. Software developers who once thought AI and machine learning were for data scientists are selecting, customizing, evaluating, and deploying foundation models. Also, many company leaders feel a feeling of urgency to ramp up on AI technology to better grasp the opportunities, limitations, and hazards. This growth in addressable audiences excites Microsoft Azure and drives us to offer more integrated and personalized experiences that make responsible AI accessible to all skillsets. It also reminds us to invest in education so that all our clients can safely and ethically benefit from generative AI, regardless of their AI journey.
This month’s interesting news focuses on giving developers and data science teams more generative AI model options and application customization freedom. In the spirit of education, check out these core learning resources:
For Business leaders
AI Success Foundation: A Leader’s Guide: Learn from Microsoft, our customers and partners, industry analysts, and AI leaders to help your company succeed in AI transformation.
Change your business with Microsoft AI: This 1.5-hour learning route gives business leaders the skills and resources to implement AI. It examines responsible AI project planning, strategy, and scale.
Career Essentials in Generative AI: This 4-hour course covers AI fundamentals, generative AI capabilities, how to use it in your daily work, and responsible AI.
For builders
Introduction to generative AI: This 1-hour course explains LLMs, Azure OpenAI Service, and responsible AI solution planning.
Start Building AI Plugins With Semantic Kernel: Beginners will learn about Microsoft’s open-source orchestrator, Semantic Kernel, prompts, semantic functions, and vector databases in this 1-hour session.
Working with Azure Machine Learning generative AI models: In this 1-hour intermediate session, you’ll learn about the Transformer architecture and how to fine-tune a foundation model using Azure Machine Learning’s model catalog.
New, strong speech and vision foundation models in Azure AI
Azure are always searching for methods to help machine learning professionals and developers find, tweak, and integrate huge pre-trained AI models. A common hub for exploring Hugging Face, Meta, and Azure OpenAI Service foundation models was launched in May as a public preview. In another milestone, the Azure AI model library released a variety of new open-source vision models for image classification, object recognition, and picture segmentation this month. These models let developers simply add sophisticated, pre-trained vision models to their apps for predictive maintenance, smart retail store solutions, autonomous vehicles, and other computer vision situations.
Azure announced in July that Azure AI services would include OpenAI’s Whisper concept.Azure launched Whisper in Azure OpenAI Service and Azure AI Speech in public preview this month. Whisper transcribes audio in 57 languages. The foundation model can translate all those languages to English and provide transcripts with improved readability, complementing Azure AI strengths. Customers may quickly and accurately transcribe huge amounts of audio content using Whisper and the Azure AI Speech batch transcription API. We hope customers will use Whisper to make information more accessible.
Apply code-first experiences and model monitoring for generative AI to application development
MLOps for LLMs, or “LLMOps,” will help enterprises realize the full promise of generative AI as adoption accelerates and develops. At Microsoft Build 2023, we announced quick flow features in Azure Machine Learning to create, experiment, evaluate, and deploy LLM processes faster. This month, we previewed a code-first prompt flow experience in our SDK, CLI, and VS Code extension. Generative AI teams may now more easily use quick testing, optimization, and version control to move from ideation through experimentation to production-ready systems.
Deploying your LLM application in production isn’t the end. Data and user behavior can affect your application over time, resulting in obsolete AI systems that hurt business outcomes and put enterprises at regulatory and reputational risk. Azure Machine Learning previews model monitoring for generative AI applications this month. Users may now collect production data, analyze key safety, quality, and token usage metrics recurringly, receive crucial issue warnings, and view the results on a beautiful dashboard.
Enterprise search is changing with Azure Cognitive Search and Azure OpenAI Service
Microsoft Bing is changing how people find relevant web content. Bing will now intelligently evaluate your question and find the finest solutions from around the internet instead of displaying a long list of links. Additionally, the search engine provides clear, concise information with trustworthy data source connections. This search experience change simplifies and speeds up internet browsing.
If organizations could search, navigate, and analyze internal data as easily and efficiently, it would alter them. This new paradigm would let employees quickly access company knowledge and leverage enterprise data. This is Retrieval Augmented Generation architecture. With Azure Cognitive Search and Azure OpenAI Service, enterprises can streamline this experience.
Improve generative AI using Hybrid Retrieval and Semantic Ranking
Microsoft found that a combination of the following search techniques creates the most effective retrieval engine for most customer scenarios, especially in the context of generative AI, after extensive testing on representative customer indexes and popular academic benchmarks:
- Chopping long content
- Combining BM25 and vector search for hybrid retrieval
- Activating semantic ranking
Developers of generative AI apps could try hybrid retrieval and reranking algorithms to improve results and please users.
Azure Cosmos DB vector search boosts Azure OpenAI Service application efficiency
Azure added sample code to their documentation and tutorials to help users understand Azure Cosmos DB and Azure OpenAI Service’s potential. Azure Cosmos DB vector search lets you maintain long-term memory and chat history in Azure OpenAI apps, enhancing user experience and LLM solution quality. Vector search lets you token-efficiently query the most appropriate context to personalize Azure OpenAI prompts. Storing vector embeddings with data in an integrated system reduces data synchronization and speeds AI app development.
[…] the “GitHub of AI,” Hugging Face An apt parallel. Developers may share and analyse code, AI models, and datasets on GitHub and Hugging […]
[…] Qualcomm is attached for AI performance right now, according to […]
[…] Generative AI could change our lives, employment, banks, and investments. It may have as much impact as the internet or mobile device. Indeed, 82% of companies exploring or employing generative AI believe it would revolutionise their sector. […]
[…] In 2023, Google introduced AI Hyper computer to train and serve generative AI models, Generative AI support in Vertex AI, Google Enterprise AI platform, Duet AI in Google Workspace, and Duet AI for […]